The right unit is the best thing ever

It’s one thing to build a team or company, it’s another thing to come up with a sane plan, and almost entirely another thing to find the right unit to that tells you where you really are.

Most companies have KPIs, in my experience, very few found the right unit of progress.

Most rely on Revenue, Users, CTRs, hours billed or deployments, tickets closed, "Burn rate", leads generated and these days, models trained; these are easy to count, easy to put on a dashboard and frequently useless.
The problem is not that the numbers are wrong, the problem is that they often describe activity rather than value.

A company can ship twenty features and improve nothing. a Sales team can generated five hundred leads and end up with no pipeline. An engineering team can increase deployment frequency while making the product less reliable. A platform can process billions of transactions per second that no customer particularly cares about.

Everything can be be moving, while the company remains stationary.

The unit beneath the metric

The right unit is smallest repeatable event in which your company creates real value.

For Airbnb, a listing is not the fundamental unit. A completed stay is much closer.
For a payments company, API calls are not the unit. Successfully completed payments for legitimate customer activity are closer.
For a security product, alert generated are a horrible unit, threats correctly detected and resolved without creating operational chaos are closer.
For an AI Coding product, lines of code generated are as meaningless as they are to measure human engineers: a task completed correctly and merged into production is a stronger unit.

I think the distinction matters because the unit determines what the company optimises, measures listing and people will create listing, measure token burn-rate and people multi-core-loop random data, measure alerts and you'll get much more alert than you can possibly verify.

Measure successful customer outcomes and the organisation has to confront the considerably harder question of whether anything useful actually happened.

A good unit contains value

The best units usually contain several things at once, I think the most important ones are the customer, action, and value of outcome.
"Monthly active users" contains almost none of that information, same as "Document created" or "Customer onboarded"..
"Customer completed the first production workflow without assistance" is much more useful, it indicates that value is being created; I also don't think you should expect all units to also point to your de-efficiencies -do we lack docs, is UI difficult, etc- that's more of a metric framework thing, like DORA.

The more closely the unit resembles the reason the customer pays you, the better.

companies naturally drift toward units that are easy to produce, easy to improve and easy to explain to investors, the easiest units to increase are often the least connected to the underlying business!

Worths mentioning that the unit is a model, not a law of nature, and no unit remains perfect (or even useful) forever.

The wrong unit forms the wrong company

Metrics don't just describe organisations, they shape them.

If engineers are measured by tickets closed, work will be fragmented into tickets.
If customer support is measured by resolution time, difficult conversations will be closed prematurely.
If recruiters are measured by hires, hiring quality will become somebody else's problem.

people are not stupid, they respond rationally to the system in front of them.

This is why a bad KPI rarely remains a reporting problem. It becomes an organisational design problem.
The metric influences planning. Planning influences incentives. Incentives influence behaviour. Behaviour determines the company you eventually become.

By the time the dashboard looks obviously wrong, the organisation may already have spent years optimising for it.

The right unit for your organisation

You'll need to identify your Atomic Event, which is the smallest event that would make a customer genuinely disappointed if it stopped happening, this event is often close the the real product.
For a consultancy, it is not an hour billed, it is a client decision improved or a business outcome achieved.

Your unit should alsop remain meaningful as the company scales, many metrics look sensible at an early stage because the founders personally understand the context behind every number. your first Ten customers are not just ten customers, the founder typically know who they are, why the bought, how much support they need and whether they are likely to remain, at ten thousand customers that intuition disappears.

for more complex businesses, no single number can be representative, a useful approach is to define one primary unit of value and then protect it with quality constraints, for a software this can be number of releases a day while constraining for reliability, retention, support burden etc.

Basically, the primary unit tells you what to maximise, the constraints tell you what you are not allowed to sacrifice.

Companies often obsess over the numerator and ignore the denominator:
"One hundred successful deployments" -> Out of how many attempts?
"Five hundred resolved incidents" -> Why were there five hundred incidents?
"Ten million in revenue" -> At what cost, from how many customer, what concentration and with what future obligation?

A metric becomes far more informative when exposed as a ratio.

A great unit aligns the company

Different function can recognise themselves inside of a great unit.

Product can improve it, Engineering can make it more reliable, Sales can bring in customers, Support can remove obstacles around it, Leadership can use it to allocate capital.

When every department has its own unrelated definition of success, the company becomes a collection of local optimisation problems, AKA KPI HELL.