Streetlight Effect
Teams optimise what's easy to measure and neglect what matters — because the easy metrics give faster feedback.
Why PMs should care
Clicks are easy to count. Satisfaction is hard. Activation can be tracked today; long-term trust cannot be measured at all without waiting a year. Short-term engagement shows up in any dashboard; the 18-month version of it only shows up after 18 months.
The predictable result: most product teams work hard on the metrics that are easy to see, and quietly ignore the ones that actually move the business. This isn't laziness. It's that easy metrics reward you with fast feedback, and hard metrics reward you with a shrug — and people naturally go where the feedback is.
The partial fix: pick one true north metric that you promise not to optimise directly. Then check, regularly, whether the easier metrics you are optimising are still moving this true north in the right direction. If engagement is up but retention is flat, the streetlight is lying to you.
Example in product work
Tutorial completion vs retention. A team optimises tutorial completion rates for three quarters. It climbs from 54% to 78% — a clear, celebrated, bright-lit win. Meanwhile, 90-day retention of users who completed the tutorial is flat; 90-day retention of users who skipped it is higher than either. The tutorial became shorter and more polished, and also more optional-feeling, and the users who genuinely needed it are the ones who now skip it because the first screen says 'skip available'. Completion is up 24 points. The actual product outcome — engaged 90-day users — hasn't moved. The team spent a year polishing the thing they could see, and the thing that mattered was in the dark the whole time.
Quibi, 2020. Quibi is the streetlight effect at $1.75 billion. Jeffrey Katzenberg's team tracked everything that was easy to measure: premium content spend, celebrity deals, minutes of content produced, turnstile partnerships, pre-launch press coverage. What they didn't measure — because it required a harder kind of research than a dashboard — was whether anyone actually wanted short-form premium video designed specifically for phone commutes. When COVID hit and phone commutes vanished, the thing they'd never validated turned out to be the entire product. Quibi shut down six months after launch. The bright-lit metrics were all green through the whole run.
What to do when you see it
- The easiest metric to measure is rarely the one that matters most.
- Fast feedback from a proxy metric is seductive. It can feel like progress even when it isn't.
- Pair every optimisation metric with a slower, harder metric that you refuse to optimise directly.
- If the two disagree, trust the slower one.
Sources & further reading
- The Streetlight Effect — David H. FreedmanAn excerpt from Freedman's 2010 Discover article on his own site — the piece that popularised the term.
- Streetlight Effect and Forecasting — LokadA concise applied take on why we search where it's easy rather than where the answer is, with a forecasting-and-decisions lens.