Skip to main content
GullySystem

Recommendation Engine Development and Testing

Recommendation engine development covers the work between an idea and something ready to show a customer. Catalogue and order history are prepared, an approach is chosen and results are measured against a plain baseline.

The data it runs on

Two things are needed. One is a catalogue with consistent categories, sizes and codes, which is rarely how a catalogue starts. The other is order history where an item is one item and not three codes for the same thing. Most of the effort goes here, with your team, before any model is trained. Skipping it produces a model that recommends confidently from a muddled picture.

Problems the build has to solve

The new visitor

Someone with no history gets category-level and seasonal suggestions until enough behaviour exists to work from.

The new product

A line with no sales yet is placed by its attributes and given deliberate exposure. Otherwise it never gets a first order.

Repeat purchases

For consumables, the useful suggestion is the reorder at the right interval, which is a different calculation.

Narrow ruts

Guardrails stop the model showing the same few items repeatedly, a common complaint after the first few months.

Measured against a plain baseline

Before the model is judged, a simple rule is run alongside it: most bought in this category this month. Sometimes the plain rule wins. That is worth knowing before you carry the cost of a model. Testing uses held-back history first, then a live split once the engine is in front of customers.

FAQ

Frequently asked questions

Does recommendation engine development need customer logins?

Not necessarily. Session-based behaviour supports suggestions for anonymous visitors, though a logged-in history gives better results, since it shows what one person bought across several visits. For dealer portals the identity is already known, which makes the work simpler.

How long does a build like this take?

It depends on the state of your catalogue and order history. Where both are clean, the engagement is short. Where item codes have never been consolidated, preparation dominates the schedule.

Can the logic be changed after go-live?

Yes, and it will need to be. Weights, blocked items and category rules are exposed as settings and not buried in code, so a seasonal change does not require a redeployment.

Who owns the model and the code?

You do, in your own repository and running on your own accounts. The features and rules are documented so another engineer can retrain it without reverse-engineering anything.

Talk to us

Tell us what you need.

Share your requirements and one of our engineers will come back to you within one business day.

What happens after you share your requirements

  1. A senior engineer reads your answers. Nobody phones first.
  2. Within one business day you get a written note: what we understood, what we would start with, and the questions we still have.
  3. If it fits, a proposal follows with the order of work, what we need from you and what you receive at each step.
  4. No call is booked and nothing is paid until you have read the note. Your details stay with us.

Start with your requirements

Seven short steps that tell us what you need. A senior engineer reads them before anyone gets in touch.

Share your requirements