March 15, 2018 by Jay Chow

How to Improve Mobile Ad Campaign Performance with Data (and Lots of It)

Postback data

Look, we get it… you have a top secret BI formula for determining customer LTV that uses 16 distinct variable calculations in addition to 11 secret herbs and spices. If partners knew how it was calculated, they’d try to maximize their score rather than getting you the highest quality users. Ending up with truckloads of “high score” installs that are of little value is a nightmare that keeps you up all night.

Relax. We aren’t asking for the formula to your secret sauce, instead just a general idea to what a quality user looks like. You know… the metrics you actually care about, what your best users look like, and of course any more data that you can share with us.

Liftoff does not have a hack team of monkeys pulling levers and making manual adjustments from campaign outputs. Everything from buying on publishers to testing creatives is automated with our advanced machine learning (ML) platform where data makes or breaks performance. Even if your campaign is already thriving, there may be more scale to unlock by sharing more data.

Leading cause of death to ad campaigns: Data starvation. Contact Liftoff if you know of any campaigns suffering from data starvation.

Remember back in high school chemistry where you dissolved salt in a solution until no more could be dissolved? Then when the solution cooled, the whole thing abruptly crystallized.

Over here it’s the same thing. Data goes in, ML trains and learns from the data, slowly improving and getting smarter. Then… saturation point! CPA models pop out, and we’re suddenly buying inventory on different strategies and predicting converting users left, right, and center.

Still not convinced? Then gather around for a true story: One of our clients, a major marketer of apparel, started running a campaign with Liftoff to drive first time purchases. However when we first started this campaign, only half a percent of installs led to purchases. With these numbers it would take over 60,000 installs to reach our CPA goal for first time purchase optimization! Nobody has the time or money for that!

Luckily this client also sent us postbacks for intermediary events such as registrations, account creations, first logins, completed onboardings, and many others. With this additional data we were able to direct our ML models to first optimize for users who had completed the registration process. Given the much higher frequency of this action, we were able optimize CPAs for this event within 4,000 installs, all while our ML continued homing in to the growing signal of first time purchases. As more data came in, our ML constructed models and optimized for further downstream events such as account creations and completed onboardings. By now our ITA (install to action rate) for first time purchases had improved to over 1% and we were scaling past $2,500 spend per day.

Finally our CPA model for first time purchases was ready, and in less than a week our ITA was up to 2.6%. We shortly scaled to over $7,000 a day while nearly doubling the client’s ROAS goals. This entire process took just over a month and 25,000 installs. What made this story a success? A pinch of patience, a bit of love, and a metric ton of data. So much data. All the postbacks they sent us contributed to the acceleration of reaching CPA goals for their ultimate a-event.

You don’t need a chart toppling app with millions of users (though we’d love to help you get there!) to benefit from sharing more data with us to feed our ML model. For a start, consider sharing all postbacks. With all postbacks, users acquired from other marketing partners can be avoided by our machine learning algorithms, saving, time, money, and increasing campaign spend efficiency.

Even something as simple as sharing revenue numbers, will let us see which of our conversions are most beneficial, allowing us to maximize return on ad spend (ROAS).

Like any good partnership, a successful campaign is built on trust. We really want to help your campaigns succeed but you have to trust us by doing the best to set us up for success.

Paraphrased from the words of visionary philosopher Jerry Maguire, “Help us help you succeed.”


Jay Chow is currently an Ad Operations Manager / Analyst for Liftoff. He manages & optimizes mobile marketing campaigns for performance, collaborating with account teams. Jay is passionate for using data analysis to influence and drive marketing decisions. He was previously a Media Buying Analyst at Rakuten.

Read more blogs