Personalisation that respects intent
Recommendations built from session behaviour and purchase history, tuned for margin and stock position as well as click-through - so you stop promoting the thing you cannot ship.
Commerce
Merchandising, demand and fraud, running themselves
Online retail runs on decisions made thousands of times a day: what to show this visitor, what to reorder, what to price, what to flag. Each is small; together they set your margin. We automate them against your live data instead of last quarter’s assumptions.
From$9,000
One component live - recommendations, search ranking, demand forecasting or fraud scoring.
Indicative starting price. The fixed fee for your scope is quoted after the two-day diagnosis, before any build begins.
What makes it work
Three things we insist on
Forecasting you can order against
Demand modelled per SKU and location with seasonality, promotions and lead times factored in, surfaced as reorder points rather than a chart somebody has to interpret.
Fraud caught without punishing good customers
Risk scored per order against behavioural and network signals, with thresholds tuned to your actual chargeback cost and false-positive tolerance rather than a vendor default.
Your catalogue and order history The data you already have, which is usually messier than the plan assumes
Buy the app or build the integration? An app is faster and stops where its roadmap stops. Custom costs more and is yours. The volume where that flips is lower than most teams expect
Wired into the store, not beside it Shopify or WooCommerce, with inventory, pricing and fulfilment state read at the moment of the answer
Judged on revenue per session Not on model accuracy. A recommendation nobody clicks is a correct answer to the wrong question
Capabilities
What is actually included
- 01
On-site recommendations and search
Product discovery ranked per visitor across home, category, product and cart.
- 02
Demand and inventory optimisation
Per-SKU forecasting, reorder points and overstock warnings before markdown season.
- 03
Margin-aware dynamic pricing
Price and promotion recommendations bounded by rules you set - never a black box on your margin.
- 04
Triggered lifecycle campaigns
Browse, cart and replenishment moments caught automatically and handed to your email and ads stack.
- 05
Fraud and abuse detection
Order risk scoring, return abuse patterns and account takeover signals.
In detail
What this covers, specifically
A category name is not a scope. These are the individual pieces of work inside this practice - take the two that apply to you and ignore the rest.
- 01
Personalised recommendations
Product ranking computed per visitor across home, category, product and cart pages, from session behaviour and purchase history, tuned for margin and stock position as well as click-through.
- 02
Semantic on-site search
Search that matches intent rather than keywords, handling misspellings, synonyms, natural-language queries and attribute filters ("waterproof jacket under $100").
- 03
Demand forecasting
Predicted units per SKU and per location, with seasonality, promotions, price elasticity and supplier lead times factored in, surfaced as reorder points rather than a chart.
- 04
Inventory and replenishment
Safety stock levels, reorder timing and overstock warnings calculated per line, so you find out about a stockout before your customers do.
- 05
Dynamic pricing
Price and promotion recommendations bounded by rules you set (floor margin, MAP, competitor position), never a black box acting on your margin unsupervised.
- 06
Markdown optimisation
When to discount ageing stock and by how much, balancing recovered cash against margin given the remaining sell-through window.
- 07
Catalogue enrichment
Product titles, descriptions, attributes, categories and tags generated or corrected at scale, which is usually the cheapest fix for poor search and filtering.
- 08
Visual search and similar items
Finding products from a photo, and surfacing visually similar alternatives when the exact item is out of stock.
- 09
Size and fit guidance
Recommending a size from the customer’s purchase and return history plus garment measurements, which directly reduces the most expensive category of return.
- 10
Fraud and payment risk
Order risk scored at checkout from behavioural and network signals, with thresholds tuned to your actual chargeback cost and false-positive tolerance.
- 11
Return abuse detection
Identifying serial returners, wardrobing and refund fraud patterns that individually look legitimate and only appear across order history.
- 12
Churn and lifetime value
Predicting which customers are worth acquiring and which are about to lapse, so retention and acquisition spend is directed by value rather than recency.
- 13
Basket and bundle analysis
Finding what genuinely sells together and designing bundles and cross-sells from it, instead of "customers also bought".
- 14
Review and sentiment mining
Extracting recurring product complaints from reviews and support tickets, which is often the fastest route to a return-rate problem.
- 15
Shopping feed optimisation
Cleaning and enriching product feeds for Google Shopping, Meta and marketplaces, where poor attributes quietly cap paid performance.
What you receive
Concrete artefacts, not a slide deck
- Recommendation and search ranking, live on site
- Demand forecast and reorder point model
- Pricing and promotion decision support
- Fraud scoring integrated at checkout
- Merchandising dashboard with margin view
01 Diagnose Days 1-2
02 Prove Week 1
03 Integrate Weeks 2-3
04 Operate Ongoing
Around three weeks end to end. That comes from scoping tightly to one workflow - not from skipping a phase. Each still ends in evidence you can check.
Further reading
Written up in more detail
Let's talk
Find the margin already in your data
Share twelve months of order and catalogue data and we will run a read-only analysis: where discovery is failing, which SKUs are consistently mis-forecast, and what that is costing per month.
Related practices