Lesson 24: Re-align Incentives So the Signal Actually Matters
You can pick the right signal, put it on a dashboard, and watch absolutely nothing change.
I’ve done it. Spent three weeks with a team agreeing on the one metric that proved we were creating value, built a clean chart, put it on a screen in the office. Everyone nodded. Six weeks later the number was flat and the roadmap looked exactly the same as it had before we started.
The signal wasn’t wrong. It just wasn’t attached to anything anyone was rewarded for.
The Rule: People Optimize What Gets Them Approved
Not what’s measured. What’s measured and then used to approve or reject their work. Those are very different things, and the gap between them is where most measurement programmes quietly die.
If your signal lives on a dashboard nobody opens before a decision, it’s decoration. If it lives in the room where the roadmap gets agreed, it’s currency.
So the move isn’t to measure harder. It’s to make the signal a required input to decisions people already care about.
A Real Example: The Nice-To-Have Problem
A product team I worked with had a genuine strategy problem that looked like a prioritization problem. Every planning cycle, roadmap items got added – sensible-sounding things, requested by real people, none of them obviously wrong. And every cycle, time to first value crept up. New customers were taking longer and longer to get anything useful out of the product.
Time to first value was the agreed signal. Everyone knew it. It changed nothing, because nobody had to answer for it.
We made one change to the planning ritual. Any roadmap change now had to arrive with two extra lines: an estimate of its effect on time to first value, and a sign-off from the person who owned that signal.
The estimates were rough. Some were close to guesses. That turned out not to matter much.
What mattered was that somebody now had to say out loud, in front of peers, “I think this adds four days to onboarding and I think it’s worth it.” Half the time, saying it out loud was enough to kill the item. The other half, the trade-off got made deliberately instead of by accident.
Within two cycles the roadmap was visibly leaner. Items that improved the signal moved faster because they had an obvious argument behind them. Support volume dropped as a side effect nobody had forecast – shorter onboarding meant fewer confused customers.
How To Do This Yourself
Put the signal in exactly two places and attach one rule to it:
- Place one: the weekly leadership review. Not as a slide at the back. As the first thing discussed, with the trend and the latest data point.
- Place two: the individual team scorecard, so the people doing the work see the same number as the people reviewing it.
- The rule: any trade-off that touches the signal must include a predicted impact and a named sign-off from the signal owner.
- Keep a two-line log of each decision – what was approved and what impact was predicted.
- Once a quarter, compare the predictions to what actually happened. Not to punish anyone. To calibrate.
That last step is the one people skip, and it’s the one that turns rough guesses into good instincts over about two quarters.
The Honest Lesson
Two things I got wrong here.
The first is that I used to wait for good estimates before introducing the rule. That’s backwards. The estimates are bad at first and they get better precisely because you’re forced to make them. Waiting for accuracy means waiting forever.
The second is subtler and it cost me more. When you make a signal into currency, people will optimize it – including in ways you didn’t intend. Give a team a number to hit and some of them will find the cheapest route to hitting it. I’ve watched a support team improve first-response time by sending faster, emptier replies.
So pair the signal with one guardrail metric that would get worse if someone gamed it. Time to first value paired with thirty-day retention. First-response time paired with resolution rate. You’re not looking to catch anyone out. You’re making sure the number can’t be won without the outcome being won too.
Look at your last three significant decisions. If none of them referenced your signal, you don’t have a signal. You have a chart.
Next time, I’ll dig into “Measure Customer Impact with Two Direct Conversations” – see you there.