Moloco, Liftoff and Mintegral Buying Against Your Own Events

Moloco and Liftoff need roughly 300 to 500 instances of your target event per week, per campaign, before their bidding is worth more than a flat CPI buy. Below that the model is guessing and you are paying a percentage of media for the privilege. So the first decision here is which event you can genuinely supply at that rate: second session, trial start, first listing enquiry, first paid order. If none of them clears the bar we bid on a proxy, write down in the plan that it is a proxy, and set the date we revisit it. What this engagement is: Buying run inside your own network seats, in your company name, with our users removed on the last day. We do not resell media and we do not hold the contract with Moloco, Liftoff or Mintegral. Rebates from those networks do not appear in our accounts because we never take them. You see raw platform spend. The event and postback layer, which is most of the work: We specify the server-to-server postbacks from AppsFlyer, Adjust or Singular to each network, decide which events each bidder sees and at what value, and write the SKAdNetwork conversion value schema. Apple sends at most three postbacks per install, in the 0 to 2 day window, 3 to 7, and 8 to 35, with coarse values in the later two for most apps. That schedule decides what is biddable on iOS. We set the iOS target inside the first window and keep the later windows for measurement rather than optimisation, because a signal arriving on day 35 cannot steer a bid placed today. Where the three actually differ: Moloco rewards a dense, stable event stream and degrades quickly when the event definition changes mid-flight, so we freeze the schema for the first six weeks and hold you to that. Liftoff bids at creative level across its own exchange, which means it needs more creative variants than the other two before frequency saturation shows up as rising cost per event. Mintegral has the widest supply and the most publisher-level noise; we run it with an exclusion list from day one and rebuild that list monthly from the MMP publisher report, never from the network's own. The Turkey question: Turkey is one of the cheapest install markets in the world, which is why worldwide campaigns quietly fill with Turkish, MENA and South Asian traffic and report a CPI the buyer is pleased with. If you are buying worldwide, those regions come out as separate line items with their own targets from the first day, so the blended number never flatters anyone. If Turkey is your actual market, we buy it in lira with weekly rate checks, because a monthly average hides what a six percent move does to a bid cap. Not included: Creative production. Playables, video and end cards are a separate engagement and we will not pretend that rotating your existing assets is creative work. Also excluded: MMP implementation from scratch, ASO, attribution disputes with a network over less than four figures, and any arrangement where we are paid on install volume. Who this is not for: Apps before product-market fit. Apps with no monetisation or qualification event inside seven days. Teams who need the install number to rise this month regardless of what happens after it. We also decline where MMP raw data export is not available to us, because buying without it is guesswork sold at a professional rate. What arrives every week: One page. Spend, cost per target event by network and campaign, D7 retention and revenue by cohort, the decisions taken since the last one, and the decisions we are holding until the data supports them.