The Tyranny of Product Metrics : When optimising ruins products

If you’re the sort who finds numbers thrilling and paragraphs exhausting, you may skip to the bottom where I’ve provided a TLDR.

There is a saying in the management circles, usually uttered with an air of gravitas, "What gets measured, gets managed". The problem,
of course, is that what gets measured is also mercilessly optimised, often until the very soul of the product has been wrung out and replaced
with a dashboard friendly husk.

In the modern product world, metrics are worshipped. Every click, every swipe, every millisecond of dwell time is dutifully recorded, charted
and presented in beautiful slide decks. It all looks terribly scientific. But if you have ever spent any amount of time in such meetings, you will
know that this science is closer to astrology than astronomy : the stars do twinkle, but their significance is mostly what we project onto them.

The Rise of the Metric Overloads

The trouble begins innocently. You launch a feature. Someone suggests measuring engagement. Perfectly reasonable. A few months later,
The engagement chart is proudly displayed in the quarterly review. Soon, it becomes the chart. All eyes dart toward it with the nervous
intensity of traders watching a volatile stock.

This is how features die: not with a whimper, but with a slow suffocation under metrics. For once a metric exists, it demands improvement.
Nobody wants to be the person responsible for the downward arrow. The PM, the engineers, the designers — all bend their creativity not
toward making the product better, but toward nudging a particular line on a graph upward.

Consider the case of social media “engagement.” What began as a rough measure of user interest soon became a ruthless dictator.
Algorithms were tuned to maximise clicks, shares, and watch time. The results were spectacular — in the sense that a fireworks factory
explosion is spectacular. Short-term numbers soared. Long-term trust collapsed.


The Problem of Goodhart's law

Economists, who occasionally say useful things, have a principle
known as Goodhart’s Law:
“When a measure becomes a target, it ceases to be a good measure".

The law applies to product management with cruel precision. Watch time
is a decent measure of video quality, until you optimize for it.
Then you discover the easiest way to boost watch time is not to improve
quality, but to feed people endless loops of outrage and distraction.
Congratulations, your product is now a slot machine.

Similarly, “active users” sounds healthy — until you realize you can inflate
the number with aggressive notifications, email spam, or by making
it just slightly annoying to log out. The metric goes up. The user experience,
meanwhile, sinks quietly into the swamp.


Metrics Vs Philosophy

The deeper problem is that metrics are not neutral. They embody philosophical choices about what a product is for. Is YouTube
for maximizing minutes watched? Or is it for helping people learn, laugh, and connect? Is Gmail for sending the maximum number
of emails per day? (Heaven forbid.) Or is it for making communication smoother and more efficient?

Metrics force you to declare - often unintentionally - what you believe matters. A badly chosen metric is not just a bad number; it is a moral failure.

The case of Multidimensional Metrics

The sensible solution is to never optimize a single metric in isolation. Growth should be paired with quality. Engagement should be paired with trust.
Efficiency should be paired with user satisfaction.

This is, admittedly, less exciting. It produces dashboards with the thrilling complexity of a cockpit, and the arguments that ensue make
medieval theological disputes look straightforward. But it is the only way to prevent the slow rot that comes from chasing one number
at the expense of all else.

Google, to its credit, often balances metrics with user trust. Chrome doesn’t just optimise for speed; it also optimises for security.
Search doesn’t just optimize for clicks; it balances relevance, freshness, and credibility. (Not perfectly, of course — nothing is.)
But the principle is there: no single graph should be allowed to hijack the product


So what I think a Product Manager should do?

  1. Treat metrics like conversation starters - A downward trend should spark inquiry, not panic. Ask why before you act.

  2. Always pair metrics with qualitative output - Numbers tell you what. Users tell you why. Ignore either at your peril.

  3. Create guardrail metrics - For every growth measure, have a safeguard. For instance, if you optimize for session length,
    track user satisfaction to ensure you are not just addicting people.

  4. Periodically reset metrics - Like stale bread, metrics grow moldy over time. Revisit whether the numbers you track still
    reflect your product’s purpose.


TLDR -

If you cling to a single metric, you’ll end up with something that’s excellent at numbers and hopeless at being useful.

vskwdas June 2025