Choosing Metrics That Help You Steer: Leading and Lagging Indicators

Most businesses have plenty of data. The harder question is which of it deserves attention.

One way I find helpful is to ask what we want measurement to do for us. Some measures exist to tell us what happened. That matters, but if we're using metrics to run a business rather than just report on it, we want something more: enough warning to change what happens next. We want metrics that help us steer.

Leading and lagging indicators are a useful way into this – not as two boxes to sort metrics into, but as a way of thinking about cause and effect.

The problem with outcomes

Take a SaaS business whose objective is to grow recurring revenue. ARR is the obvious measure. If the objective is growth, revenue has to grow; we can't declare victory on the strength of lots of leads, lots of sales meetings and lots of features shipped if none of it turns into money. So we need outcome measures.

The trouble is that the outcomes we care most about tend to show up late. By the time the quarterly revenue figure tells us we've missed the target, most of the quarter is already behind us. That's fine if measurement is about reporting performance. It's much less use if we want to manage it.

Now suppose we know from experience that hitting the revenue target needs about £600K of qualified pipeline, and that we normally convert around 40% of proposals. Halfway through the quarter we have £180K of qualified pipeline and conversion is running at 25%. We haven't missed the target yet, but we already know a good deal about where we're likely to end up. Those earlier measures are doing the job of leading indicators.

Leading and lagging are relative

It's tempting to treat this as two buckets: leading indicators over here, lagging over there. I don't think that holds up.

Consider a simple chain:

Sales conversations → Qualified opportunities → Proposals → Customers → Revenue

Revenue lags customer acquisition. Customer acquisition lags proposals. But customer numbers also lead revenue, and proposals lead customer acquisition while lagging opportunities. The same measure is leading and lagging at the same time, depending on what you compare it with.

So rather than two categories, I'd rather picture a chain of cause and effect. A measure leads whatever sits further down the chain and lags whatever sits further back. Once you look at metrics this way, it gives you a practical method for finding them.

Work backwards from the outcome

Start with something you care about and ask what causes it.

Say a rail operator wants to improve the dependability of its service. The obvious measure is the proportion of trains arriving on time. But punctuality is the result of a lot of other things, so ask what makes trains late: signalling problems, crew shortages, congestion, long dwell times at stations, rolling-stock failures.

Take rolling-stock failures and ask again. Perhaps preventative maintenance isn't being completed. Perhaps particular components keep failing. Perhaps faults aren't being caught early enough, or parts aren't there when they're needed.

Keep going and you end up with something that looks less like a list of KPIs and more like a map of the system you're trying to manage:

Preventative maintenance → Rolling-stock reliability → On-time running → Passenger confidence

The real system is messier than this, with more branches and more interactions between them. That's fine. We're not building a mathematical model of the railway, just understanding enough about cause and effect to make sensible choices about where to look.

And that's what the map gives us: choices. On-time running tells us directly whether we're delivering the service we promised. Rolling-stock failures give us an earlier warning of one thing that might stop us. Preventative-maintenance performance gives us an earlier signal still. Each says something different about the same outcome.

Earlier isn't necessarily better

You might conclude from this that we should keep walking backwards until we hit the earliest measurable thing. There's a trade-off, though. The further upstream we go, the more time we have to intervene and the less certain we are about the eventual outcome.

Back to the sales example. Revenue is definitive – we earned it or we didn't. Customer count is a strong predictor, though what each customer spends will vary. Proposals are a step further removed; qualified opportunities a step further again. Eventually we arrive at something like the number of outbound emails the sales team sends. That's easy to measure and easy to influence. We could double it tomorrow. It doesn't follow that revenue would double with it.

Every step away from the outcome adds an assumption about the causal link. So the most useful leading indicator isn't the earliest thing we can count. It's the measure that's early enough to give us time to act and still closely enough tied to the outcome that movement in one tells us something about the other.

Look for leverage

There's a second test I'd put alongside predictive value: can we do anything about it?

A measure can predict the outcome very well and still be useless as a management lever. External factors often tell us a great deal about what's coming while being entirely outside our control. Those are worth knowing about, but when we're picking the handful of measures that get regular management attention, I'm looking for the places where prediction and influence overlap.

If preventative-maintenance performance is slipping, the operator can do something about scheduling, capacity or practice before reliability follows. If qualified pipeline is behind plan mid-quarter, there's still time to change prospecting activity, look at where opportunities are being lost, or shift effort towards a particular market.

That's the point at which a leading indicator earns its place. It doesn't just warn us about the future; it gives us a chance to change it.

You probably need both

None of this makes lagging indicators less important. Relying on leading indicators alone would be dangerous. If we measured preventative maintenance and stopped measuring reliability, we could get very good at maintaining trains without ever knowing whether passengers were getting a better service. The outcome measure keeps us honest.

So I don't choose between them. I want a small number of each. The lagging measures tell us whether we're getting the outcomes we wanted; the leading measures tell us whether the conditions that tend to produce those outcomes are developing as expected. Together they're closer to an instrument panel than a performance report – some instruments tell you where you are, others where you're heading.

The point is to shorten the feedback loop

One last thing, because none of this is much use if the metrics just sit on a dashboard.

Suppose qualified pipeline turns red halfway through the quarter. That only creates value if someone notices, asks why, decides whether to intervene, and then does something differently. Measure, understand, decide, act, measure again. That's the feedback loop, and the earlier we can spot a meaningful deviation, the smaller the correction needs to be.

This is why leading indicators matter so much in a management system. They're not substitutes for the outcomes we care about; they're the signals that let us see those outcomes coming. Measure only the end of the chain and you'll get very good at explaining what happened. Choose sensible measures along it and you stand a much better chance of changing what happens next.