This case study adapts Xiao Wei's 2022 article about user quality when app campaigns optimize for application, loan disbursement, or repayment events. The numbers describe one product and period; they are not transferable bidding or credit-risk thresholds.
Download the event quality review worksheet
App Ads Event Quality Review (PDF, English and Simplified Chinese). A blank worksheet for comparing application, disbursement, and repayment cohorts with consistent definitions and mature outcomes. It contains no benchmark thresholds.
What did the three event groups show?
The source compares campaigns from October 1–15, 2022. “Top three tiers” refers to an internal risk ranking, not an observed repayment rate.
| Optimization event | Spend (source basis) | Applications | Top-three-tier share |
|---|---|---|---|
| Application | 1,478.07 | 3,428 | 38% |
| Disbursement | 1,238.25 | 2,339 | 46% |
| Repayment | 455.84 | 679 | 33% |
The disbursement group had the highest share of top-tier users, while repayment had the lowest. Yet the author reported that the repayment-event group performed about 3%–5% better on repayment than other groups, and disbursement about 8% better than application. The source does not fully specify the outcome definition, observation window, or statistical uncertainty, so these are observations rather than proof that one event always wins.
Why can risk tiers and outcomes disagree?
- A risk tier is a proxy. Actual repayment also depends on approval rules, user behavior, and servicing.
- Repayment is closer to business value but has fewer observations and a longer delay. Missing postbacks or short attribution windows can weaken the optimization signal.
- Spend and sample sizes differed; markets, creative, and budgets may also have differed. These campaigns cannot be ranked causally from the table alone.
- Disbursed users have already passed a risk screen, creating selection bias.
Did repayment data shift placements?
The source observed one disbursement-event campaign whose budget rose from about 100 to 200 between October 11 and 23. As repayments arrived, spend shifted toward Google Display and YouTube. The author proposed two explanations: the new repayment signal or the budget change. Before-and-after screenshots cannot separate them. Review budgets, event timestamps, markets, and creative changes together. See Google's App campaign placement reporting guide and our Google App campaign guide.
How to test event choice more reliably
- Define the final outcome first, such as mature-cohort collections or risk-adjusted ROAS.
- Keep market, product, creative, budget, and attribution as comparable as possible; use a controlled experiment where feasible.
- Record applications, approvals, disbursements, due-date repayments, acquisition cost, and sample size for each cohort.
- Wait for outcomes to mature and report uncertainty before changing bids or credit policy on a short-term difference.
For campaign setup, see our Meta app ads guide.
FAQ
Is the latest possible optimization event always best?
No. It may be closer to value, but volume, delay, and selection bias matter.
Can the top risk-tier share alone measure campaign quality?
No. Validate it against mature disbursement, repayment, and cost outcomes.
Does a shift toward YouTube prove repayment signals worked?
No. The example also changed budget, so the cause remains uncertain.