Free Tool
Decompose — Which Lever Moved Sales
household penetration · three-lever waterfall · Cinderhaven series #3
Sales went up. That tells you nothing about why. Decompose splits a period-over-period sales change into the only three levers that can move it — buying households, purchase frequency, spend per trip — and says, in plain language, whether the growth is real or erosion in disguise.
Every sales change is some combination of three things: more or fewer households bought you, your buyers came back more or less often, and they spent more or less each trip. That is the whole list. The first lever — household penetration, the How Brands Growmetric — is the one most POS dashboards cannot show, because it needs panel data, not scan data. A brand can post growing revenue while its buyer base shrinks, and the dashboard will report a good quarter.
The worked example is Cinderhaven Provisions — a fictional $25M specialty food brand on a seed-locked synthetic household panel. The erosion story in the demo is not scripted in: the panel is generated with realistic mechanics (prices rise, price-sensitive households buy less, some lapse) and the decline is computed from the resulting transactions. The tool finds the story; it does not assert it.
Growth that is actually erosion
Comparing Cinderhaven’s two most recent like quarters, sales are up $1,954. Good news, on the surface. The waterfall says otherwise. Buying households: −$2,459 — fewer households bought the brand at all, with penetration down 2.2 points. Purchase frequency: −$3,597 — the households that stayed came back less often. Spend per trip: +$8,010 — prices went up, and the buyers who remained absorbed it.
Two levers are bleeding; price is carrying the entire number and then some. The brand is held up by a smaller, more loyal, higher-spending core while its base quietly shrinks. That is erosion wearing growth’s clothes, and it is exactly the pattern a scan-data dashboard reports as a good quarter.
The verdict the tool writes for that comparison, computed from the data and never scripted: “Sales rose $1,954 from 2024-Q4 to 2025-Q4, driven mainly by higher spend per trip.” Drive the same number with new households instead of price, and the verdict reads differently.
The three bars sum to the exact dollar change
The attribution is an exact Shapley decomposition: each lever is credited its average marginal effect across every order the three changes could have happened in, so the waterfall reconciles to the sales delta to the dollar. No interaction term hides in a rounding gap. Underneath the waterfall, the buyer flow shows the households behind the penetration number — new, retained, and lapsed, quarter over quarter — so you can see whether you are replacing the households you lose or just squeezing the ones who stay.
When price is doing all the work, the next move is not another price increase; it is winning households back before the base gets too small to hold the number up. Whether the households you do win stick is the next question — the integrity check on this tool is Leaky Bucket: #3 counts the buyers, #4 says whether they stuck. Both sit on the same panel behind the three meanings of penetration most dashboards conflate.
What you get
A three-lever waterfall that bridges Period A sales to Period B and reconciles exactly to the change, the new/retained/lapsed household flow behind the penetration number, and a board-ready verdict sentence a CFO cannot misread. It takes “sales moved” — which tells you nothing — and turns it into “sales moved because of this, so pull that.”
The lever the dashboard was never going to show
The full argument — how sales growth on price can hide a household penetration decline, and why the three-lever split catches erosion disguised as growth — is in Sales decomposition: the growth that hid a household penetration decline →
Start in writing.
A few minutes by form — no call. Tell me your last two like periods — sales up or down, and by how much — and I’ll tell you what data you need to split the change across the three levers, and what it costs to keep not knowing which one moved. No deck, no obligation.