Fixing App Growth Measurement: A Practical Playbook
Growth teams often treat bad ads as the main problem, but the deeper issue is usually a measurement system that no one trusts. This guide turns a conversation with Eric Seufert into a working playbook: how to concentrate spend without adding overhead, how to decide when a channel has earned more budget, and how to test creatives when terminal LTV and incrementality are still unknown.
Why measurement disorganization blocks growth
The most common pitfall Seufert describes is not a weak ad account but a chaotic measurement setup. Teams run competing tools, cannot tell which output to trust, and treat several disconnected data points as if they formed a single picture. The fix is organizational before it is technical: stakeholders need to agree on one measurement plan that everyone can interpret the same way. Seufert frames the measurement system as the heartbeat of the company, because budget, creative, and channel decisions all flow from it. If the heartbeat is irregular, every downstream decision inherits the noise.
Concentration vs. diversification: the waterfall method
Putting most of your ad spend into one or two large channels can feel risky, but diversifying for its own sake adds overhead and complexity. Seufert's alternative is a waterfall. Start with your largest channel and keep adding budget there until its return reaches the ROAS threshold you have set. Only then do you open the next channel, which typically has a lower spend ceiling, and push it to the same threshold. Repeat for the third channel and any that follow. The key metric is spend-adjusted ROAS, not raw ROAS, because the question is what the next dollar buys, not what the average dollar bought. The waterfall keeps the channel mix concentrated where returns are strongest while still allowing controlled expansion.
A worked example of the waterfall
Suppose your ROAS threshold is 1.5 and your biggest channel is currently returning 2.0 at its current spend. Under the waterfall, you keep increasing budget there until the marginal return falls to 1.5. Only then do you open the second channel. If that channel returns 1.8 at a small spend but drops to 1.4 after a modest increase, you stop at the point where it crosses 1.5 and move to the third channel. The example is illustrative, not a recommended threshold; the point is the sequence. Each channel is pushed only to the edge of acceptable return before the next one is opened.
Creative testing: kill losers fast, let winners age
Seufert's creative testing approach is deliberately lopsided. Underperformers tend to reveal themselves quickly, while a strong creative needs a longer runway before its value is clear. In the early days after launch, terminal LTV and incrementality are unknowable, so the practical move is to increase spend methodically in the places that are working while gathering more data. A creative that is clearly not a winner should be cut immediately, because at best it performs at the average and will not grow spend. Apparent winners should be left alone long enough to see how they perform over time. Track ROAS for each cohort at day 20, day 30, day 50, and beyond, and raise the budget frontier as the evidence supports it.
A decision checklist for the measurement plan
Before scaling spend, confirm the following with the stakeholders who own budget and reporting:
- One agreed measurement framework, with a named owner and a written definition of each metric.
- A stated ROAS threshold that determines when a channel is considered maxed out.
- A spend-adjusted ROAS view, not only a blended ROAS view.
- A waterfall order for channels, with the spend limit of each channel recorded.
- A creative review cadence that separates fast loser cuts from slower winner validation.
- A cohort ROAS schedule (for example, day 20, 30, and 50) agreed in advance.
- A rule for when the budget frontier moves, so the decision is not relitigated each week.
Failure distinction: losing creative vs. unproven creative
Two failures look similar in a dashboard but call for opposite responses. A losing creative is one that is clearly not a winner; keeping it running will not grow spend and at best matches the average. An unproven creative is one that has not had enough time or cohort data to show its long-term value. The first should be cut immediately. The second should be allowed to age while cohort ROAS is tracked. Confusing the two leads teams to either keep dead weight in rotation or kill a future winner before its cohort has matured.
Frequently asked questions
Is it risky to put most of my ad spend into Meta and Google?
Concentration carries risk, but diversifying for its own sake adds overhead. Seufert's advice is to diversify strategically using a waterfall: max out the biggest channel until it hits your ROAS threshold, then the second-biggest, then the third, and so on. Optimize for spend-adjusted ROAS rather than raw ROAS.
What is the biggest pitfall across growth teams right now?
Measurement disorganization. Teams use competing tools, do not know which output to trust, and cannot interpret several data points as a coherent ensemble. The fix is a measurement plan agreed with stakeholders so everyone reads the same system the same way.
How do I make confident decisions with incomplete data?
Identify losers quickly and let winners age. Early on, terminal LTV and incrementality are unknowable, so increase spend methodically in the places that are working. Cut creatives that are clearly not winners, and track cohort ROAS at day 20, 30, 50, and beyond before raising the budget frontier.
What is spend-adjusted ROAS?
Spend-adjusted ROAS looks at the return on the next increment of spend rather than the average return across all spend. It is the metric Seufert recommends for deciding when a channel has been maxed out and the waterfall should move to the next channel.