Spot the opening
The agent watches sell-through, stock position, seasonality, and competitor promo. When a worthwhile promo window opens, it flags it with the expected lift.
Before a campaign ships, the AI checks real lift (new sales, not reshuffled ones), halo (what else rises with it), and cannibalization (what it steals from). It drafts the offer, the calendar, and the customer segments using real numbers from every store. Your team approves what goes to the till.
A weekly promo cadence replaces the "plan in Excel, ship in Canva, measure never" workflow.
The agent watches sell-through, stock position, seasonality, and competitor promo. When a worthwhile promo window opens, it flags it with the expected lift.
Mechanic, depth, duration, scope (category × store cluster), creative copy. Drafted against elasticity and halo — not a guess.
Customer passport segments (value, RFM, category affinity). WhatsApp, email, Telegram sequences wired directly to the till.
Incrementality per segment, per store. Not aggregate uplift — actual causal attribution against a hold-out.
Mechanic (% off, BOGO, bundle, tiered), depth, duration, category/SKU scope. The agent chooses what the data supports.
Pre-flight estimate of causal lift vs. cannibalization. Campaigns below a threshold never reach the calendar.
Which SKUs rise, which fall, how long the halo lasts. Shown per category, per store cluster.
WhatsApp Business, email, Telegram. Personalized copy drafted by the LLM in the tone of your brand.
Offers honored at Kwanta POS. Redemption and attribution write back to the customer passport.
Causal lift against a hold-out, per segment, per store. Not last-click — actual incremental revenue.
Promo reaches customers wherever they are — and closes the loop back to the till.
A hold-out group is defined per campaign, matched on pre-period behavior. Lift is the difference in purchase behavior between exposed and hold-out over the campaign window — causal, not correlational.
Yes. We train the LLM layer on your tone of voice, historical campaigns, and product vocabulary. Output is drafted for approval — your team always signs off before anything ships.
No. The passport shares notification preferences and consent, not purchase history. If two Kwanta retailers see the same customer, each one sees their own transaction history — only the consent layer is shared.
Talk to the founders. Get early access, honest pricing, and a direct line to the team.