Baseline sales: the non-promoted demand line

What baseline sales are

Baseline sales are the units you would have sold with no promotion running, the modeled non-promoted demand line underneath the actual sales curve. It is the answer to a question that has no clean observation: what would have happened anyway? When I ran the annual launch model at a natural-products brand, every promo evaluation I ever defended started here, because you cannot call a lift real until you know the line it lifted off of.

Picture a tortilla-chip SKU selling steadily at Sprouts, then spiking for four weeks during a $2.99 feature, then settling back. The total during those four weeks is easy to read off a SPINS export. The baseline, the part of that total you'd have earned at the regular $3.99 price with no display, is the hard part. It is not measured. It is estimated, and how you estimate it decides whether the promotion looks like a win or a waste.

How baseline gets estimated

The standard approach is a pre-period and post-period read, de-seasonalized. You take the weeks of clean, non-promoted selling before the promo and the weeks after it, fit a demand line through them, then extend that line across the promoted weeks. That projected line is the baseline. Syndicated providers like SPINS and Circana run a more sophisticated version of the same idea, modeling each SKU's normal week against price, distribution, and seasonality, but the intuition is identical: reconstruct the non-promoted trend and lay it under the promo spike.

De-seasonalizing is the step amateurs skip and regret. A sparkling-water baseline in July is genuinely higher than in January, so if you benchmark a summer promo against a winter base you'll credit the season's lift to your discount. Two more traps live here. Forward-buy pulls future baseline demand into the promo window, so the weeks right after a deal often dip below the true baseline as shoppers work through what they stockpiled. And pantry-loading does the same at the household level. A clean baseline accounts for both, which is why the post-period weeks matter as much as the pre-period ones.

Base versus promoted weeks, side by side

Take a single granola SKU at 90 Sprouts stores over an eight-week stretch. Weeks 1 through 3 and 7 through 8 run at the everyday $5.49. Weeks 4 through 6 run a $3.99 TPR. Here is the per-store-per-week read.

The five base weeks average 13.4 units/store/week, and that flat line is the baseline. Notice week 7 dips to 11, below the pre-promo run rate. That is forward-buy: shoppers stocked up during the deal and stayed home the week after, so the raw post-period understates true baseline. A careful model weights both shoulders rather than reading week 7 alone, which lands the baseline near 13 to 14, not 11.

With that baseline fixed, the promoted weeks become legible. The three promo weeks averaged 38 units/store/week against a ~13.4 base, and only the gap above baseline counts as incremental. Everything at or below the line would have sold anyway, at full price, with no markdown funded.

Why baseline is the metric everything hangs on

Baseline is the denominator for promotion analysis and the cleanest read on true demand health. Get it wrong and every downstream number inherits the error. Set the baseline too low and a mediocre promo looks heroic. Set it too high and you'll kill a deal that was actually paying. It also sharpens velocity: a SKU's base velocity, the non-promoted units/store/week, is the honest fitness signal, because a blended velocity that includes promo weeks flatters a product that only moves on deal.

The full mechanics of separating base from promo in SPINS data, including the de-seasonalization and the forward-buy correction, are walked through in post-promo lift and baseline in SPINS.

Where Scout fits

Estimating a clean, de-seasonalized baseline by hand, per SKU, per retailer, every time a promo runs, is the slow part of promo evaluation. Scout connects your SPINS or retailer data and models the non-promoted demand line so you can read base versus promoted weeks without rebuilding a regression in a spreadsheet each quarter. It estimates and surfaces the baseline. It does not run your promotions or fund your markdowns. That stays on your side of the line.

The short version

• Baseline sales are the modeled units you would have sold with no promotion, the non-promoted demand line under the actual curve.
• It is estimated, not observed, usually from de-seasonalized pre-period and post-period weeks, with corrections for forward-buy and pantry loading.
• Baseline is the denominator for every promo evaluation. Set it wrong and the lift number lies in whichever direction the error points.