Instrument-and-Improve S-Curve Prioritization
Let retention curves and diminishing returns tell you when to optimize and when to bet on something new
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 85%
A system for deciding where to invest across a product portfolio by tracking each feature's growth curve and detecting when marginal returns approach zero. Use it to allocate resources between improving what you have and betting on new formats, and to diagnose a metric drop in a mature product.
Origin
Asked how he balances 'just make things better' against hitting KPIs, and how he splits resources between optimizing Google and betting on new things, Stein rejected the either/or and laid out how metrics act as a pilot's instruments across a product's lifecycle.
Core principles
- 01Vision and metrics are intersected, not opposed: start from a problem/vision, then instrument to know if you're on track.
- 02Every feature rides its own S-curve — young features grow, mature features plateau, some decline.
- 03Metrics show you the way but never tell you exactly what to do — you still have to think for yourself how to make it better.
How to run it
- 1
Ship v1 and check the retention curve
After building the first version, judge whether people actually like it using a J-curve: the percentage of people still using it at day 7, day 30, day 90. Does retention flatten, or do people drip out toward zero?
Pro tip Pair the analytics with talking to people — scrutinize qualitatively, don't rely on the curve alone.
Watch out If retention never flattens above zero, on a long enough timeline no one uses it — you're toast; don't scale it.
- 2
Gate on 'good enough to talk about' then 'how big'
Next gate: is it good enough that people talk about it so it grows? Then ask how big it can actually get — small, medium, or large. No product starts big; even fast-growing ones started with 100-200 people internally.
- 3
Watch for diminishing marginal return
For each feature, estimate the expected value of further investment. When you could put 50 people on a project and it still won't move the needle, marginal return is approaching zero — that's the plateau of the S-curve.
Pro tip These zero-return moments usually coincide with something fundamental changing (user expectations, market saturation) — the signal to go first-principled for the next growth engine.
- 4
Root-cause any drop by slicing
When a core metric drops (say 5% this week), do root cause analysis: is it a region, a device, a demographic, a use case? Isolate where the problem lies, understand it, then apply the specific fix — the 'treatment for that disease' — and return to growth.
Pro tip Treat metrics as flight instruments: they tell you if you're flying correctly, not where to steer.
In the wild
AI Overviews showed people asking harder questions Google couldn't answer, and users literally appending 'AI' to their queries. Rather than optimize the plateauing core search UI, Stein's team treated it as a moment to open a new growth curve and built AI Mode.
→ A new growth engine that expanded search rather than a marginal tweak to a mature product.
Common mistakes
Optimizing a plateaued feature
Pouring 50 people into a feature at diminishing marginal return feels productive but won't move the needle — the discipline is recognizing the plateau and reallocating to a fresh curve.
Building without a quantified goal
Stein warns that without knowing where you're headed you just 'congratulate yourselves' — you can't tell if the improvement mattered to anyone.
Is it for you?
Best for
Product leaders and founders allocating limited resources across a portfolio of features
Not ideal for
Pre-launch products with no usage data yet, or pure zero-to-one research bets where curves don't exist
From the transcript
“You might look at something like a J curve. So this is the retention, the percentage of people still using the product day seven, day…”
“you get to these points of just diminishing marginal return in every system where it feels like you could put 50 people on this project”
“you need to be really close to root cause analysis there”
From the episode
Inside Google's AI turnaround: The rise of AI Mode, strategy behind AI Overviews, and their vision for AI-powered search
Robby Stein (VP of Product, Google Search)