Step 06 — Results

What each campaign taught the model.

The results below are illustrative placeholders from the fictional retailer, shown to demonstrate the learning loop the production system would run. Each past campaign is measured, compared against the others, and turned into a plain lesson that shapes the next cycle.

Campaign results

Six campaigns, measured

Reach, response, redemption, repeat purchase, revenue, and margin for every past campaign on the fictional dataset.

Campaign	Segment targeted	Sent	Audience	Response	Redemption	Repeat purchase	Incremental revenue (EGP)	Margin impact	Reactivated
Loading campaign results…

Comparison

Response and repeat purchase, side by side

The clearest signal is not reach but whether customers came back: response rate against repeat-purchase rate across the six campaigns.

What the model learned

One lesson per campaign

The learning loop

How learnings feed the next cycle

Results adjust segment priority. When a segment responds and comes back, it earns a higher place in the next ranking; when reach is spent on customers who do not return, that segment moves down. The priority board you started from is meant to be re-scored by what actually happened, not fixed once.

Offer-type effectiveness reweights the action matrix. A no-discount, recognition-led campaign that protected margin argues for more of the same; a blanket price cut that lost margin argues against it. Each result nudges the weight the model gives an offer type for a given reason.

Channel response updates channel fit. The channel that earned replies for a segment gains confidence for that segment next time; the one that went unread loses it. Over cycles, the channel plan bends toward where each group actually engages.

In this prototype the loop is illustrated rather than live: the weights are shown, not retrained. In production it would run on your own campaign outcomes, on a cadence your team sets, with every change kept auditable.

Run the flow again Back to dashboard