You can answer detailed questions about what happened. Revenue by month, projects delivered, clients gained and lost, hours spent. The historical record is genuinely good.
Now ask what next quarter looks like, and the answer changes register entirely. It becomes a feeling, hedged, with a couple of specific deals attached.
Not because you're careless. Because there's nothing in the business to consult.
Here's the version that should actually bother you: every serious problem you've had in the last three years announced itself in the numbers only after it had fully happened. The client who left showed up as churn. The quiet quarter showed up as revenue. In every case the data confirmed the thing after it was too late to act on — and you experienced that as bad luck rather than as a property of what you measure.
Why the free numbers are all about the past
Trace where a lagging metric comes from.
An invoice exists because money had to move, and moving money requires a record. A signed contract exists because agreement requires an artifact. Churn exists because ending something requires an action in a system.
Every one of those is recorded as a byproduct of the business functioning. Nobody built the capture — the capture is the transaction.
Now trace a leading indicator. A conversation that didn't happen. A follow-up nobody sent. A client whose replies have slowed from a day to a week. A project whose next step hasn't been scheduled.
None of those involve a transaction. So none of them generate a record. So none of them are visible to any tool you own.
Every metric that's easy to obtain is a lagging one, and it's easy to obtain because the thing already finished.
That's the whole mechanism, and it explains why the effort is so lopsided. Lagging metrics arrive whether you want them or not. Leading ones require somebody to define a behaviour, decide how it'll be captured, and maintain the capture — for a number whose only value is that it might save you from something that hasn't happened yet.
Which is also why being told to track leading indicators is useless advice. You didn't fail to think of it. It costs something and the lagging ones didn't.
The four that work
You don't need a system of leading indicators. You need three or four, and they should be behaviours you can already name because you already know what precedes trouble in your business.
For most founder-led service businesses, the same four keep working.
Net new conversations started this week
Leads revenue by roughly a quarter, and it's the earliest signal you have of a drought that hasn't arrived yet. It also has the useful property of being entirely within your control, which most metrics aren't.
Open opportunities with no scheduled next step
This is the pipeline's decay rate made visible. A deal with nothing scheduled isn't waiting — it's decaying, and the count of them is the single most actionable number in a small pipeline.
Days since a client last heard from a human
Measured against that account's own normal interval, not an absolute threshold. A client who has always been in touch monthly hasn't drifted by being in touch monthly. This leads churn by months, and it's the cheapest early warning available.
Work items past their promised date
Leads the delivery problems you'll hear about in three weeks, and it leads them from the inside — you find out before the client does, which is the entire difference between a conversation and a complaint.
Notice that all four are counts of things not happening. Leading indicators in a service business are almost always absences, which is why they're invisible to systems built to record events.
The capture is the actual work
This is where it dies, so be honest about it up front.
Days since last contact requires that contact is logged. Next step scheduled requires that next steps live somewhere structured rather than in someone's head. Past promised date requires that a promised date was recorded at all.
Each of those is a small change to how work is done. And each will be resisted — correctly — if it's imposed as reporting overhead, because a logging task that returns nothing to the person doing it has a shelf life of about six weeks.
So the design constraint is absolute: the capture has to be a byproduct of the work, not an addition to it.
If logging a client conversation is a separate act, it won't happen. If the conversation happens in a system that records it, it's free. If setting a next step is a form somebody fills in afterwards, it decays. If it's a field on the thing they're already updating, it holds.
Build the systems as you do the work, not instead of the work. Every leading indicator that survives in a real business is one where the capture rides along with something people were doing anyway.
Start with one
Don't build four. Pick the behaviour that most reliably preceded your worst surprise of the last two years, instrument only that, and run it for a quarter.
One leading indicator that actually fires is worth more than four you designed and never captured.
And run it manually first. A tally in a notebook is a valid leading indicator. Doing it by hand for a month tells you whether it's the right one, what threshold matters, and what the capture would need to contain — all of which you'd otherwise be guessing at while building.
Steering by the mirror
Every intervention happens at the most expensive moment. A drifting client is a phone call in month four and a lost account in month seven. Lagging metrics only ever show you month seven. Same problem, different price, decided entirely by when you found out.
You can't steer, only react. Planning against outcomes means adjusting after results arrive, which in a service business is a lag of one to two quarters. The business is being driven by looking in the mirror, and the mirror is honest.
Forecasting becomes a personality trait. Without leading indicators the only forecast available is your intuition — which can't be checked, can't be delegated, and when it's wrong, nobody can say why.
Your team stops doing the things that produce outcomes. If only outcomes are measured, only outcomes get attention. The follow-up, the check-in, the early conversation become invisible work nobody is credited for. They stop happening — which shows up as an outcome, two quarters later.
The trap inside a leading indicator
One warning, because a leading indicator that goes wrong goes wrong more expensively than a lagging one.
Lagging metrics describe the past, and the past can't be gamed. Leading metrics describe behaviour, and behaviour responds to being measured — which is the whole point and also the risk.
Measure net new conversations started and you will get more conversations started. Some of those will be conversations that shouldn't have happened, with people who were never going to buy, because the number rewards the act rather than the judgment behind it.
That isn't an argument against the metric. It's an argument for pairing it. A volume indicator needs a quality one next to it — conversations started alongside the proportion that reach a second conversation — because the pair is much harder to game than either alone.
The same applies to contact frequency. Days since a client last heard from a human will produce more contact, and a business that hits the number with four templated check-ins has satisfied the metric and damaged the relationship. Pair it with something that reflects whether the contact was any good.
A leading indicator is an instruction to your team, whether you meant it as one or not. Write it as though somebody will follow it exactly.
The layer you're actually missing
The three layers are the pulse, the scoreboard and the soul. Most businesses have some version of the scoreboard: what's working over time.
What's missing is the pulse — what's happening right now. And a pulse assembled from invoices isn't a pulse, it's a slow history.
That's the specific gap, and it's the one that requires deliberate capture, because right now in a service business is made of behaviours rather than transactions. Nothing processes a conversation. Nothing invoices a follow-up.
Which is the whole point of the Telemetry pillar: stop reacting, start recognizing. Recognition means seeing the pattern while it's forming — and a business measuring only transactions has, by construction, forfeited that.
What a leading indicator is not
Two things get mislabelled as leading indicators often enough to be worth separating out.
A faster lagging metric is still lagging. Weekly revenue instead of monthly revenue is more current and it's still a report on completed transactions. Frequency isn't the axis — what the number is of is the axis. A daily churn figure tells you about people who have already left.
A forecast is not an indicator. A weighted pipeline number is a prediction assembled from judgments, and its accuracy depends entirely on the judgments. That's useful and it isn't a measurement. The distinction matters because a forecast can't surprise you — it only ever contains what somebody already believed.
A real leading indicator has one property: it changes before the outcome does, for reasons outside anyone's opinion. Contact intervals lengthen before a client leaves. Deals stop having next steps before revenue drops. Those movements happen in the world whether or not anybody has formed a view about them, which is exactly what makes them worth watching.
The reconstruction to run first
Take your worst surprise from the last two years and reconstruct it. When did it become true? What behaviour changed first? Would anybody have seen it?
That behaviour is your first leading indicator. Count it by hand for a month before building anything.
If you know what you should be watching and there's nowhere for it to come from — because the capture would be a task nobody sustains — that's the constraint. Worth naming which of the seven pillars is actually binding before instrumenting anything.
Get The OPERATE Report →