Your Ecommerce Forecast Should Be Wrong (On Purpose): Forecasting as a Steering Tool

The first thing I look at when I open a new client's ad account isn't the ROAS. It's whether there's a forecast pinned anywhere, and what they did with it.

Most of the time the answer tells me everything. There's a single number for the year, built back in January, sitting in a spreadsheet nobody has opened since. One figure, treated like a prophecy. By the time I'm looking at it in June, reality has drifted so far from that number that the whole document has quietly become useless.

Here's my take, and it's the thing I end up saying in almost every kickoff: a forecast that's trying to be right is close to worthless. A forecast that's built to show you where you're wrong is one of the most useful tools you own.

That sounds like word play. It isn't. It changes what you build, how often you look at it, and what you do when the month goes sideways.

A forecast is a diagnosis, not a prediction

In ecommerce, your guaranteed future revenue is zero. I mean that literally. You don't own any customers. Nobody who bought from you last month is obligated to buy again. If you stopped all your marketing tomorrow, the revenue doesn't coast, it falls off a cliff.

That's the part that makes our forecasting fundamentally different from a software business, where a chunk of next month's revenue is already locked in by contract. Our revenue is a choice. It's the sum of actions we decide to take - the emails we send, the campaigns we launch, the offers we run. It is not an extrapolation of the past unless we choose to repeat the past exactly, which no brand ever does.

So when you forecast a single confident number and sit back to watch, you're playing guess-the-jar. You wrote down a figure and you're hoping. There's no mechanism connecting the number to the work.

What I want instead is a forecast I can hold up against reality every month and ask one question: where am I importantly wrong? Not "did I nail it." Where's the gap, how big, in which direction, and what does it tell me to do next. The number is a trellis. The plant is the actual business growing against it.

Build the number you feel great about, then stress it

I'm not saying don't put a stake in the ground. I'm saying build it honestly, in layers, from the most predictable thing outward.

The base layer is your existing customer base. The cohorts you've already acquired will produce some revenue next month whether you do anything or not. That's the most stable part of the cake, and if you have decent cohort data you can model it with real confidence.

On top of that sits what you can reach for free - your email list, SMS, organic search positions you hold, your social audience. Less predictable than the base, but still yours to influence.

And the top, thinnest, wobbliest layer is paid acquisition. CPMs you don't control, conversion rates you can't promise, an auction that does what it wants. This is the layer everyone wants to lead with, and it's the one you should trust the least.

Most struggling brands have that cake upside down. The biggest slice of their forecast depends on the most volatile, least controllable channel. So the moment CPMs spike, the whole number falls apart and they've got no idea why.

Build it the right way up. Then here's the bit people skip: forecast a number you genuinely feel great about, and layer an upside case on top of it. One you'd be thrilled to hit. The point isn't to pick the middle and pray. It's to hold a range you understand.

Don't bet on one number, scenario-plan against the things that hurt

Once you've got your honest base, the real value comes from asking what would make you wrong. Not vaguely. Specifically.

What happens to this forecast if CPMs jump 20% in your peak weeks? What happens if there's a broad demand pullback and conversion rate softens by half a point? What happens if a key product goes out of stock for ten days?

When you model those, you stop having one forecast and start having a few. I think about it as three numbers. There's the stretch target - the one I'd tie a team bonus to, deliberately set above what's likely so people have to solve something new to reach it. There's the middle case, the one I'd actually plan cash flow against, because it's what I honestly expect. And there's the floor, the lowest acceptable outcome, the number I'd show a board or a bank as the bottom of the range.

Three numbers, each doing a different job. The day a CPM spike or a soft launch knocks you off the middle case, you already know which lever you're reaching for, because you thought about it cold rather than in a panic.

The monthly gap is where the actual work lives

Here's the cadence I'd run, and it's genuinely what we do with clients month to month.

At the start of the month, the forecast sets the expectation - not a prediction, an expectation of what needs to happen. Then you watch actuals against it. Not at month-end, when it's too late to do anything. Weekly. On the big days, daily.

And when the gap shows up - it will, every single month, in some direction - you treat it as information, not failure. Behind target by mid-month? That's not a number to feel bad about. It's a signal telling you to act: more sends, a sharper offer, fresh creative into the campaigns that are still working.

To put this in perspective with a made-up but very ordinary example: say a client is forecast to do around $300k for the month and by the 18th they're pacing closer to $230k. The old way, you find that out on the 1st of next month and shrug. The steering way, you saw it on the 12th, you understood it was soft email plus slow spend, and you had two weeks to close it. That's the entire difference. Same forecast, completely different outcome, purely because of what you did with the gap.

It's like driving with a maps app. The destination doesn't change. But you're going to hit traffic, you're going to get rerouted, and the value isn't in having guessed the perfect route in advance. It's in how fast you see the delay and adjust.

The contingency you fire when you're pacing behind

This is the move I'd push you to build in advance, because nobody makes good decisions in the last week of a bad month.

Plan to be wrong. Going into the month, you have your main plan - ideally one that doesn't lean on discounting at all. But you also pre-build a contingency. A backup offer, a campaign, something with a clear call to action that you keep in your back pocket and do not touch unless the pacing tells you to.

Then if you hit the final stretch meaningfully behind, you pull the chute. You fire the contingency. Because it was designed ahead of time, when you were calm, it's a better offer than anything you'd improvise at 11pm on the 28th watching the number stall.

That's the whole philosophy in one move. You assumed you'd be wrong, you decided in advance what you'd do about it, and the forecast told you when.

So if your forecast right now is a single number you wrote months ago and haven't looked at since, I'd gently suggest it isn't doing any work for you. Pull it back up this week. Hold it against where you actually are. Find the gap. That gap is the most useful thing the forecast was ever going to give you - and it's the only part most people throw away.