Frequently asked questions

A text file with two columns separated by a semicolon: customer identifier and product identifier, one row per purchase. Names, addresses, amounts and dates are not needed.

There is no fixed threshold. Suggestions come from comparing customers, so the more customers with purchases in common you have, the more stable the scores. With few customers, or anonymous purchases, there is little to compare.

Yes, if you can tie purchases to a customer: loyalty card, customer record on the order, account. Without that link the history does not say who bought what.

No. A customer code and a product code are enough: no names, addresses, emails, amounts or dates. The customer code is the one you already have in your ERP.

Your system makes two kinds of HTTP request to RecommenderPro, authenticated with a token: one for a customer and one for a cart. The integration guide has the format, the requests and a sample response. The web app does the same by hand.

Yes. You use them wherever you like: product page, cart, checkout, email after the order, agents’ app. You set the minimum threshold, the number of suggestions and separate archives by season, department or type of customer. Any ERP or online shop that can export purchases and make an HTTP request can be connected, and we look after the integration.

In the cloud, or on your own servers. On your servers you need a 64-bit Windows machine with PostgreSQL.

When you load a new history. Loading is complete: it replaces the previous archive in a single operation. There is no row-by-row update.

Yes. In cart mode the products the customer has in the cart at that moment are enough, even just one, and you do not need to know who they are. If you do know them, history mode starts from everything they bought in the past.

Yes. The web app generates a file listing, for each customer, the products to offer them (up to five), ready to import into your email marketing tool. Sending the emails stays in your tool: RecommenderPro does not send them.

They follow the purchases you load: the score is recalculated on the archive every time. If you load the updated history, or one archive per season, the pairings change with your purchases. The engine has no seasonal model and does not update itself: reloading the data is up to you.

Italian, English, German, Spanish and French.

Yes. We offer training on integrating with the software you already use, and support packages. We agree the details after the demo.

We do not quote a percentage, because it depends on your history, your catalogue and where you show the suggestions. In the demo you see what the engine suggests and judge whether it makes sense for your customers.

It is the percentage of customers with purchases in common who also bought that product. A score of 72 means 72 customers out of 100. It is not the probability that the customer will buy.

It depends on the amount of data loaded and the number of requests. We set it out in a quote after the demo.

With the demo: we show you the web app with sample data and how it connects to your system. Request it here.