Somewhere in your organisation, there’s probably a slide with one enormous number on it. It’s the North Star Metric. Every team is supposed to contribute to it, and every initiative is supposed to move it in the right direction. One metric to rule them all.
Lovely.
Your product is a complex system, not a bloody compass, and complex systems have an irritating habit of ignoring our PowerPoint presentations.
The North Star Metric promises clarity. With competing priorities and uncertain outcomes everywhere you look, wouldn’t it be wonderful to have one number that told you whether you were heading in the right direction?
Of course it would. I’d also like a crystal ball.
The metric is green. The product is on fire.#
I’ve lived this one.
I joined a large utility company as head of product. Their North Star Metric was time on site.
We’d built customers tools to understand their bill, to see which of their decisions cost them the most, and to forecast their bill based on the weather where they lived. If people were spending time with those tools, the thinking went, they must want them. Time on site was the proof.
I didn’t believe it.
When we interviewed customers, a lot of them didn’t understand the basics, like what a kilowatt-hour is, or that their unit rate had far more to do with the size of their bill than anything our clever tools could show them.
Some had bought into pay-as-you-go meters, which had been marketed as a lifestyle choice. Propositions sat in my remit, so I knew the numbers. Those customers were paying twice as much for their energy. That one wound me up, if I’m honest.
They wanted a bill they could read, and a way to work out which tariff suited the way they lived.
The leadership team’s message was ‘find the right features and the user will stay’.
So time on site kept going up. And satisfaction kept going down, until our NPS was negative.
A customer staring at a bill they don’t understand is spending time on site. Confusion looks a lot like engagement when all you count is the minutes.
It took a negative NPS to get the OK to fix the basics.
So we went back to fabulous basics. We made the bill clearer and helped people choose the right tariff.
Within months of the switch, our NPS had swung from negative to strongly positive.
Nobody was fiddling the numbers. The leadership team believed time on site proved customers wanted what we were building. The measure had become a substitute for understanding the system.
This is where Goodhart’s Law usually gets wheeled out: when a measure becomes a target, it stops being a good measure.
Fair enough. But you don’t need anyone gaming the metric for this to happen. With accurate data and good intentions, a North Star can still lead you astray, because the product’s behaviour comes from the relationships between things.
‘You just picked the wrong North Star’#
That’s the obvious comeback. Time on site was a bad choice, so pick a better one.
We tried. We looked at other North Star Metrics, and the more we looked, the more we came to recognise how hard it is to choose a good one.
One number can’t balance the competing demands of retention, engagement and monetisation. Optimise aggressively for one and you’ll often damage another further downstream.
And what counts as value doesn’t stand still. A market shift or a product pivot changes it, but teams cling to an outdated North Star out of institutional inertia, even when it no longer reflects value.
Complexity doesn’t care about your alignment workshop#
The usual defence of a North Star is that it creates alignment. And it does.
But alignment and understanding aren’t the same thing. You can have an entire organisation beautifully aligned around the wrong reading of reality.
Alignment can even make it worse. When everyone’s pushing the same number, other interpretations get harder to see, and conflicting signals get treated as distractions.
At the utility, leadership was lined up behind finding the right features, and what customers were telling us in interviews kept losing the argument.
‘But we have supporting metrics and guardrails’#
This is usually where someone tells me I’ve misunderstood North Star Metrics.
You’re not supposed to use the number on its own. You have input metrics, supporting metrics and guardrails - measures that tell you whether you’re improving the North Star at the expense of something else.
That’s better than blindly chasing one number. But if you need a collection of other measures to tell you whether improving your North Star is good for the product, what exactly makes it the North Star?
I was in exactly that spot at the utility company, with two numbers pointing in opposite directions. Which measure wins?
For a while, the North Star did. Then NPS won the argument.
I know, NPS is a single number too, and it has its own critics. It wasn’t our choice. Finance had tied every point of NPS movement to a funding conversation, so it was the number the business listened to.
Which means time on site was never really our ultimate measure of success. The moment it disagreed with how customers felt, someone had to decide which one to believe.
That decision came down to judgement, based on what customers were telling us in interviews and what we thought would make them stay.
Congratulations. You’ve arrived at product management.
I don’t think the product needs a North Star at all.
Read the constellation#
‘So if we don’t have one metric, what are we supposed to do? Measure everything?’
No. That would be just as ridiculous.
I once inherited a team dashboard with 19 widgets, and the team used none of them to make a decision. Within 2 weeks we’d cut it down to 4 metrics and 2 charts, and the data finally started steering the conversation. (I wrote about that one in Stop Admiring the Data. Start Steering With It.)
Instead of asking ‘What’s the one number that tells us we’re succeeding?’, ask:
- What are we trying to achieve, and for whom?
- What evidence would suggest we’re making progress?
- What other things might change as a consequence?
- Where might improving one dimension damage another?
- What assumptions are we making about how this product behaves?
- What would cause us to change our minds?
Then put the measures those questions point you to side by side. At the utility, that meant reading time on site next to NPS, and both of them next to what customers were telling us in interviews.
Once time on site went, we dropped the North Star altogether. After that, every roadmap item we wanted to look at came with a splattering of metrics we were trying to change, along with the ones that shouldn’t be affected.
Please don’t combine them into a more sophisticated super-metric. As I argued in You’re Measuring Value Wrong, value isn’t a number. Each one gives you a different view of a system you’re trying to understand, and the interesting bit is where they disagree.
If time on site rises while NPS falls, that’s interesting.
If engagement and churn go up together, you usually have an understanding problem.
If revenue goes up while customer trust goes down, that’s very interesting indeed.
Resist the urge to tidy those contradictions away before the next steering committee. They’re telling you your understanding is incomplete, and that’s worth more than another green arrow on a dashboard.
Judgement is the job#
We reach for a single answer, whether it’s a root cause, a priority score or a North Star, anything that makes the hard decisions feel less hard.
Unfortunately, the hard decisions are the job, and a North Star never makes one of them. At the utility, people made every call that mattered, and for too long they let time on site speak for them.
So make the judgement visible. Say out loud which number you’re choosing to believe, and why.
The last time your North Star disagreed with another number you care about, who decided which one won?
