Article 3 of 8 · Predictive Intelligence · 10 min read
The Drivers of a Forecast: Finding What Actually Moves Your Numbers
A trend line projected forward tells you where a number is heading if nothing changes. A forecast built on drivers tells you why it's heading there — and what would change it. The second is a decision tool; the first is just a ruler laid against the past.
Two companies forecast next quarter's sales. The first draws a line through last year's numbers and extends it: "we grew 8%, so plan for 8% more." Clean, fast, and blind. The second asks a harder question — what actually drives our sales? — and finds that most of the movement comes from three things: the number of qualified leads about sixty days earlier, a strong seasonal pattern, and one large client's ordering cycle. The second company doesn't just have a number. It knows what moves it, which means it can see a dip coming two months out, and can answer "what happens if we change something."
That's the leap this article is about. In the series so far we've said a forecast is a statement about what's likely, and that the first job is separating signal from noise. This is the next step: once you know a pattern is real, you ask what's behind it. Because a forecast that knows its drivers is worth far more than one that only knows its own past.
Extrapolation vs. driver-based forecasting
Most casual forecasting is extrapolation: take the shape of the past and extend it forward. It assumes tomorrow is a continuation of yesterday — which is fine until anything changes, at which point it's silently wrong. Driver-based forecasting does something fundamentally different: it connects the outcome you care about to the things that actually move it, and projects the outcome by projecting its drivers.
The difference isn't academic. An extrapolated forecast is a dead number — it can only ever say "more of the same." A driver-based forecast is alive: it responds to the world, warns you early, and tells you which levers you can pull.
Why drivers are worth the effort
Finding the real drivers of a number takes more work than drawing a line. Three payoffs make it worth it.
- Early warning — If leads drive sales with a 60-day lag, a drop in leads warns you two months before sales fall — while you can still act. The driver is a leading indicator of the outcome.
- "What if" becomes answerable — Know the drivers and you can ask: what happens to sales if we double marketing, or lose the big client? The forecast turns from a prediction into a decision tool.
- Wrong forecasts explain themselves — When a driver-based forecast misses, the drivers tell you why — a driver moved unexpectedly. An extrapolation just shrugs: it never knew why it was right in the first place.
An extrapolation can only say "more of the same." A forecast that knows its drivers can warn you, answer "what if," and explain itself when it's wrong.
How to find the real drivers
1. Start from the business logic, not the data
Before touching a spreadsheet, ask what plausibly causes this number. For sales: demand, price, marketing, pipeline, seasonality, maybe a macro indicator or even weather. You're looking for a short list of candidates with a real mechanism — reasons they would move the outcome — not a fishing expedition through every column you happen to have.
2. Check whether the driver leads the outcome
A true driver usually moves before the outcome, and does so consistently. Leads rise, then sales rise sixty days later. Look for that lagged, repeatable relationship in the data — not just that two lines wiggle at the same time.
3. Beware correlation that isn't causation
Two numbers can move together for reasons that won't help you forecast: both are driven by a hidden third thing, or it's plain coincidence. The guard against this is the mechanism test — is there a believable reason this drives that? — plus checking whether the relationship holds on data you didn't use to find it. A correlation with no mechanism is a trap waiting to spring.
4. Prefer a few real drivers to many weak ones
Three drivers you understand beat twenty you found by trawling. Every extra driver you bolt on to squeeze out a better fit is a chance to fool yourself — a point we'll take apart in the next article. The strongest forecasts rest on a handful of drivers with clear mechanisms.
This is where econometrics earns its place: estimating, from real data, how much each driver moves the outcome and how confident we can be in that — so the relationship isn't a guess, it's a measured quantity with a stated uncertainty.
A quick example
Say you've established that qualified leads drive sales with a two-month lag. This month, leads come in 20% below normal. An extrapolated sales forecast notices nothing — sales are still fine this month, so the line still points up. The driver-based forecast flashes a warning now: expect a sales dip in about two months. One of these companies spends the next eight weeks fixing the lead problem. The other finds out when the sales number finally drops, too late to do anything but explain it.
Find out what actually moves your numbers.
Strategiq Softworks builds predictive tools and custom analytical platforms grounded in peer-reviewed research — identifying the real drivers behind your key numbers, estimating how much each one matters, and turning that into forecasts you can steer by. Need a driver-based forecasting model built for your business? We take a limited number of consulting engagements each quarter.
Signal vs. Noise: What's Actually Worth Predicting · Why Good Models Still Fail: Overfitting and the Illusion of Precision