Chapter 28 of 29 · Organising a place for discovery
1 The bottom line
Treat a working prototype as the start: test it cheaply, measure what each unit costs and whether it holds up, and scale only when it beats what it replaces. Many working results fail on cost, reliability or demand, not on the science.
2 Key takeaways
- Test the core idea with the simplest, cheapest experiment you can. It exposes real problems before you make costly commitments.
- Echo, a passive balloon satellite, cost under $2 million and about three dozen men
Jon Gertner, Bell Labs.
- Echo, a passive balloon satellite, cost under $2 million and about three dozen men
- Count a prototype as the start of the work, not the end. Reliability, repeatable manufacture and a market all still have to be built.
- The point-contact transistor worked in December 1947 but changed behavior with humidity or a slammed door
Jon Gertner, Bell Labs.
- The point-contact transistor worked in December 1947 but changed behavior with humidity or a slammed door
- Measure cost per unit of value before you scale. A pilot can work perfectly and still be a financial failure.
- Solar cells cost several hundred dollars per watt, and Chapin estimated in 1956 that powering a house would cost nearly $1.5 million
Jon Gertner, Bell Labs.
- Solar cells cost several hundred dollars per watt, and Chapin estimated in 1956 that powering a house would cost nearly $1.5 million
- Ask buyers what they would pay, not whether they like it. Crowds trying a product show interest, not demand.
- AT&T surveyed 700 Picturephone users at the 1964 fair, and at $40 to $60 a month only 12 percent of couples wanted one at home
Jon Gertner, Bell Labs.
- AT&T surveyed 700 Picturephone users at the 1964 fair, and at $40 to $60 a month only 12 percent of couples wanted one at home
- Put real installations with paying customers. Working performance proves a product better than more lab refinement.
- The Bloomfield engine, started in March 1776, made about fifteen strokes a minute and drained nearly sixty feet of water in under an hour Jenny Uglow.
But to an innovator, being early is not necessarily different from being wrong.
Jon Gertner, Bell Labs
2.1 Run the cheapest test first
A small, simple experiment shows what breaks before you spend real money.
- Pick the simplest version that tests the core idea, even if it earns nothing.
Jon Gertner, Bell Labs
- Run it with a small team and a small budget.
- Let its job be to expose real problems before you commit.
- "There's a difference, you see, in thinking idly about something, and in setting out to do something."
- Once the experiment works, hand it to a development team built for scale and reliability.
Jon Gertner, Bell Labs
- Expect the build to be far bigger than the test.
- The active Telstar that followed Echo needed more than five hundred scientists and engineers and fifteen thousand tested parts.
- Ask what must be done to turn "possibly" into "probably".
Jon Gertner, Bell Labs
- Ask whether the device can be built now. If not, ask why and what science is missing.
- Ask whether that knowledge will come in time, and what raises the odds before the deadline.
- 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.
2.2 Test demand with money, not applause
Interest is cheap to get, so ask what people would pay and who is answering.
- Ask for willingness to pay at concrete monthly price bands, separately for home and business.
Jon Gertner, Bell Labs
- Offer price bands, not a yes or no on interest.
- Record sample bias, since visitors at a showcase are not a cross section of customers.
- Even at $60 to $80 a month, 29 percent of Picturephone business users were interested.
- For social and communication products, ask whether users want to be seen, not only to see.
Jon Gertner, Bell Labs
- A survey that asks only about the viewer's benefit misses the cost to the person viewed.
- Most Picturephone users said seeing the other person was important, but no one asked if they wanted to be seen. Rudi Kompfner later put a photo of himself in front of his set.
- Weigh a negative market study against field evidence when the product creates a new behavior.
Jon Gertner, Bell Labs
- Studies can only gauge demand for things people already know.
- Run pilots with paying users to settle it.
- 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.
2.3 Check cost per unit against what you replace
| Relay switching | Transistor switching | |
|---|---|---|
| Speed | ~1,000 ops/sec | up to ~1,000,000 ops/sec |
| Hardware | 65,000 relays | Transistors |
| Justified on cost then? | Yes | No |
| Adoption | The incumbent | Took nearly 20 years |
A result that works but costs more than the alternative is not yet a product.
- Measure what each unit of output costs and compare it to the alternatives.
Jon Gertner, Bell Labs
- Do this before scaling a technically successful pilot.
- Treat a brilliant but uneconomic invention as waiting for its problem, not as a product.
- The silicon solar battery ran a six-month test in Americus, Georgia, starting October 4, 1955.
- Require any replacement to do the job better, cheaper, or both.
Jon Gertner, Bell Labs
- Novelty alone is not a reason to switch.
- If it is neither better nor cheaper yet, keep improving it rather than forcing adoption.
- 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 be justified on cost, and electronic switching took nearly twenty years to fully deploy.
- Cost every production stage when sales are high but cash is short. Jenny Uglow
- Build a price book that costs every stage through to the shop counter.
- Use longer runs to spread overheads and allow lower prices for a wider market.
- Wedgwood found cash tied up in expenses and unsold stock, and ornamental prices unrelated to cost. In April 1773 the Potters' General Assembly agreed to cut prices by 20 per cent.
2.4 Prove it holds up and can be made in quantity
A device that works once is not a product until it works every time and is easy to build.
- Plan for reliability, reproducibility and materials consistency before you declare success.
Jon Gertner, Bell Labs
- A discovery observes nature, an invention applies science in a new way, and an innovation is the long work of making the thing in quantity and selling it.
- Jack Morton's team made five thousand working transistors by summer 1949, and the transistor was ready for large-scale production only by summer 1951.
- Back the version that is both better and easy to make, and back it fast.
Jon Gertner, Bell Labs
- A first working device can be irrelevant if fabrication is too complex.
- 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. His words: "This looks like the transistor we've been waiting for. It should be a cinch to make."
- Where failure is costly, choose proven components with long test records.
Jon Gertner, Bell Labs
- Set the payback condition first, such as years of failure-free operation.
- Then pick the least novel parts that meet it.
- Run long durability tests early so the data exists when a high-stakes project arrives.
- Kelly calculated that 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.
2.5 Put it with real customers
Paying use proves the product and shows what the lab cannot.
- Ship real installations for paying use instead of refining in the lab. Jenny Uglow
- Boulton pushed Watt to build two large engines for immediate paying use instead of endless improvement.
- Their working performance proved the engine against Newcomen's.
- A supplier able to machine parts precisely, Wilkinson's boring bar, solved the hardest bottleneck. Boulton put the pitch this way: "I sell here, sir, what all the world desires to have, POWER."
- Put one trusted, skilled engineer on site with customers. Jenny Uglow
- Few men could assemble engines to precise plans, and wrong or late parts from Soho caused failures.
- Have that engineer fix problems and win over local experts.
- William Murdoch went to Cornwall in autumn 1779, stayed twenty-one years and became the only man Watt fully trusted. His plea to customers: "Have pity on them, bear with them, give them another trial."
2.6 Check the world is ready before you scale
Scaling pays only when the enabling parts, the network and the buyers are in place.
- Check whether the enabling technologies have matured before betting on the idea.
Jon Gertner, Bell Labs
- Big ideas often arrive before their parts exist.
- Keep shelved plans alive for when the window opens.
- 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. As the book puts it: "Cellular is a computer technology. It's not a radio technology."
- If the product is a network, plan how it reaches critical mass before launch.
Jon Gertner, Bell Labs
- A network product gives each user little value when few others have it.
- Price cuts cannot rescue a network that is too small.
- 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.
- Where you can, design the product to work with an existing network from day one.
Jon Gertner, Bell Labs
- A product that plugs into a network people already use delivers value to the very first customer.
- 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.
2.7 Where the books differ
The right tactic depends on how costly failure is, how big the build must be, and whether buyers can picture the product.
- Start small to test, but do not plan to stay small. Jenny Uglow
- Cheap experiments answer whether the idea works. They do not justify the tools, trained workers and dedicated works that only pay off at full scale.
- Boulton refused narrow regional rights to Watt's engine. He argued a dedicated works could build engines about 20 per cent cheaper, and as much more accurately as an instrument maker is than a blacksmith. His line: "I find it very well worth my while to make for all the world."
- Ship and fix on site where failure is cheap to repair. Use proven parts where failure is costly.
Jon Gertner, Bell Labs
- The Watt engines went to paying users, with Murdoch on hand to fix problems Jenny Uglow.
- For TAT-1, Kelly picked the least novel parts because the payback condition was years of failure-free running.
- Survey people who have tried the product. For products nobody has imagined, trust pilots over market studies.
Jon Gertner, Bell Labs
- The Picturephone survey asked real users about price and gave usable answers.
- The mobile phone study asked about a behavior people did not yet know, so it only measured demand for products that already exist.
Sources
- The Idea Factory: Bell Labs and the Great Age of American Innovation · Books
- The Lunar Men: The Inventors of the Modern World, 1730–1810 · Books
Part of Organising a place for discovery in the growth library.