All books

Innovation · 2012

The Idea Factory

Bell Labs and the Great Age of American Innovation

Jon Gertner

21 chapters30 lessonsNotes from the full text
In short

Bell Labs turned hard technical problems into products deployed at massive scale by pairing patient research with development, manufacturing and a mission someone depended on. Its wins and misses, from the transistor and cellular to the Picturephone and the integrated circuit, show how new ideas displace old ones and why invention alone is not innovation.

The notes

Zoom from one line per idea to every case and number

The bottom line

Big breakthroughs come from institutions that combine hard real problems, patient money, and mixed disciplines working side by side. For product and research teams, the test is shipping something at scale that does a job better, cheaper, or both.

Key takeaways

  • Tie research to a large system that keeps throwing up hard, real problems. Good problems, not ideas, are the scarce input, and a growing network never runs out of them. Ch. 3 · System
    • Parts had to last forty years, so engineers built a dropping machine that slammed receivers into cradles tens of thousands of times. Walter Shewhart invented statistical quality control to keep products to spec.
    • In 1937 Kelly told Shockley he wanted the relays out of telephone exchanges, replaced by something electronic. That clear target later pointed toward the solid-state amplifier.
  • Put theorists, experimenters and builders where they cannot avoid each other. Ideas came from collisions across disciplines, and phone calls did not replace proximity. Ch. 5 · Solid State
    • Murray Hill Building 1 connected every wing, ran a 700-foot physics corridor, and kept labs and offices apart so people had to walk past each other. More than eighty industrial labs visited to copy it.
    • With no office free, Bardeen joined in October 1945 and set up in the lab Brattain shared. That accident started the theorist and experimentalist partnership.
  • Manage innovation as a system: long money, a sure use for results, skilled support staff and in-house training. Most attempts at the new fail, so people need years of runway and confidence their work will be used. Ch. 9 · Formula
    • Kelly twice handed the AT&T president his resignation over underfunded work. Both times he got the money.
    • Bell Labs employed thousands of technical assistants, some with only a high school diploma, and started Kelly College in 1948 to train staff beyond graduate school.
  • A working prototype is the start, and innovation means making the thing in quantity and selling it. The first transistor worked but was too fragile and inconsistent to use. Ch. 6 · House of Magic
    • Humidity, a hand wave or a slammed door could change the point-contact transistor's behavior. Jack Morton's team made five thousand working units by summer 1949, and the device was ready for large-scale production only by summer 1951.
    • Gordon Teal worked secretly from 5 p.m. to 3 a.m. on borrowed equipment to pull single germanium crystals. Shockley called the result the essential missing ingredient.
  • Back a new technology only when it does the job better, cheaper, or both, and can be manufactured. Raw performance did not decide which designs won. Ch. 10 · Silicon
    • The first silicon transistor, in January 1954, mattered little because it was too complex to mass produce. The 1955 diffused silicon transistor beat every germanium device and looked easy to make, so the future became silicon.
    • The silicon solar battery was at least fifteen times more efficient than prior converters, yet cost several hundred dollars per watt. In 1956 Chapin put the cost of powering a house at nearly $1.5 million.
  • Crowds and friendly surveys are not demand, and a networked product with few users is worth little to each one. Even the strongest lab shipped a product almost nobody would pay for. Ch. 15 · Mistakes
    • A 1967 study of ninety-nine business employees found many willing to pay more than $50 a month for Picturephone, and Jim Fisk compared skeptics to those who doubted Alexander Graham Bell.
    • By March 1972 Pittsburgh had eight paying customers at $160 a month, and Chicago had 46 at $75.

The very point of new ideas was to make them into new things.

Jon Gertner, Introduction: Wicked Problems

1Some walls fall only to new science, and big organizations can supply it

When a goal stalls because the needed knowledge does not exist yet, more tinkering fails, and a large firm with a long horizon can hire the science to break through.

  • Trial and error carries a product far, but it cannot explain why something works. Ch. 1 · Oil Drops
    • Around 1910, American industry advanced mostly through cut-and-try work by people with little formal training.
    • Edison was the model. He searched systematically until something worked and hired mathematicians for the theory.
    • In 1877 Edison found that a compressed carbon button greatly improved a telephone mouthpiece. He got there only after trying lead, copper, manganese, graphite, osmium, silicon, boron, platinum and many other materials.
    • A decade later he switched to tiny roasted carbon granules made from coal. That change made the telephone truly marketable.
    • Edison was not anti-intellectual. He read widely and kept a large library, but he scorned theory.
  • Some limits are gaps in science, not gaps in engineering. Ch. 1 · Oil Drops
    • Thick copper, loading coils and mechanical repeaters topped out around 1,700 miles, roughly New York to Denver.
    • Mechanical repeaters added distortion when a line used more than a few, so stacking them undid the gain.
    • A coast-to-coast call needed an amplifier that science had not yet produced.
    • In 1909 Jewett and chief engineer John J. Carty debated in San Francisco whether they could link New York to the coast in time for the exposition. They decided the crux was an amplifier.
    • Jewett framed the bet as four questions. Can we build it now? If not, what has science not yet shown? Will science show the way, and in time?
    • The fourth question set the work plan: What must we do to make "possibly" into "probably" in two years?
  • Recruit top research talent and give it one concrete, dated problem. Ch. 1 · Oil Drops
    • In fall 1910 Jewett asked Robert Millikan for his best doctoral students. Harold Arnold accepted.
    • Arnold's sole task was a repeater. His high-vacuum tube carried the transcontinental line, which opened in 1915.
    • The win justified hiring more young scientists, including Kelly in late 1917 at $2,100 a year.
    • Arnold improved Lee De Forest's audion, which reached AT&T in 1912, by finding better materials and a higher vacuum.
    • The tube beat rival repeaters in a Philadelphia listening test, where people heard calls through each design.
    • The exposition was delayed to 1915. The line ran four copper wires on 130,000 poles, and a three-minute call to California cost about $21, about $440 today.
  • A protected monopoly bought the patience to shelter research that cannot be scheduled. Ch. 2 · West to East
    • Vail accepted regulated prices and profits in exchange for a cleared path to near monopoly. That gave the company a decades-long horizon.
    • Bell Labs was spun off in 1925. A small group under Arnold looked five to ten years ahead, on his view that invention cannot be scheduled.
    • Clinton Davisson was the engineers' last resort when they could see a tube fail but not why. He was kept on because those answers had financial value.
    • Bell Telephone Laboratories began January 1, 1925, owned half by AT&T and half by Western Electric, with a budget of about $12 million. About three hundred of two thousand technical experts worked under Arnold in research.
    • Jewett was clear that the lab existed to raise phone revenue. Basic research was tolerated for what it did for applied work.
    • Davisson's own output was considered meager, partly because colleagues kept interrupting him. Kelly still called him perhaps the father of basic research at Bell Labs.
  • Large organizations can drive major innovation, so the startup-only story is too simple. Intro · Wicked Problems
    • The pattern runs in three steps. Grasp a hard problem, gather ideas, then push to a product deployed at massive scale.
    • That pattern fits today's wicked problems, such as information overload, infectious disease and climate change.
    • At its late-1960s peak, Bell Labs employed about fifteen thousand people, including some twelve hundred PhDs.
    • Connecting everyone and their machines took an effort rivaling Apollo and the Manhattan Project in size, expense and duration.
    • The model had flaws. It served a regulated monopoly watched by the government, and it had personality clashes, institutional arrogance and later strategic missteps.

2Tie research to a growing system that never runs out of hard problems

A lab stays inventive when it serves a large, expanding system whose daily demands keep producing worthwhile problems to solve.

  • Scientists were hired to serve the phone system, so they learned the system before doing research. Ch. 2 · West to East
    • Harvey Fletcher held a physics PhD. Even so, he spent his first year at Western Electric climbing poles, installing phones and running switchboards.
    • Later recruits sat through weeks of seminars on how the Bell System worked.
    • The mission was concrete. Any two people anywhere should be able to talk as clearly as face to face, at low cost. Regulation required service to get cheaper as well as better.
    • AT&T publicity later suggested scientists came for the sake of science. That was incidental. They were hired for the system.
    • Jewett was clear that the lab existed to raise phone company revenue, not to add to human knowledge. Basic research was tolerated for what it did for applied work.
    • The Labs stayed healthy only as long as the phone business did. Managers had to track technology, politics and finance together, and keep scientists shielded from all three.
  • Demands for durability pushed invention down to the smallest part. Ch. 3 · System
    • Engineers built test rigs that imitated every kind of wear: a woodpecker machine for finishes, an artificial mouth for handsets, a simulated finger for dialing.
    • They set chemical limits for linemen's two-inch leather belts.
    • The 512A teletypewriter oiler was designed to dispense up to fifteen drops and no more.
    • At an outdoor branch in Chester, New Jersey, lodgepole pine poles were buried ten feet deep. Staff studied their decay for decades and graded stains and fungicides.
    • Gophers, squirrels and termites caused hundreds of thousands of dollars in losses a year. Engineers found that steel tape on cables kept gophers off.
    • Donald Quarles's team found that eight inches of vertical wire spacing prevented abrasion. Company practice had suggested twelve.
    • Each cable carried many calls at different frequencies, separated by quartz filter plates. Only Brazil had crystals of the needed size and quality, so managers built a supply chain to get them.
  • A clear target outcome gives direction even when nobody knows the method. Ch. 3 · System
    • Relays and vacuum tubes were the weakest parts of the system. They were delicate, power hungry, hot and slow, and they wore out.
    • Kelly's goal of replacing them named the result he wanted and left the means open. Shockley remembered that talk all his life as the moment his ambition came loose.
    • On December 29, 1939, Shockley sketched a copper-oxide solid-state amplifier in his notebook, modeled on vacuum tube design. Brattain laughed because he had already tried a variation. He still built units to spec in early 1940. They showed nothing, and the war halted the work.
    • Breakthroughs rarely arrive as one flash. People, ideas, problems and timing converge slowly. Shockley's own team did not see how close they were to transforming technology.
  • Each invention creates new problems, so the list of work renews itself. Ch. 9 · Formula
    • Kelly sorted the system's needs into four kinds. Economic needs meant cheaper calls through multiplexing. Operational needs meant direct dialing. Cultural needs meant a cross-country link or the car phone. Military needs meant radar and gun controllers.
    • A new device often requires further devices before it can be used properly. Novelty breeds more novelty.
    • There is always a larger volume of work that is worth doing than can be done currently.
    • This chain reaction helps competitors just as much, unless the lab belongs to a regulated monopoly.
    • A constant supply of need does not guarantee repeat success. Most attempts at the truly new still fail.

3Engineer collisions between disciplines, theory and practice

Murray Hill Building 1 forced people to walk past each other
Labs and offices sit on opposite sides of one 700-foot corridor with every wing connected, so each trip between desk and bench crosses paths with someone else's, and that is how Bardeen ended up working in Brattain's lab
diagram ·

New ideas spread and combined when the place, the teams and the habits kept pushing theorists, experimenters and builders into each other's work.

  • Design space so people cross paths and teams can resize overnight. Ch. 5 · Solid State
    • Every wing connected, so no department had a fixed boundary on the map.
    • Research and development staff were mixed together, not housed apart.
    • Steel walls moved in six-foot steps, so a team could grow or shrink its space fast.
    • Each module was piped with air, water, steam, gas, vacuum, hydrogen, oxygen, nitrogen and AC and DC power. Any room could host any experiment.
    • The 225-acre purchase was announced in July 1930, stalled in the Depression and revived in 1938. The building cost about $4.1 million, opened in 1942 and held about nine hundred scientists and assistants. Nearby suburbs came to be called 'research row.'
    • The move had scientific reasons, since physics, chemistry and acoustics needed distance from city dirt, noise and vibration. It also suited the executives, because the site was near their homes.
  • Build teams that mix skills and temperaments, and pay whatever it takes for the missing piece. Ch. 5 · Solid State
    • In July 1945 Kelly regrouped research into interdisciplinary teams. He promoted younger solid-state experts like Shockley and demoted older supervisors.
    • Shockley and Fisk judged that only about three people in the country could fill the group's theory gap. Kelly offered Bardeen a salary well above any university's.
    • Theorists predicted on blackboards and experimenters tested on breadboards. The group met at least weekly, sometimes daily.
    • Kelly read the new structure aloud to the supervisors, and it took most of the day. One demoted man wept. Addison White still called it enormously good and brave management. No papers record Kelly's reasons.
    • Bardeen came from the Naval Ordnance Laboratory. The lab he moved into was shared by Brattain, who studied semiconductor surfaces, and Pearson, who studied the interior.
    • The loop did not guarantee results. Shockley's field effect came out at least 1,500 times weaker than predicted, and by January 1946 the group was still groping. Talk carried on at lunch and at Snuffy's.
  • Informal channels carried new physics faster than formal ones. Ch. 3 · System
    • Karl Darrow summarized European advances, quantum mechanics among them, in the Bell System Technical Journal.
    • Brattain brought back Sommerfeld's lectures and taught them to colleagues.
    • Shockley's Thursday group worked through one textbook chapter a week, with members taking turns to lecture.
    • Darrow's trembling hands made him a poor experimenter. Harold Arnold kept him on at a lower salary than engineers got and let him spend summers in Europe.
    • Brattain heard Sommerfeld at a Michigan summer course around his 1931 world tour. He then gave an informal lecture series at West Street.
    • Shockley, Brattain, Fisk, Townes and Woolridge met by invitation at a quarter to five over cafeteria tea and cookies. Then came drinks at Brattain's Village apartment, and they were still back at 8:45 a.m. in shirt and tie. Depression-era short hours also freed young staff to take Columbia courses, mostly with I. I. Rabi.
  • Make expertise reachable and helping a duty, not a favor. Ch. 7 · The Informationist Ch. 9 · Formula Ch. 10 · Silicon
    • A patent study found one shared trait among top inventors. They often ate with Harry Nyquist, who drew them out with good questions.
    • Mathematicians were expected to drop their own work for any good problem. The resident expert could not refuse a junior's question.
    • Staff cooperated across buildings without asking managers. New hires toured labs before choosing a project.
    • Thornton Fry's group became a formal department around 1930 as a consultancy for engineers, physicists and chemists. Henry Pollak summed up its role: Our job was to stick our nose into everybody's business.
    • Supervisors sent stuck newcomers to the author of the definitive book: Shockley on semiconductors, John Tukey on statistics, Shannon on information. Some young staff were terrified to go.
    • The norm had limits. Shannon worked with his door closed and watched the department's 'Convergence in Webster' game from the doorway instead of joining.
    • Pearson noticed that a silicon rectifier was sensitive to light and called his friend Chapin. Chapin needed power for remote phones where batteries failed in humid weather. The partners sat in different buildings and needed no sign-off.
  • Keep asking questions until confident claims show their limits. Ch. 14 · Futures, Real and Imagined
    • Bill Baker sat at lunch with whoever he saw first and asked about their work, their life and their ideas.
    • He folded every conversation into a constantly updated picture of the lab's science.
    • He treated "completely understood" as a signal to start probing.
    • At one monthly meeting, the director of chemical research called a topic completely understood. Baker never challenged the phrase. He asked a question, then a question about the answer, layer by layer, until it was clear the topic was not understood at all.
    • Pollak read this as Baker enforcing a belief that science rests on inquiry, not certainty.
    • The habit had a cost. Baker revealed little of himself, and even close colleagues were unsure what he did or meant.

4Run innovation as a managed system: structure, patient money, people and powerful allies

Innovation can be built on purpose when the organization's structure matches the system it serves, money runs for years, a skilled supporting cast holds the know-how, and powerful allies have a stake in its survival.

  • Give each stage of the pipeline its own job, and keep the handoffs loose. Ch. 9 · Formula
    • Research filled a reservoir of new knowledge, materials and methods.
    • Systems engineering judged whether an application was possible, plausible, necessary and economical. Development engineers then designed the devices.
    • Branch labs sat inside Western Electric factories, so scientists stayed close as their work moved into manufacturing.
    • The transistor was the model case. It left the solid-state team, went into development and manufacture, and systems engineers worked out how it would fit into the phone network.
    • On March 23, 1950, Kelly described the Labs to the Royal Society in London as an "institute of creative technology." He never said "innovation." Jack Morton and others took up that word years later.
    • The three stages were only a sketch. Real handoffs were casual and ran in every direction, and Kelly compared the lab to a living organism.
  • Treat money, hiring and tacit skill as infrastructure, and build them when others cannot. Ch. 2 · West to East Ch. 9 · Formula
    • Kelly hired in the mid-1930s, when universities could offer only half, or at best two-thirds, of the Labs' $3,000 starting salary.
    • Stable funding let the lab plan five, ten and twenty years out. Guaranteed markets kept even pure researchers focused, because good work would surely be used.
    • Training went beyond graduate school through the Communications Development Training Program, which was unaccredited.
    • Kelly personally hired Shockley and Fisk from MIT. Pierce and Woolridge came from Caltech, Baker from Princeton, and Townes joined too. Woolridge called $3,000 fantastic.
    • Physicist John Rowell called the technical assistants keepers of practical information, secrets, tricks and lore. PhDs treated them as top colleagues.
    • Kelly's own rule: Never underestimate the importance of money. He ran the Labs in practice while Oliver Buckley, who had undisclosed Parkinson's disease, held the title until 1951.
    • Caveats: recruits may have come for the salary in a collapsed academic market. The whole formula also relied on monopoly funding and captive markets.
  • Urgency and federal money can compress decades into years. Borrowed technology still needs a full redesign. Ch. 4 · War
    • By mid-1940, about 75 percent of the Labs' work was military. Staff doubled from about 4,600 to 9,000.
    • Existing lab work paid off fast. Synthetic quartz replaced scarce South American supply.
    • Fisk's team found that every magnetron parameter had to be redesigned for each use.
    • Technicians tried about a hundred chemical formulas and grew crystals to six inches from a small seed. The Labs produced hundreds of thousands of them.
    • Fisk, about 30, worked in a former biscuit factory and X-rayed the British device to get blueprints. His team produced fifteen working designs. Western Electric made more than half the radar sets used in the war.
    • Reverse engineering ran against Kelly's preferred order of research first, then development. The labor shortage also opened hiring to women and, for the first time, Jews.
  • Pick leaders for technical judgment and for skill with people, not for brilliance alone. Ch. 4 · War Ch. 20 · Echoes
    • Shockley's operations analysis set depth charges at 30 feet instead of 75. U-boat kills rose fivefold within two months.
    • Yet he struggled to manage a few dozen men he judged less able than himself.
    • In 1997 Pierce listed four conditions. Management must be technically competent all the way to the top. Researchers never raise funds. Work can be supported for years. A line of research can be stopped without damning the researcher.
    • Shockley kept separate calendars for military, scientific and home life. Biographer Joel Shurkin later found an unsent note to his wife saying he had played Russian roulette.
    • Kelly planned ahead while the war was still on. In May 1943 he circulated a 29-page memo on the postwar Labs. In January 1944 Buckley told staff there would be no postwar planning.
  • Make the lab indispensable to the state, and accept the entanglement that comes with it. Ch. 9 · Formula Ch. 14 · Futures, Real and Imagined
    • Bell took on Sandia, the Nike missiles and the DEW line.
    • Those ties helped blunt the 1949 suit to split Western Electric from AT&T. An intact Bell System was worth more to cold war defense.
    • Baker carried Bell's knowledge of how information is coded and moved into Washington's codebreaking and spy satellite work.
    • Sandia had fifteen hundred staff building the non-nuclear parts of atomic weapons. AT&T took it on the condition that it earn no profit, and it became a rotation post for rising managers.
    • Nike's guidance system needed about 1.5 million parts and went into operation in 1953. By 1950, government work took half of Kelly's hours, and his security clearance equaled the CIA head's.
    • Baker's 1957 committee urged solid-state computers for codebreaking. He also helped found the National Reconnaissance Office in 1960.
    • The cost: civilian communications became tied to nuclear weapons, and part of the motive was keeping regulators away.

5Invention is the start; innovation means making it in quantity, reliably, better or cheaper

A working device proves an idea, but value comes only when the thing can be made reliably in volume and beats what it replaces.

  • Discovery, invention and innovation are separate stages, and only the last one pays. Ch. 6 · House of Magic
    • Discovery observes nature. Invention engineers science into something new. Innovation turns that into a product in wide practical use.
    • Karl Jansky's radio astronomy was a scientific triumph that never produced a profitable telecom device.
    • The transistor came from naming the obstacle, Bardeen's surface states, then varying metals, points, spacings and electrolytes until small gains added up.
    • Jack Morton treated discovery, product development, manufacturing technique and market creation as one integrated process.
    • Jansky built a movable antenna at Holmdel in the early 1930s to study atmospheric noise and heard a steady hiss from the Milky Way. Inventions are patentable. Discoveries are not.
    • On November 17, 1947, Brattain and Robert Gibney found an electrolyte that cut through the surface barrier. On December 16, Brattain slit gold foil on a plastic triangle with a razor, wiggled it, and got real gain. Neither man fully understood the physics.
    • Ralph Bown's test: show gain on real speech, feed output back into input, and confirm a steady oscillation. News then climbed one management level at a time, a week or two apart. Kelly heard about a month later.
    • Kelly gave Morton a month to write a development plan while he toured Europe. Morton spent twenty-nine days in terror and settled the plan on the thirtieth. The Bell system could absorb output, so market risk was small and production problems dominated.
  • Materials and process work is the hidden precondition for both mass production and new designs. Ch. 5 · Solid State
    • Permendur diaphragms failed on trace contaminants, so ingredient purity became part of the fix.
    • Ohl's melts produced p-type and n-type silicon long before anyone could explain them.
    • Crystal pulling and Bill Pfann's zone refining made consistent germanium possible Ch. 6 · House of Magic.
    • Plastic cable sheathing saved more than the Labs' research budget for its decade Ch. 14 · Futures, Real and Imagined.
    • Permendur was cobalt and iron with about 2 percent vanadium. Traces of carbon, oxygen or nitrogen spoiled it. Jewett later called the old belief that tiny impurities had no effect absolutely wrong.
    • Ohl found one sample that rectified both ways and gave a large voltage under bright light. He showed it in Kelly's office and no one could explain it. The team could smell phosphorus when sawing n-type silicon, though no instrument detected it.
    • Pfann woke from an after-lunch nap with the idea of sweeping a molten zone along a rod. The result held perhaps one impurity atom per 100 million, like a pinch of salt in a thirty-eight-car train of sugar.
    • A study Baker commissioned put the sheathing savings at about $2.5 billion. Staying with lead would have needed 80 percent of all lead produced in the US.
  • Manufacturability, not peak performance, picks the winning design, and a monopoly can wait to deploy. Ch. 10 · Silicon
    • Germanium is rarer, costlier and often useless above 150 degrees Fahrenheit. Silicon was harder to work.
    • Cal Fuller's diffusion gave precise control of impurity layers, which made a buildable silicon device possible.
    • Kelly's test applied to the phone system too, so electronic switching took nearly twenty years to fully deploy.
    • On March 17, 1955, Tanenbaum was back at the lab because his wife was hosting a bridge game. He melted an aluminum wire straight through the top layer to reach a p-layer finer than a hair. He judged it a cinch to make.
    • Morton cut short a European trip to back the device, and Kelly deferred to him. Fuller's interest began when he noticed germanium could be contaminated by touching a brass doorknob.
    • A switching office with 65,000 relays handled slightly under 1,000 operations a second. Transistors could potentially do a million. Parts already in place were built to last thirty or forty years, so early replacement needed economic or clear technical justification.
  • Where failure is costly, build from proven parts, then squeeze more out of what exists. Ch. 11 · Empire
    • Kelly barred the new transistor from TAT-1 and used late-1930s vacuum tubes.
    • The cable had to run twenty years without failure to pay for itself.
    • TASI later doubled its channels by reusing the pauses in speech.
    • Some tubes had been tested continuously for sixteen years and were built in a clean room at a new Hillside, New Jersey factory. Cheaper radio links were rejected because weather could ruin calls for days.
    • TAT-1 opened September 25, 1956, through fifty-two repeaters. Its technology never failed once in twenty-two years, though trawlers sometimes broke it near shore.
    • A talker sends signals only about 35 percent of the time. TASI checked channels two thousand times a second and reconnected a talker in a millionth of a second, taking capacity from thirty-six to seventy-two channels.
  • Spreading the technology on purpose grew the market and kept the originator ahead. Ch. 6 · House of Magic
    • Bell gave away samples, licensed for $25,000, and taught hundreds at Murray Hill conventions in 1951 and 1952.
    • Outside engineers improved the device, and the originator was best placed to profit from their advances.
    • Secrecy lasted only until patents were filed, under the code name Surface States Phenomena. Licenses were free for hearing aid makers out of respect for Alexander Graham Bell's work with the deaf.
    • Expert users found uses Bell had not planned. MIT's Jay Forrester wrote that transistors might serve high-speed digital computing. The openness was partly forced: hoarding the device risked antitrust scrutiny.

6Geniuses, instigators and managers do different work, and the institution holds them up

Breakthroughs need the lone genius, the instigator who spreads ideas and the manager who keeps a team intact, and each did their best work only while the organization carried them.

  • A rare single mind can found a whole field, and hard practical problems are what feed it. Ch. 7 · The Informationist
    • Shannon's tinkering with the relays of Bush's differential analyzer led him to describe switching circuits with Boolean algebra.
    • Wartime cryptography showed him that English is 75 to 80 percent redundant, so messages can be compressed.
    • PCM was his catalyst for a general theory of bits, channel capacity and error-correcting codes. The engineering itself bored him.
    • The switching thesis, about twenty-five pages, followed a 1937 summer at West Street and won an engineering society prize in 1939.
    • His 1945 classified cryptography paper ran 114 pages. The 1948 communication paper appeared in the Bell System Technical Journal in July and October, and the word "bit" came from John Tukey.
    • The transistor would likely have arrived elsewhere within a couple of years. Shannon was different: he founded a field, stated its main results and proved most of them at once. He still insisted his work did not rank with relativity.
  • Unstructured freedom gave the genius insight and fame, but his output faded. Ch. 8 · Man and Machine
    • His 1949 paper was the first ever on chess programming. It framed chess as a wedge toward call routing, translation and military decisions.
    • That framing reassured managers who feared regulators would see a monopoly paying an eccentric to play games.
    • Theseus, the maze mouse, made him a public figure in a way information theory never had.
    • Theseus stored walls and turns in relays under the maze. After one run it could finish in as little as fifteen seconds from any starting point. Time covered it in 1952.
    • Hagelbarger's guessing machine beat human players about 53 percent of the time. Shannon built a rival with less memory and more speed. An umpire machine scored the match, and Shannon's machine won about 55 to 45.
    • Unfinished papers piled up in boxes, and most letters went unanswered, even a 1951 CIA request routed through Kelly. In 1956 he left for MIT. He felt stale after fifteen years and saw his freedom as a special favor.
  • The instigator matters less for what he finishes than for the ideas he spots, names and pushes onto others. Ch. 12 · An Instigator
    • Pierce found Kompfner's traveling wave tube in a 1943 British report and brought it to Bell Labs.
    • He named the transistor. He also published fast under the pen names John Roberts and j.j. coupling to skip internal clearance.
    • He believed freedom works like food: people need it, but in moderation.
    • In November 1944 he toured twenty British labs in four weeks to find Kompfner. The 1946 launch promised 10,000 simultaneous calls coast to coast, yet perfecting the tube took thirteen more years. His summary: Rudi invented the traveling wave tube and I discovered it.
    • His 1950 tube book drew about one correction per page. He cared more about getting the gist right than about avoiding errors.
    • Most colleagues took his suggestions seriously. When he told Shannon he "should" do something, Shannon asked what "should" meant, and both laughed.
    • His harsh reviews had costs. People feared his office, and one blunt remark sent Chuck Elmendorf to a lowly post.
  • A manager who competes with his own reports can wreck a winning team. Ch. 6 · House of Magic
    • Shockley felt left out of the point-contact invention. Over New Year's Eve he filled 19 notebook pages in 2 days and designed the junction transistor in secret.
    • He revealed it only when John Shive's results made him fear someone else would get there first. This broke the Labs' taboo against competing with subordinates.
    • The design was better, but the team collapsed. Bardeen left for the University of Illinois. Brattain complained to Kelly, who released him from reporting to Shockley.
    • Brattain hinted to Kelly that he could not lie about timing in a patent dispute.
  • Promotion should follow management fit, because a passed-over star may leave and build elsewhere. Ch. 11 · Empire
    • Kelly judged Shockley unfit for upper management and refused to promote him.
    • Shockley sent Kelly peevish memos in 1953 and 1954 and appealed as high as AT&T's president. Then he left for Palo Alto.
    • Kelly introduced him to investors, though probably not out of pure kindness. Shockley offered Morris Tanenbaum double his salary but lured only one person from the Labs.
    • He then hired Gordon Moore, Robert Noyce, Jean Hoerni and Eugene Kleiner.
  • Outside the Labs the stars mostly faded, which suggests the organization made their work possible. Ch. 18 · Afterlives
    • In 1957 Shockley asked staff to take polygraph tests, and Moore and seven others left. His science dried up without supervisors and technicians to offset his flaws.
    • Kelly's skill at judging research carried over to IBM. So did his habit of attacking the biggest unknown first.
    • Pierce found Caltech unmoored. Nobody there deeply cared what a professor did, while at Bell Labs the network needed the work.
    • Kelly's method: ignore what current technology allows, find the missing piece, solve it first, then build the project around it. At IBM he listened to talks with his eyes closed and reported privately to Thomas Watson Jr.
    • Shockley's firm was sold and folded in the late 1960s. Several of the eight he drove out became billionaires.
    • Bardeen is the counterexample. At Illinois he won a second physics Nobel in 1972 for superconductivity.
    • Shannon turned to the stock market and juggling and worried he had nothing left to say. Pierce kept instigating: in 1957 he urged Max Mathews toward computer music. His complaints may partly reflect a grumbling nature.

7Big ideas wait for their parts; prove them cheaply, then hand them to development

A grand idea becomes real only when its separate parts, money and rules line up, and the safe path is a cheap research test before a full development build.

  • An idea can arrive decades before the parts that make it work. Ch. 13 · On Crawford Hill
    • Pierce's satellite drew on parts already made for other jobs: the rugged transistor, the traveling wave tube and Friis's low-noise horn antenna.
    • The silicon solar cell had failed on cost for rural phones. In space nothing else could recharge batteries, so its weakness stopped mattering.
    • Rockets and money came only after Sputnik in October 1957, then Explorer I and the creation of NASA.
    • The maser began with Townes, who left Bell Labs for Columbia and got it working in 1954. By 1957 a Bell team had one built on a fingernail-sized crystal cooled to about minus 456 degrees Fahrenheit.
    • Pierce missed the maser's use for satellites. Kompfner saw it while home sick in bed and rushed to tell him on recovery. Pierce agreed at once it changed everything.
  • Test the core idea with the simplest, cheapest version, even if it can never earn money. Ch. 13 · On Crawford Hill
    • Pierce picked a passive reflector over a better active satellite. Bell could not yet build a durable active one, and research had few people.
    • The goal was to hit practical problems early, a lesson from the transistor, whose troubles surfaced only in production.
    • Friis's check, whether a project is too big a ball of wax, guided the scope. ARPA's ambitious Advent tried too much too soon and failed.
    • Echo cost under $2 million and used about three dozen men.
    • Kelly killed the idea over lunch in summer 1958, backed by a three-page memo on Brock McMillan's numbers. On Bill Baker's likely advice, Pierce never sent his rebuttal.
    • After Kelly retired, a January 6, 1959 proposal asked $1 million. It argued satellites might beat cable on cost and that someone else would act if Bell did not. AT&T approved it as Project Echo.
    • The May 13, 1960 launch failed when the rocket's third stage missed orbit. On August 12, 1960 a recorded Eisenhower message bounced off the 100-foot balloon, and the maser amplified it four thousand times.
    • Pierce admitted he would have flubbed the ground terminal. Bill Jakes ran it, and before launch he cleared everyone from the control shack, Pierce and Kompfner included. They obeyed.
  • Once research proves the idea, a development team builds the durable version from many older inventions. Ch. 13 · On Crawford Hill
    • Telstar combined sixteen Bell inventions patented over twenty-five years. None was made for space.
    • Chaffee's 1937 frequency modulation feedback receiver, thought largely useless, proved vital for Echo and Telstar.
    • Engineers chose the orbit after testing how callers tolerated delay.
    • Geostationary orbit at about 22,300 miles meant about six tenths of a second of delay. Murray Hill subjects found the delay and echo annoying, so Bell kept Telstar at no more than 3,000 miles.
    • Telstar took more than five hundred engineers and fifteen thousand parts, each stressed far past expected loads. NASA charged $3 million to launch it on June 10, 1962.
    • Technical success still lost the business. The Communications Satellite Act of 1962 created COMSAT and effectively shut AT&T out of international satellites. Pierce wondered whether Kelly had been right.
  • Cellular sat shelved for decades until enabling technology made it ripe. Ch. 16 · Competition Ch. 17 · Apart
    • The December 1947 Ring and Young memo laid out a honeycomb of reused frequencies. It went unpublished and served only to support a spectrum petition.
    • In the late 1950s the FCC gave the needed spectrum to UHF television instead.
    • Frenkiel, Porter and later Engel kept drawing hexagons in spare time. Google's 20 percent time echoes this (Ch. 20 · Echoes).
    • Microprocessors and the programmable ESS (Ch. 14 · Futures, Real and Imagined) made locating and handing off moving callers solvable.
    • That UHF block could have yielded eight hundred to a thousand phone channels. Pierce said the FCC had picked mass communication over individual communication.
    • Frenkiel noted locating and hexagon ideas on January 17, 1966, then shelved the plan when the FCC stalled. Engel held that getting ahead meant spending 20 or 30 percent of your time beyond your assignment.
    • The hard part was the logic, not the radio: Cellular is a computer technology. It's not a radio technology. Intel's 4004 held 2,300 transistors.
  • Field data and a paying pilot answered what market studies could not. Ch. 17 · Apart
    • Jakes rode a van for months, and DiPiazza's team ran a gutted trailer overnight, to measure real signal behavior.
    • A fall 1971 study found no market at any price. The team ignored it.
    • Cellular could reach anyone on the existing network from the first day.
    • A 1978 Chicago trial with paying subscribers settled the open questions.
    • Jakes drove through tunnels, hills, woods and near lakes across New York and New Jersey. He called the moving-vehicle radio problems almost beyond imagining.
    • DiPiazza came from radar decoy work on Kwajalein. His team measured suburban noise and range at antenna heights of ten, twenty and fifty feet, and its tower choices became lasting standards.
    • Engel concluded such studies only measure demand for products that already exist. AT&T spent $100 million on cellular before market.
    • The Chicago service had two thousand paid subscribers over two thousand square miles. It showed the FCC how customers rated it and how Motorola handsets worked on Bell's network.

8Where it breaks: misread demand, eclipsing outsiders, and a mission that disappears

The lab failed where received wisdom, sunk bets and friendly signals went unchallenged, and it faded once the monopoly and the mission that funded it were gone.

  • Name the kind of mistake you are making, because each one has a different cause. Ch. 15 · Mistakes
    • Missed opportunity: Bell Labs developed diffusion and photolithography. Kilby at Texas Instruments and Noyce at Fairchild used them to build the integrated circuit.
    • The barrier was belief, not ability. Morton held that chips with many components could never be reliable.
    • Mistake of perception: years and millions went into hollow waveguides. Corning's low-loss glass fiber then made them obsolete.
    • The outsider saw it first. Kao had no sunk investment in waveguides and needed cheap short links between city switching centers.
    • Morton answered the tyranny of numbers with functional devices that did several jobs at once. They proved very hard to engineer. His intolerance of opposition kept the lab from changing course. Kilby had the single-chip idea in summer 1958, probably a few months before Noyce. Both bet that the manufacturing problems could be solved later.
    • Tanenbaum saw an early chip at Noyce's shop. He thought it neat but did not grasp how far it would go.
    • Kao, of International Telephone and Telegraph, visited in 1966. At the time, existing glass lost the signal within a few dozen meters. Bell had a fiber program by 1969, but Baker still championed waveguides in a 1968 speech. Corning announced its fiber in September 1970. By late November Kompfner wrote that fiber would take center stage.
    • Memories of the visit differ. Kao recalled a skeptical welcome. Tingye Li and Ira Jacobs say no one dismissed him, they just expected slow progress. Pure glass coming from Corning, with a century of glass expertise, was understandable.
  • A sound forecast can still back the wrong product or the wrong pathway. Ch. 14 · Futures, Real and Imagined Ch. 15 · Mistakes
    • Fisk read the trends correctly: faster, more digital, more congested, moving to higher frequencies. His bet on millimeter waves in waveguides was the error.
    • Picturephone's warnings came early. Only 12 percent of couples would pay $40 to $60 a month for one at home.
    • The survey asked whether seeing the other person mattered. It never asked whether people wanted to be seen.
    • AT&T quietly surveyed 700 fair users and knew fairgoers were not a cross section. Business users were warmer: 29 percent were interested even at $60 to $80.
    • Pierce warned in the mid-1950s that acceptability had not been evaluated. Molnar still predicted it would be the main way people communicated by 2000.
    • Irwin Dorros blamed groupthink. Rudi Kompfner put a photo of himself in front of his own set. Modern video chat does not redeem the product: being early is not necessarily different from being wrong.
  • The lab's greatest work dissolved the monopoly that paid for it. Ch. 11 · Empire Ch. 16 · Competition
    • The 1956 decree licensed about eighty-six hundred patents royalty-free. Others could copy Bell technology fast.
    • Shannon's digital information let any company code and decode messages. That undercut the case for one owner of the network.
    • Leaders denied the threat. An executive said only AT&T had the network, and Baker called the industry mature.
    • Max Mathews' hindsight view: an analog call passed through many amplifiers, so one entity had to design and control them all. Digital signals removed that need. Cheap transistors made the shift easier for rivals.
    • MCI built nothing to last forty years and subsidized no local service. Its service was sometimes worse but much cheaper. In 1971 the FCC ordered AT&T to connect it.
    • Kelly had warned in 1943 that telecom was converging with fast, competitive electronics. Justice sued on November 20, 1974. Sandia and military work did not buy safety at Justice or the FCC.
    • The January 1982 settlement split off the local companies. AT&T kept long distance, Western Electric and Bell Labs, and was freed from the 1956 decree. Thirty years later Tanenbaum regretted not fighting the trial to the end, for the Labs' sake. Ch. 17 · Apart
  • Without someone depending on its output, a lab keeps its assets and loses its point. Ch. 19 · Inheritance
    • AT&T had never had to sell, so it struggled in open markets. Lab funding tightened as its parent weakened.
    • Odlyzko argued in 1995 that open-ended research no longer paid for a single firm. Payoffs took too long, and others captured the value.
    • Pierce's 1986 verdict was that the lab inherited buildings, equipment and people, but the mission was gone.
    • AT&T's purchase of NCR failed. It lost phones and equipment to cheap Asian rivals. Its products now aged in three or four years instead of decades.
    • Lucent was spun off in 1996 and was worth $98.5 billion two years later, peaking at $270 billion. Its profits had been inflated by financing customers' purchases. The stock fell from about $84 to below $2, and the workforce shrank from 150,000 to about 40,000.
    • In 2002 Schon was found to have faked data. Under Jeong Kim, Nature counted four researchers in basic physics. Holmdel closed in 2006: 1.9 million square feet on 472 acres, unsold for years, watched by one guard.
    • There was no clear turning point. Nobel work went on in the 1980s, including Chu's atom trapping, while insiders felt things were fine.
  • The model depended on its era, so its closest heirs sit outside the market. Ch. 20 · Echoes
    • Monopoly security let Kelly back the transistor, which made his own vacuum tube work obsolete.
    • Pay was compressed: the top officer earned about twelve times the lowest-paid worker. By the late 1990s large-firm CEOs typically earned a hundred times.
    • Venture capital will not fund open-ended science, and tech giants answer to shareholders.
    • In the mid-1960s Baker and a New Jersey consortium hired Terman to design a hub around Bell Labs. New Jersey was too spread out, and the Labs were less fluid than the Valley. His graduate-only Summit University would have cost $15 million to start. Drug companies declined, Baker could not justify paying for it alone, and the plan was shelved.
    • Janelia Farm opened in 2006 on 689 acres. About three hundred researchers and a hundred visiting scholars work there with no teaching, grading or grant chasing.
    • Chu told a Senate committee in 2009 that the Department of Energy must be the modern Bell Labs of energy research. His hubs remain a modest first step in size and scope.
    • In 2006, 77 of the 88 US entities producing significant innovations got federal funding.

9Put it to work

Pick the hardest open problem behind your product, find the piece nobody understands yet, and prove that piece cheaply before you build or sell at scale.

  1. Name the unknown that blocks the goal. Before planning features, ask whether the thing can be built now. If not, ask what knowledge is missing and who could supply it. Ch. 1 · Oil Drops
    • Jewett and Carty decided that a coast-to-coast call came down to one device, an amplifier that science had not yet produced. Jewett asked Millikan for his best doctoral students and gave Harold Arnold that single task.
    • The trap is adding more tinkering when the gap is missing understanding. Mechanical repeaters hit a hard limit that no amount of tweaking could pass.
  2. Check that the enabling parts have arrived. An idea can be right and still early. Keep a shelved plan alive until the technologies it depends on exist. Ch. 17 · Apart
    • The honeycomb of cells was sketched in a December 1947 memo. It became practical only once integrated circuits, the microprocessor and programmable electronic switching were available.
  3. Prove the core idea with the smallest, cheapest build. Run a research experiment whose job is to surface practical problems. Once it works, hand it to a team built for scale and reliability. Ch. 13 · On Crawford Hill
    • Echo, a passive balloon, cost under $2 million and needed about three dozen men. Telstar, the active follow-on, took more than five hundred scientists and engineers and fifteen thousand tested parts.
    • The military's Advent active satellite tried to do too much too soon, and it failed.
  4. Make it safe to doubt the received wisdom. When a senior belief rules out a design, test whether the blocking problem can be solved later before dropping the idea. Ch. 15 · Mistakes
    • Bell Labs had developed diffusion and photolithography. But Jack Morton held that chips with many components could never be reliable, and he did not tolerate opposition. Kilby and Noyce bet the manufacturing problems could be worked out and got the integrated circuit.
  5. Test demand with real prices, not crowds. Lines at a demo show curiosity. Ask what people will pay each month, and ask both sides of a social product what they want. Ch. 14 · Futures, Real and Imagined
    • AT&T quietly surveyed 700 Picturephone users at the 1964 fair. At $40 to $60 a month, only 12 percent of couples wanted one at home.
    • The survey asked whether seeing the other person mattered. It never asked whether people wanted to be seen.
  6. Make the product useful to the very first customer. Prefer designs that plug into an existing network, so value does not wait for others to adopt. Then run a small pilot with paying users. Ch. 17 · Apart
    • A cell phone user could call anyone already on the phone network, which the Picturephone could not offer. A 1971 study found no market for mobile phones at any price. The team ignored it because such studies only measure demand for products that already exist.
    • In summer 1978 a Chicago trial served two thousand paid subscribers to show the FCC how customers judged the service.
  7. Assume your best work will come back as a competitor. Shared or licensed technology gets copied. Cheaper entrants go after your most profitable customers without carrying your obligations. Ch. 16 · Competition
    • An AT&T executive told Bell Labs staff not to worry because only AT&T had the network. MCI built a cheaper microwave network anyway, and in 1971 the FCC ordered AT&T to connect it.

10The book's words

These terms help you name the kind of problem, team and bet you are dealing with.

  • Problem-rich environment A growing system that never runs out of hard, real problems worth solving.
    • Phone parts had to last forty years, so even a teletypewriter oiler was engineered to dispense no more than fifteen drops.
  • Innovation Turning an idea into a product made in quantity and sold, as opposed to a discovery or an invention.
    • Karl Jansky's discovery of radio noise from the Milky Way founded radio astronomy but never yielded a profitable device.
  • Better, or cheaper, or both Kelly's test for whether something new deserves to replace what already works.
    • Transistors were not yet justified this way in the phone system, so electronic switching took nearly twenty years to fully deploy.
  • Too big a ball of wax A check on whether a project is beyond what a team can deliver right now.
    • Harald Friis asked it often. The same caution led Pierce to try a passive satellite before an active one.
  • Steam engine time The moment enough enabling knowledge has come together that an innovation becomes nearly inevitable.
    • Named for James Watt's steam engine. Cellular reached it only once microprocessors and programmable switches existed.
  • Instigator Someone whose main contribution is getting others to work on ideas that had not occurred to them.
    • John Pierce named the transistor and pushed Max Mathews into computer music. "Rudi invented the traveling wave tube and I discovered it."
  • Critical mass Enough top people across disciplines in one place for ideas to collide productively.
    • Borrowed from nuclear physics. Kelly wanted it in every department.
  • Circumscribed freedom Freedom for researchers, bounded by a broad but directed mission.
    • Morry Tanenbaum's phrase. Pierce felt unlimited freedom was disorienting, and that freedom, like food, is best in moderation.
  • Mistake of judgment A product that fails to meet the need you assumed, through poor function, cost or timing.
    • The Picturephone is the example. A mistake of perception is different: the waveguide was eclipsed by glass fiber from Corning.

11Chapter by chapter

Where each idea is in the book.

  • Introduction: Wicked Problems Bell Labs, the research arm of monopoly AT&T, was for decades the world's most innovative scientific organization, and the way it turned hard problems into mass-deployed products still offers lessons for today's near-intractable 'wicked problems'. Large organizations, not only small nimble startups, can produce major innovation.
  • 1. Oil Drops In the early 1900s American industry ran on tinkerers like Edison, but AT&T discovered that hard technical walls, like sending a voice across the continent, could be broken by hiring academic scientists. That bet, made through a friendship between Frank Jewett and Robert Millikan, set the pattern that later brought Mervin Kelly to the phone company.
  • 2. West to East Bell Labs grew out of Western Electric's engineering shop in 1925 as a large, semiautonomous lab where a small research group was given room to pursue fundamental questions, on the bet that unpredictable insight would pay off for the phone business alongside routine engineering. Mervin Kelly learned there that knowing why things work matters, and that the hard-to-make vacuum tube needed a replacement.
  • 3. System Bell Labs hired top scientists to serve the phone system, not science for its own sake, and the system's endless, problem-rich demands, combined with informal channels for spreading new physics, set up the conditions for leaps like the solid-state amplifier.
  • 4. War World War II turned Bell Labs from open research into secret, high-speed development, compressing ten to twenty years of technical progress into four, and it reshaped the people who would lead the postwar electronics revolution. Even mid-war, Kelly was planning how the Labs would lead that revolution rather than join it.
  • 5. Solid State Bell Labs after the war set up the conditions for a breakthrough rather than the breakthrough itself: a building engineered for collisions, an interdisciplinary solid-state group funded for basic knowledge, and a deep base of materials work on purity and controlled impurity. By January 1946 that group still had nothing to show.
  • 6. House of Magic The transistor was discovered and invented at Bell Labs in a few weeks of late 1947, but turning it into an innovation took years of manufacturing, materials work, licensing and diffusion. Invention alone is not innovation: you must make the thing in quantity and sell it.
  • 7. The Informationist Claude Shannon, a solitary Bell Labs mathematician, founded information theory almost single-handedly in 1948, showing that all communication can be measured in bits, bounded by channel capacity, and made nearly error-free with redundancy. His work shows that innovation starts in one mind but is fed by an environment rich in hard problems and good questions.
  • 8. Man and Machine Shannon treated the phone network as a giant computer and used playful machines, chess programs, a learning maze mouse, guessing machines, to probe what computers could do and how minds work. His curiosity-driven freedom produced insight and fame but also drifted away from publishing, and he left Bell Labs for MIT in 1956.
  • 9. Formula Kelly believed innovation could be managed on purpose. His formula was a steady supply of real problems, a critical mass of top people kept physically close, skilled support staff, in-house training, long-term money and guaranteed markets. To protect that machine, he also tied Bell Labs tightly to the military and the state.
  • 10. Silicon A regulated monopoly can afford to be slow in deploying new technology while its lab races ahead in science, but inventions only matter once they do a job better, cheaper, or both. Silicon won the transistor's future through a manufacturable process, while the brilliant silicon solar cell stalled on cost.
  • 11. Empire By the late 1950s Bell Labs under Mervin Kelly had built an empire: a transatlantic phone cable that never failed, a Nobel Prize, and a monopoly secured by the 1956 consent decree. Yet the transistor and Shannon's information theory, its greatest achievements, held the seeds of the monopoly's end.
  • 12. An Instigator An instigator is a rare kind of innovator, distinct from both the genius and the skilled manager. John Pierce mattered at Bell Labs less for what he finished himself than for the ideas he spotted, named, pushed onto others, and framed ahead of their time.
  • 13. On Crawford Hill Big ideas often arrive before their parts exist, and they become real when separate inventions, rockets and money finally line up. Proving such an idea works through a cheap research experiment first, then handing it to development, is what turned Pierce's satellite concept into Echo and Telstar. Even so, engineering success did not guarantee a business.
  • 14. Futures, Real and Imagined Bell Labs in the early 1960s had to bet on a future no one could predict. Its flashiest bet, the Picturephone, drew crowds but weak willingness to pay, while invisible infrastructure like electronic switching, digital transmission and high-frequency pathways quietly built the base for services nobody had yet imagined. Bill Baker, the secretive research chief, linked that technical future to Washington's intelligence apparatus.
  • 15. Mistakes Even the strongest research lab misjudges the future: Bell Labs missed the integrated circuit, bet on waveguides that glass fiber made obsolete, and launched a Picturephone nobody wanted. These were different kinds of mistakes, missed opportunity, perception and judgment, and each has its own causes.
  • 16. Competition In the 1970s AT&T's regulated monopoly came under attack from competitors like MCI and the Justice Department, while Bell Labs, amid the fight, pushed forward optical fiber and mobile telephony, two projects that showed the network was far from finished.
  • 17. Apart By the 1960s and 1970s, big breakthroughs like cellular telephony came from large, mixed teams of systems, research and development engineers. They worked when converging technologies made the idea ripe. Meanwhile, the spread of Bell Labs' own inventions helped erode the monopoly that paid for the lab, and the 1982 breakup left its future unclear.
  • 18. Afterlives Bell Labs' leading figures mostly did their landmark work inside the institution. Outside it, they drifted into consulting, teaching, hobbies or disgrace, which suggests the enveloping organization, more than lone genius, made the breakthroughs possible.
  • 19. Inheritance After the 1984 breakup, Bell Labs did not vanish so much as get split and split again. It survived as a respectable industrial lab but lost the essential mission that had made it great, and so stopped mattering to American technology.
  • 20. Echoes Bell Labs' lasting legacy is ideas, not buildings, and the parts of its formula worth salvaging are patient funding, technically competent leadership, a broad but directed mission, and a critical mass of disciplines working together to understand problems deeply. Today's venture and tech-giant models excel at incremental products but leave a gap for fundamental, risky breakthroughs.

Lessons from The Idea Factory

30 things to do, each with the chapter it comes from. The library's chat and course draw on these.

  1. Require any replacement to do the job better, cheaper, or both

    A switching office with 65,000 relays handled slightly under 1,000 operations a second, while transistors could potentially do a million. The transistor still could not yet be justified on cost, and electronic switching took nearly twenty years to fully deploy.

    10. Silicon
  2. Treat lines at a demo as curiosity and test real willingness to pay

    AT&T surveyed 700 Picturephone users at the 1964 fair. Reaction was favorable, but at $40 to $60 a month only 12 percent of couples wanted one at home, while 29 percent of business users were interested even at $60 to $80.

    14. Futures, Real and Imagined
  3. Ask users whether they want to be seen, not only to see

    Most Picturephone users said seeing the other person was important, but no one asked whether they wanted to be seen. Rudi Kompfner later put a photo of himself in front of his set.

    14. Futures, Real and Imagined
  4. Plan how a networked product reaches critical mass before launch

    By March 1972 Pittsburgh had eight paying Picturephone customers at $160 a month, and Chicago had 46 at $75. Bob Lucky felt he had the last set in the world with no one to call.

    15. Mistakes
  5. Design new products to work with an existing network from day one

    A mobile user could call anyone on the existing phone network, so cellular did not need other users to adopt first. The Picturephone did, and failed.

    17. Apart
  6. Ignore market studies that measure demand for products nobody has imagined

    In fall 1971 a marketing study for AT&T concluded there was no market for mobile phones at any price. The team disagreed and kept going. Joel Engel later concluded such studies only measure demand for products that already exist.

    17. Apart
  7. Run a cheap, simple experiment first to surface practical problems

    Echo, a passive balloon satellite, cost under $2 million with about three dozen men. The active Telstar that followed needed more than five hundred scientists and engineers and fifteen thousand tested parts.

    13. On Crawford Hill
  8. Count a prototype as the start, not an innovation

    The point-contact transistor worked in December 1947 but changed behavior with humidity or a slammed door. Jack Morton's team made five thousand working transistors by summer 1949, and it was ready for large-scale production only by summer 1951.

    6. House of Magic
  9. Define a hard verification test before escalating any claimed breakthrough

    Research VP Ralph Bown asked Bardeen and Brattain to make their amplifier oscillate. When it did, he called it a basically new thing in the world. Mervin Kelly was not told until about a month later.

    6. House of Magic
  10. Judge a breakthrough by whether it can be mass produced

    The first silicon transistor in January 1954 was too complicated to mass produce and did not change the industry. Tanenbaum's 1955 diffused silicon transistor beat every germanium device and looked easy to make, so Jack Morton cut short a European trip to back it.

    10. Silicon
  11. Track cost per unit of value before scaling a technically successful pilot

    The silicon solar battery ran a six-month test in Americus, Georgia, starting October 4, 1955. Cells cost several hundred dollars per watt, and Chapin estimated in 1956 that powering a house would cost nearly $1.5 million.

    10. Silicon
  12. Find a new problem where a failed product's weaknesses stop mattering

    The silicon solar cell was too costly for rural phones, but in space nothing else could recharge batteries. Joe Chaffee's 1937 feedback receiver, considered largely useless, was revived for Echo and Telstar.

    13. On Crawford Hill
  13. Classify a failure as missed opportunity, perception or judgment

    Bell Labs missed the integrated circuit, bet on waveguides that glass fiber made obsolete, and launched a Picturephone nobody wanted.

    15. Mistakes
  14. Do not let a leader's received wisdom close off a direction

    Jack Morton believed multi-component chips could never be reliable and pursued functional devices instead. Kilby and Noyce bet reliability could be fixed later and built the integrated circuit on Bell Labs processes like diffusion and photolithography.

    15. Mistakes
  15. Watch outsiders with different expertise and no sunk investment

    Bell Labs spent years and millions on hollow waveguides, and Baker still championed them in a 1968 speech. Charles Kao argued for pure glass fiber in 1966, and Corning announced very low loss fiber in September 1970.

    15. Mistakes
  16. Never assume your existing network is a moat rivals cannot copy

    MCI built a cheaper microwave long-distance network without durable gear or local subsidies. In 1971 the FCC ordered AT&T to connect MCI to its local switching centers.

    16. Competition
  17. Ask whether your own breakthroughs remove the conditions protecting you

    An analog network needed one entity to control many amplifiers. Shannon's digital information and cheap, freely licensed transistors let any company code and decode messages, which undercut the case for the monopoly.

    11. Empire
  18. Look for unused capacity in existing systems before building new infrastructure

    A talker sends signals only about 35 percent of the time. TASI reassigned idle channels during pauses and doubled the transatlantic cable from thirty-six to seventy-two channels without a new cable.

    11. Empire
  19. Choose proven components with long test records where failure is costly

    Kelly calculated TAT-1 had to run twenty years without problems. He barred the transistor and used late-1930s vacuum tubes, some tested continuously for sixteen years. The technology never failed once in twenty-two years.

    11. Empire
  20. Check whether enabling technologies have matured before betting on an idea

    The 1947 cellular memo sat unpublished. Cellular became practical only with integrated circuits, microprocessors like the Intel 4004 with 2,300 transistors, and programmable electronic switching.

    17. Apart
  21. Invest in invisible infrastructure and sell it through concrete new features

    ESS No. 1 took two thousand man-years and cost more than $100 million. AT&T sold it through call forwarding, call waiting and conference calling, and the network's ability to find people later proved hugely valuable.

    14. Futures, Real and Imagined
  22. Frame hard bets with questions that turn possibly into probably

    Jewett and Carty used this framing in 1909 for a New York to San Francisco line. They hired Millikan's student Harold Arnold, whose vacuum tube repeater let the line open in 1915.

    1. Oil Drops
  23. Wait and repropose a killed idea with cost and competitive arguments

    Kelly called the satellite a nonstarter in summer 1958, and a rebuttal was never sent. After he retired, a January 6, 1959 proposal asking $1 million was approved as Project Echo.

    13. On Crawford Hill
  24. Design shared space so different disciplines keep running into each other

    Murray Hill connected all wings, used a 700-foot corridor and six-foot modules with full utilities. More than eighty industrial labs visited to copy it.

    5. Solid State
  25. Mix theorists with experimentalists and pay to fill a missing capability

    Shockley and Fisk judged only about three people in the country could fill the theory gap. Kelly offered Bardeen a salary well above any university's, and he landed in Brattain's lab, starting the theorist and experimentalist partnership.

    5. Solid State
  26. As a manager, guide your reports and never compete with them

    Shockley secretly designed the junction transistor and sprang it on his team. Bardeen left for Illinois and Brattain asked to stop reporting to him. Later his staff left after he asked them all to take polygraph tests.

    6. House of Magic
  27. Value the colleague who asks good questions over raw output

    A patent lawyer study found the only common thread among top Bell Labs patent holders was that they often ate lunch or breakfast with Harry Nyquist.

    7. The Informationist
  28. Give people slack time for self-chosen projects beyond their assignment

    Joel Engel believed advancement required spending 20 or 30 percent of your time beyond your assignment, and used it to plan cellular. Google lets workers spend up to 20 percent of their time on projects of their choosing.

    17. Apart
  29. Tie a research group to a concrete need someone depends on

    After the 1984 breakup Bell Labs still won Nobel Prizes, yet it was split repeatedly. By 2005 it was about a third of its former size, and Holmdel closed in 2006.

    19. Inheritance
  30. Separate incremental product work from foundational platform bets in funding

    Kelly spent decades improving the vacuum tube, then backed the transistor that made that work irrelevant. Monopoly security likely let Bell Labs escape the innovator's dilemma for a time.

    20. Echoes