Show each shopper what they may want.
AI Personalisation suggests products, content blocks and offers for each visitor or customer from what they browsed and bought, inside your shop, email tool or app. You write the rules before anything is shown, covering which items may be promoted, which offers keep a margin floor and which customers must never be contacted. Marketers review rules, not shoppers.
A regular is treated like a stranger. Last spring she bought a wedding jacket, and every mailer since has been the same offer sent to the whole list, so most of it is ignored.
Personalisation here means a short, explainable list. It holds items often bought with this one, the category the customer keeps returning to and an offer not yet used, and nothing else. A merchandiser keeps control through rules and exclusions, and every suggestion carries a plain reason, so staff can judge it at a glance.
The problems this is meant to ease.
What AI personalisation does.
What this looks like in a working week.
The people who open this every day.
From first conversation to daily use.
Discovery
Find what you already hold. We sit with the e-commerce manager, the marketer and the merchandiser and list where orders, browsing, consent and stock live today, and who owns each list.
Data readiness
Clean lists come first. Order history is matched to customer accounts, products are given categories and margin limits, and consent records are gathered into one list so opt-outs are honoured before any suggestion.
Trial on one category
Start with one category. Suggestions for one category are generated from past orders and shown only to the merchandiser, who compares them with her own picks and hides the pairs she dislikes.
Staff review of suggestions
Staff judge the reasons. The merchandiser, a store manager and an email marketer read a sample of suggestions with their plain reasons, mark the odd ones, and those marks become exclusions.
Going live
Switch on one place first. Recommendation rows go on one page first, then email blocks follow after a preview, and each message that starts from a trigger waits for the marketing head’s approval.
Reviewing afterwards
Read the results with care. After a cycle the team compares clicks, orders and opt-outs with the figures taken before go-live, removes rows that only added clutter, and revises the rules for offers.
How you would know it is working.
Record where you stand before it goes live, so the change is measured against your own figures.
- Click-through rate counts clicks on a recommendation row or email block against the times it was shown, and the baseline is the rate on the same placements before go-live.
- Conversion uplift is the change in orders from shoppers who saw a suggestion compared with a held-back group, and the baseline is the order rate of that group.
- Customer engagement counts repeat visits, replies and saved items per customer, with the same counts from your shop and email records before go-live serving as baseline.
- Opt-out and complaint counts are logged for every send, and the baseline is the count from the last campaigns mailed to the whole list before personalised blocks began.
- Margin on recommended orders is checked against the floor set in the rules, and the baseline is the average margin of comparable orders in the three months before go-live.
Where a person stays in charge.
Rules outrank suggestions
The merchandiser has the last word. A rule on margin floors, offer frequency or blocked pairs always wins over a suggestion, and nothing in the setup changes a list price on its own authority.
Consent decides who is included
Opt-outs are applied first. Customers who opted out or asked not to be contacted are removed before any list is built, and browsing is linked to an account only where permission exists.
Thin data gives thin suggestions
Small shops see bestsellers. A shop with few repeat buyers or a small range gets little beyond popular items, and the page marks such suggestions as weak instead of guessing.
Privacy law differs by country
Tracking rules are local. Consent for tracking, profiling and direct messages is set by the country each customer lives in, so agree the capture and storage steps with your own privacy adviser. It prepares data, not legal advice.
People review the output
Previews come before sending. Email blocks and trigger messages are shown on sample recipients first, and the marketing head approves each campaign before the first customer receives anything.
It is one part of a system, not an island.
A module earns its place by what it passes to the next one. These are the connections we set up most often.
- Online shop or product catalogue
- Point of sale and order history
- Email and messaging tools
- Loyalty and offers module
- Website or app, for recommendation rows
Products that include it.
How we would put it in.
Questions owners ask about AI personalisation.
How much history do we need?
Enough repeat customers for patterns. Many orders across many items count, not a few big buyers, and a shop with a small range and mostly one-time buyers will see little beyond bestsellers.
What do unknown visitors see?
Bestsellers or the merchandiser’s picks. There is nothing personal to base a suggestion on yet, and the software does not pretend otherwise, so suggestions improve only as consented browsing and order data builds up.
Is customer tracking allowed?
It depends on the country. Privacy and marketing consent law varies between countries, so your own lawyer approves how consent is captured and how suppression works, while the software only prepares the data.
Can it raise prices for loyal customers?
No, and we would not build it. Offers follow rules your team writes and approves, including a margin floor, and nothing in this setup changes a list price on its own.
Do we need a developer?
Only for rows on your site. Rows on your website or app need a small change to the shop code, which we handle, while email blocks usually work through your existing email tool.
How does it explain a pick?
Each pick carries a plain reason. Staff see notes such as viewed three times or bought with a similar item, so they can judge a pick and the merchandiser can hide it.
When is this not worth doing?
When customers buy only once. A shop that sells large items once in several years gains little from history-based suggestions, and a few hand-picked rows are simpler.
What effort does setup need?
Mostly data tidying and rules. Someone must match orders to customers, set categories and margin limits, and write the offer rules, after which connecting the shop and email tool is the smaller part.
Businesses that ask for this.
Modules that work with it.
Tell us how you handle AI personalisation today.
A spreadsheet, a register or another system: say which, and we will tell you plainly what is worth changing.
- No obligation
- A reply within one business day
- Your details stay private
Start with your requirements
Seven short steps that tell us what you need. An engineer reads them before getting in touch.
Share your requirements