Predictive marketing
Predicts and recommends inside your systems.
We have spent years building customer data software. It predicts who will buy and what, recommends one-to-one and writes the result into the CRM and the store you already have.
Not a product you have to learn to operate. It is what keeps you from starting from scratch.
Fifteen years building with data and more than two hundred projects delivered, for companies that cannot afford for it to fail.
AWS technology partner
What it is
From your data to your channels, in four steps.
One profile per customer, built from what you already store. On top of it, the models. And back into the systems where you already work.
- Sources ERP, CRM, ecommerce and documents, as they are today.
- Single profile One customer, one record, even if the data lives in four systems.
- Models Predictive and generative, on your data.
- Activation Back into the CRM, the store and the sending channel.
Integrates with
PrestaShop and Shopify integrate through a native module, installable from the store itself. The rest through REST API and JavaScript SDK, with response times under 100 ms.
Predictive models
What each customer will do.
See these casesThey are trained on your history, recalculate on their own and write the result into each customer’s record. Three are standard; the fourth is trained for the question you have.
- How much they will spend in total Customer Lifetime Value (CLV): what they will bring in while they remain a customer, if nothing changes. It shows who is worth retaining and where not to invest.
- How often they buy again The consumption cycle: in which week they will need it again. This is where sends that arrive on time come from.
- What they will buy next time Purchase propensity, the buyer-persona: which categories are likely to be in the next order and which are not.
- Whatever you want to predict Custom models, trained for a question specific to your business. They start from the same history as the other three.
Segmentation and promotions
What to offer each customer, and when.
See these casesEvery customer sits in a segment according to their purchase cycle and what they buy. Each segment calls for a different action, and no promotion goes to someone who was going to buy anyway.
- Buys when they are due On cycle. No discount needed: recommendation and content, and the margin stays whole.
- Running late Their expected date has passed. A reminder or a small incentive, before they go cold.
- Has not bought for a while Dormant. A win-back promotion in proportion to what they were worth, not the same one for everyone.
- One of the highest contributors High value. Preferential treatment and early access, not a discount: the repeat purchase is already there.
That way each customer gets the promotion that moves them and none that is not needed: more repeat purchases, less dormancy and more value per customer.
Recommendation networks
What to show each person.
See the store casesNeural networks trained on real purchasing behaviour. They decide which products each person sees in the store, in the app, in the newsletter and in automated sends.
- One-to-one That person’s products, not the average’s.
- Product to product What gets bought together with what they are looking at.
- One-to-one ranking The same listing, ordered differently for each visitor.
- Bestsellers The default for first-time visitors.
In production
Where it is running.
See all projectsThe four layers running together on each company’s systems. Each project has an article on what was implemented and the results: that is where the figures this page does not publish can be found.
Does it fit your business?
Tell us which systems you have and what you want to achieve. We will tell you whether the platform shortens the path, and if it is not worth it, we will tell you that too.
Frequently asked questions
What people ask before deciding.
Do we have to migrate our data?
No. The platform reads from your systems and writes back into them. Nothing moves and nobody changes tools.
Are the models yours or a third party’s?
The predictive models and the recommendation networks are ours, trained on your data. For generative tasks we use the models available on the market.
Can we use only part of it?
Yes. Some clients only enrich the CRM and others only personalise the store. The four layers are not sold as a bundle.
What if our store is not on the list?
It integrates through API and SDK, like any other development. The native module saves the integration work, but it is not the only way.
Can we buy the platform without a project?
We do not offer it as self-service: there is always someone on our team accountable for the result. What you do not need is to hire a specific service to be able to use it.