Analytical use cases

Two uses of the same marketing-mix model, run backwards or sideways.

Not agentic workflows — direct uses of the fitted model: one answers a supply question by inverting the response curve, the other answers a spend-allocation question via the interactive Solver.

01

Inventory-stretch via marketing throttle

The usual question a marketing model answers is "spend more, get more". This is the opposite: stock is running low, the next production run is two weeks away, and the question is how much to slow demand so the shelves don't empty before then.

The reasoning chain

At the current sell-rate, stock hits zero in about 12.5 days — short of the next production run, 14 days out. Cutting daily marketing spend by roughly $93 brings demand down to the 107 units/day the remaining stock can sustain, saving an estimated $1,299 over the gap.

Under the hood

This runs the Solver's logistic saturation curve backwards. Instead of asking "what revenue does this spend produce," it asks "what spend produces this lower demand." The numbers above are precomputed from synthetic inventory data and the real fitted Meta response curve. Nothing here is interactive, by design — the point is to watch the model reason from a supply constraint to a marketing decision, not to hand a visitor sliders they'd have no real basis to set.

02

ROAS optimisation via the Solver

Spend is currently split across channels by habit more than by evidence. The marketing-mix model already knows each channel's response curve — the question is whether reallocating spend under the same total budget can beat the current split.

Starting from the current allocation and its measured ROAS, you can drag spend between channels and watch projected ROAS update live off the real response curves — or let the model solve for the best allocation under a fixed budget and compare it to your own attempt. This only works with more than one channel to move spend between. Marketing is over a third of this business's budget, so moving spend to the right channels changes the result.

Open the Solver →