Stop Building One Campaign per Offer: Let AI Pick the Promotion

Most sites run promotions as campaigns. One banner, one audience, one set of dates — and then the next offer gets a campaign of its own. Ten offers later there are ten campaigns, each one a guess about who should see it, two of them fighting over the same spot on the category page, and no clear answer to which one actually sold anything.

Promotions in Personyze works the other way round. You keep your promotions in one list, and for every visitor the AI picks the one most likely to convert them — learned from what visitors like them bought — then keeps learning from every view, click and purchase.

What a row in the list holds

A promotion is one of three kinds: an image banner, a text promotion (a line such as “Free returns for 60 days”) or a coupon with its code. Every row carries what a campaign used to carry:

• Dates — it starts and ends on its own, and the list sorts it into Live, Scheduled, Ended, Paused or Draft.
• Pages — the categories it belongs on, so the outerwear offer stays on outerwear.
• Who sees it — country or distance from a place, cart value, lifecycle stage, CRM and profile fields, categories viewed lately, device, or having redeemed another promotion. Set with pickers, per row.
• Language — one promotion in several languages; each visitor gets theirs, and the design’s own buttons (“Copy code”, “Ends soon”) follow the visitor’s browser language.
• Code — a shared code, a code that depends on the cart, or a code of the visitor’s own — from a pool you upload, or generated from a pattern such as SAVE-{6}.

Rows go in by hand, from a CSV or Excel file, from a Google Merchant Center promotions feed, or from a feed address that Personyze reads again on a schedule. Write with AI drafts headlines, variants to test and translations; its drafts wait in the Draft tab until you approve them.

How the AI picks

Only promotions that fit the visitor are in the running: their dates, pages, language and audience must all match. Among those, the widget chooses. The default learns two ways at once:

• Optimize shows each promotion, sees which ones earn clicks, purchases or the goal you choose, shows the winners more — and keeps trying the rest, so a better offer added next week still gets its chance.
• Personalized by audience learns, every six hours, which kinds of promotion work for visitors like this one — by device, traffic source, and new or returning.

A widget can also use a simpler rule. There are ten in all — Matches this category, From their recent categories, Matches their search, Personalized, Nearest location’s offer, Most popular, Popular near the visitor, Newest this week, Optimize alone and Personalized by audience — and each one tells you whether your data supports it yet.

One page, three visitors

The same list behaves differently for every business that uses it. Three examples, from the illustrations on the Promotions page:

Illustration: the businesses and visitors are examples, not customers.

Nobody built a campaign for “Spanish speakers on outerwear” or “trial users in their last days”. Each of those promotions is a row with its own audience, and the widget on the page chose among the rows that fit.

The details it looks after

Shown where it fits, wherever your visitors are

Promotions come with designs of their own — a hero banner, a slider, a text strip, a bar with a countdown, a coupon card and a coupon bar. Place one in a spot on the page, as a bar, a slide-in or a pop-up, and fire it on page load, on scroll, after a delay, on exit intent or when the visitor goes idle, with the same frequency caps as any Personyze campaign.

The same list feeds the other channels too: a personalized banner in your emails, web push, mobile apps, and the AI Chat, which can offer a visitor who asks about a discount the promotion that fits them — with its code.

Promotions or Banners & Pop-Ups?

They answer different questions. With Banners & Pop-Ups you decide who sees each message: you build the audience, choose the page and the moment, and the campaign shows exactly that. With Promotions you list the offers and the AI decides which one each visitor sees, learning as it goes. Many sites use both — a targeted pop-up for one audience you care about, and a promotions widget on every category page.

See AI Promotion ManagementHow the list, the AI pick, personal codes and the lift report fit together — for a store, a travel site or a SaaS product.

Questions people ask