Michael Paxman
Global Product Communications Manager at Adjust
Where to hear and read about Michael Paxman1 episode or test
Michael Paxman is Global Product Communications Manager at Adjust, a mobile attribution company that bought the bot-detection startup Unbotify. The most useful thing to learn from him is to judge campaigns by the value of the users they bring in and to act on attribution data fast, including signs of bot traffic.
If something looks inefficient, act on it.
Mobile Ad Fraud Bots & Inefficient Marketing Campaigns
Michael Paxman: the story and the lessons
1Michael Paxman on acting on attribution data
Paxman says to act as soon as data shows inefficiency. A properly set-up attribution tool gives more data points than most teams imagine. Paxman says teams that ignore them are turning a blind eye to data they already pay for.
Waiting on a losing campaign doubles down on the loss.
Paxman argues monthly campaign reviews are too slow. Buying a month of ads and checking numbers at month end no longer works because network prices keep rising. He suggests setting up the data views you prefer and marking the red flags to catch.
"The days of like buying a month's worth of ads... let's look at our numbers at the end of the month, it's kind of not good enough anymore."
Paxman warns against rebuilding strategy every three months. Even new mobile-first companies tell him their marketing strategy changes every 3 months. That can leave 2 months and 3 weeks of data showing the approach was wrong.
His fix is to use the data during the cycle and correct course earlier.
2Paxman on judging campaigns by user value
Paxman wants teams to price a customer, not an install. Cost per install hides whether a user ever becomes a customer. He points to cost per subscriber or paying customer as the target, and says he reacts strongly to teams that only count installs.
He says players of all sizes skip this and he is horrified at how many companies do. The question to ask: "you've got to look at the value of these users and are they worth what I'm spending?"
Paxman says define a good user first, then a good campaign. Once you know what a good user looks like, you can work out what a good campaign looks like. In his words, that gives you all the tools needed for relatively smart marketing.
Until you know, he advises sitting tight and riding the early volatility, then looking for those signals.
Paxman gives no fixed rule for how long to wait. For a brand new app he says you will not know much for at least a week, and seasonality and competitors still blur the picture. He calls his answer on how much spend is enough "weaselly" because it differs for every app.
A new app may have to hop on the roller coaster and go with it until it knows what to look for.
Paxman ties the analysis window to the user lifecycle. A hyper casual game has short sessions, so it can be analyzed early. A house-selling app needs a full rotation of its lifecycle before users can be judged.
His rule: "look at what you would expect a typical user's life cycle to be."
3What Paxman learned about bots after Adjust bought Unbotify
Adjust bought Unbotify to tell human users from bots. Unbotify is a Tel Aviv startup, and the deal was announced at the start of the year. Its SDK judges whether device interaction is human from pressure, timing and phone movement.
The aim was first to protect ad spend by flagging bot installs and in-app actions so advertisers could act on those sources. Paxman calls the technology insanely complicated.
Paxman says bot abuse in apps is far bigger than expected. Once the team looked inside Unbotify's database, they saw the tip of the iceberg. Paxman cannot share numbers or a percentage, but says it is so much worse than you think.
"The amount of abuse that goes on in apps by bots is unreal." Shopping apps, ticket sales and PvP games are common targets.
Mobile apps rarely test whether a user is human. Websites use CAPTCHAs, but on mobile no one is tested, so bots meet little resistance. Paxman warns that bot developers are extremely smart people.
Bot makers earned millions from websites, and he notes ticket scalping by bots has gone on for over a decade. They did not leave apps alone.
Adjust tests the user every time, not just at install. Some people play for a couple of days like humans and then use a bot to win. Bot checks therefore have to continue across acquisition, advertising and the whole in-app experience.
Asked whether the user is tested every time, the answer is yes. Paxman says that with click spam handled, the market can go deep on harder problems.
Paxman frames bot defense as protecting honest players too. In a PvP game, a bot making perfect moves wipes real players out instantly, which he calls a horrendous user experience. Detection protects ad budgets and also supports retention.
Mobile Ad Fraud Bots & Inefficient Marketing Campaigns
Attribution data only pays off if you act on it quickly, judge campaigns by user value instead of installs, and watch for bot traffic that hides inside apps. Adjust's acquisition of Unbotify showed that bot abuse in apps is far larger than most teams assume.
Paxman says to act as soon as data shows inefficiency. A properly set-up attribution tool gives more data points than most teams imagine. Paxman says teams that ignore them are turning a blind eye to data they already pay for.
Waiting on a losing campaign doubles down on the loss.
Paxman argues monthly campaign reviews are too slow. Buying a month of ads and checking numbers at month end no longer works because network prices keep rising. He suggests setting up the data views you prefer and marking the red flags to catch.
"The days of like buying a month's worth of ads... let's look at our numbers at the end of the month, it's kind of not good enough anymore."
Paxman warns against rebuilding strategy every three months. Even new mobile-first companies tell him their marketing strategy changes every 3 months. That can leave 2 months and 3 weeks of data showing the approach was wrong.
His fix is to use the data during the cycle and correct course earlier.
Paxman wants teams to price a customer, not an install. Cost per install hides whether a user ever becomes a customer. He points to cost per subscriber or paying customer as the target, and says he reacts strongly to teams that only count installs.
He says players of all sizes skip this and he is horrified at how many companies do. The question to ask: "you've got to look at the value of these users and are they worth what I'm spending?"
Paxman says define a good user first, then a good campaign. Once you know what a good user looks like, you can work out what a good campaign looks like. In his words, that gives you all the tools needed for relatively smart marketing.
Until you know, he advises sitting tight and riding the early volatility, then looking for those signals.
Paxman gives no fixed rule for how long to wait. For a brand new app he says you will not know much for at least a week, and seasonality and competitors still blur the picture. He calls his answer on how much spend is enough "weaselly" because it differs for every app.
A new app may have to hop on the roller coaster and go with it until it knows what to look for.
Paxman ties the analysis window to the user lifecycle. A hyper casual game has short sessions, so it can be analyzed early. A house-selling app needs a full rotation of its lifecycle before users can be judged.
His rule: "look at what you would expect a typical user's life cycle to be."
Adjust bought Unbotify to tell human users from bots. Unbotify is a Tel Aviv startup, and the deal was announced at the start of the year. Its SDK judges whether device interaction is human from pressure, timing and phone movement.
The aim was first to protect ad spend by flagging bot installs and in-app actions so advertisers could act on those sources. Paxman calls the technology insanely complicated.
Paxman says bot abuse in apps is far bigger than expected. Once the team looked inside Unbotify's database, they saw the tip of the iceberg. Paxman cannot share numbers or a percentage, but says it is so much worse than you think.
"The amount of abuse that goes on in apps by bots is unreal." Shopping apps, ticket sales and PvP games are common targets.
Mobile apps rarely test whether a user is human. Websites use CAPTCHAs, but on mobile no one is tested, so bots meet little resistance. Paxman warns that bot developers are extremely smart people.
Bot makers earned millions from websites, and he notes ticket scalping by bots has gone on for over a decade. They did not leave apps alone.
Adjust tests the user every time, not just at install. Some people play for a couple of days like humans and then use a bot to win. Bot checks therefore have to continue across acquisition, advertising and the whole in-app experience.
Asked whether the user is tested every time, the answer is yes. Paxman says that with click spam handled, the market can go deep on harder problems.
Paxman frames bot defense as protecting honest players too. In a PvP game, a bot making perfect moves wipes real players out instantly, which he calls a horrendous user experience. Detection protects ad budgets and also supports retention.