Product Intelligence – A product-centric AI opportunity

A paradigm shift is on the way, and we are on the AI cutting edge of that change.

We have developed state-of-the-art prediction software for the shopping industry.

this is

Something new

With a focus on complete privacy protection, we have developed state-of-the-art prediction software for the shopping industry. This new, intelligent system is built upon a set of AI models, utilizing millions of products, billions of interaction points, and trillions of data combinations. We call it Product Intelligence

Hello Retail’s software translates product information into vectors of bits and bytes, a series of numbers, discovering dependencies and similarities. This information trains our AI models in product understanding and awareness, providing core value to any professional shop owner, it being traditional commerce with physical stores or full-blown ecommerce businesses. 

Product Intelligence has become the foundation of our product offering and lays the groundwork for business development in the future of commerce. Our cutting-edge AI improves customer and shopping experiences, provides powerful industrial insights, and assists in building product strategies for marketing, customer acquisition and retention. A holistic building block for your business strategy.

Learn more

Hello Retail specializes in eCommerce personalization and optimization. Our product offering aims to facilitate the best online shopping experience and provide a solid foundation for business decisions. A personalization engine has always been at the core of our products. It has, until today, been developed in a traditional and orthodox way, but we are about to turn that world upside down! Or perhaps – inside out.Our journey began over ten years ago. For a long time, we followed an established approach to building personalization software, which included user tracking and machine learning based on the specific behaviors of individual users.Using the traditional method, we needed cookie tracking to track the individual through a long tail of product visits, monitoring interests and other purchase patterns. For example, if I were browsing for white sneakers in a store, we would view me as a visitor viewing different products, write that information to our system, and look at what I might be looking at next as part of the tail in the product visitor journey.
Skaermbillede 2023 01 06 kl. 13.11.04

As a result, we had to stick with the visitor in the shop and place a hook on them – a glued and exclusive tracking cookie. This cookie would be used to serve them with products as they continued on their journey. In a way, it is similar to the experience of being followed across platforms by the same ad for one specific product just because you glanced at it once – in my case, the same white sneakers everywhere I went!

And then, in early 2016, the European Council agreed on data protection regulations. GDPR did not try to put down obstacles for tracking companies, but was agreed upon to safeguard privacy on the internet, and this resonated with us.

We always felt we had to take a stand and achieve our goal of providing 100% privacy-compliant eCommerce optimization software, by doing things differently.

We knew that eCommerce optimization software had to be built in a new, innovative way without compromising the personal touch and feel of visiting a well-known store.

Change the world upside down

Building optimization software in new ways

However, providing a software solution, we had to think differently and It is not as simple as it may seem to make a machine understand what a human mind would easily conclude.

Consider the situation where you walk into a fashion shop in the mall, and the salesperson examines you as you enter. She first notices your general approach – do you make eye contact? Are you here for a specific purpose, or are you just browsing? Which brand of clothes are you wearing? Is it expensive or not? These, together with various other subtle or more exact details, make up her initial impression of you.

She’ll use all the information she has to size you up for how to provide you with the best possible customer experience – and of course, to get you to buy from her.

She does not only know what to suggest or respond to but also how her merchandise fits well together. She knows what is popular, what she thinks is trending with her customers, what is often the best product to sell, and which products to introduce later in the sales process. It is all these decisions that make her an expert in her field.

The human mind is a fantastic wonder, and it combines and concludes in milliseconds based on input through our senses. 

That has to be a perfect analogy of how we strive to develop the best possible solutions for your business. We build our software in a manner that assists and finds the best products to present to each customer in any context possible. 

We want to be the solid data foundation for building the right strategy for your commerce solution, being an eCommerce or physical store. We develop algorithms and AI models to mimic the same expertise as the shop owner behind the counter in your local shop but with the experience of a thousand shop managers.

We succeeded by building what we call Product Intelligence – true Artificial Intelligence.

As described previously, in the past, all information related to products’ views and purchases would stick to a user’s profile. So when I looked at the white sneakers, that product view was attached to my user profile in our system. With Product Intelligence, we now turn things upside down and attribute everything to the product instead of the user.

Just as the shop seller from the fashion store described above, noticed every little detail as you walk into the shop, so are we utilizing an increasing amount of anonymous data beacons, tracking information, and parameters. All to train an AI model to learn from and understand the behavior, and instead of attaching the knowledge to the user, we attach all the data points to the product itself. This form a collective knowledge between and in the intersection of product relations – and is used for predictive algorithms and insights. 

Product Intelligence is the new AI backbone of our offering. It is the engine that sets the new foundation of all our solutions- existing or future. We use Product Intelligence to determine the products recommended to visitors, personalize their search experiences, and provide unique intelligence insights valuable for any eCommerce store.

So, what is this?

The foundation of Product Intelligence consists of our most important building block – the products in our customer’s inventories. We have millions of products in our system, all assigned in closed compartments for each shop, but when we look at them all from a global perspective, we start to learn. We use this knowledge to automatically improve the AI, and become wiser. This doesn’t mean that any customer-specific product information is reused with other customers – rather, it means that, as you, as an example, become a more experienced human being and train your brain to understand your surroundings – you gain the intelligence needed to make decisions.

In order to build Product Intelligence, we must first understand our building blocks( the products), in new ways – not just a product image and a title and a price. It’s important to build a machine that sees them as a human would, but in a way, a machine can understand – bits and bytes.

Our success has been to develop an exact translation method for actual products into vectors. A vector is down-to-earth math and is really a series of numbers, which can be expressed in a map, holding magnitude and direction, and what is explicitly important is that it is in the same map as other vectors.
In other words, other similar products, closely related products, or groups of products in more dimensions will be easier to identify. Vector representations in this context can be looked at as you would understand DNA in the real world – although you cannot reproduce a product from a vector (as you can with living objects from DNA in advanced genetics), you can still recognize similarities. 

Let’s take an example.  – If you take two DNA strings and compare them, you might see that they both come from a four-legged animal so it tells you something about what this DNA string originates from – we can, in the same way, see that two vectors in our system are the same type of product, for instance, a laptop.
Furthermore, if we zoom in on the details, we might see that the DNA strings we are examining are not only from a four-legged animal, it is in fact, from an 8-year-old white cat. We know that because we can compare DNA strings –  once again, in the same way, we can see closer similarities in our vector system, so that the two vectors mentioned before represent a specific space grey MacBook Pro laptop.

So we have the possibility to see both groups of products with general parameters and attributes, and we have the option to zoom in and compare for a more fine-grained level of details. This is especially important when you want to identify similar products to ensure reusable knowledge. 

This is really one of our greatest achievements because the data we have on one specific product is very often not complete in terms of actual product information, maybe we are missing a specific product number, maybe we are lacking a good description, maybe the image is not good enough – but as a human, you would put the pieces together and identify the product, our system can do the same. Advanced genetics are able to reconstruct full DNA strings from small corrupted DNA sequences, we can equally do the same for products and vectors – if we only have very little product information, we are able to translate this into a corresponding vector and place it very accurately on our map, and hereby construct a fulfilling product understanding

When this is applied to millions of products,
we can develop true Product intelligence.

The brain - Product Intelligence

Let’s dig a level deeper to understand Product intelligence further

Product Intelligence is an AI engine that simulates how human brains work with specialization in the e-commerce and personalization industry. 

At a somewhat crude level, one can think of a human brain as a complex network consisting of an enormous amount of nodes – the neurons. Moreover, the neurons are connected to each other by nerves so that they can communicate information and accomplish complex tasks through collaboration. It turns out that we can use almost the same language to describe our Product Intelligence Engine. The details are, however, quite involved, and we shall break down the details into digestible pieces in what follows.

Naively speaking, the neurons in our Product Intelligence are simply products from all of our customers, i.e., vectors -, and the nerves carry all the information about how different products, either from one shop or from an entire industry, are related to each other. One fundamental issue with this naïve approach is the extremely high fluidity of the industry, where products get updated almost constantly. It is, therefore, meaningless to keep track of each individual product. Instead, our engine aims at learning patterns using the following strategy. 

The actual neuron can be thought of as a tiny cloud or cluster of product vectors that are close variants and are the most up-to-date. 

For example, a neuron in our context could be a collection of MacBook Pros with white, black, or gold color. Here there are two important points to keep in mind:

  • Vectors in a single neuron don’t necessarily come from one shop. For example, the exact same MacBook from two different shops will technically be different products but they belong to the same neuron.


  • The size of a neuron is not a priori fixed. Indeed, it should scale according to the task one wishes to accomplish. For example, if the task is to analyze the trend of Apple products within the electronics market, then one should scale up the neuron to combine all MacBook Pros, regardless of the precise versions or specs, to get the big picture. However, if the task is to recommend accessories to a customer who already has a specific MacBook in the cart, then the size of the neuron should be scaled down in order to find, for example, the most compatible cables and external hardware to the specific model.

Let’s bring the neurons online

Next let’s talk about the nerves in the engine which bring the neurons to life.

Just like a human brain, the nerves in this context come in different species, which are designed to deal with different tasks.

For example, there is a type of nerve in our scenario, which may be called the “accessories nerve“, and it is responsible for recognizing products that are typically purchased together. 

Imagine a customer looking at a coffee machine. Then a Product Intelligence neuron called the “coffee machine“ (i.e., a group of very similar coffee machines, say from the same manufacturer within the same model) will be fired up and send out signals through the accessories nerve to activate a number of other neurons which are probably called “coffee beans“, “coffee filters“, and so on. 

Now suppose our customer added a coffee machine to the cart and moved on to a certain brand of coffee beans. In this scenario, both the coffee-machine-neuron and the coffee-beans-neuron are fired up, and collectively they send out new and potentially stronger signals toward coffee filters and mugs, and so on.

Other types of nerves could potentially be “alternative products”, “add-ons at checkout“, etc.

So the nerves in our human brain analogy are really algorithms using the product vectors to accomplish the needed tasks. 

Another important point to highlight here is that although each set of algorithms in the engine has its own functionality, they can be combined to accomplish more complex tasks. For example, if a shop aims at improving the click-through rate or conversion rate or whatever metric of interest, the engine can simultaneously spin up several algorithms and find out the best combination for the task through continuous improvements. 

Out of our endless list of capabilities, here are a few sample use cases of how Product Intelligence can be utilized.

  • Providing insights into which type of products work well within industries and customer segments. For instance – Do we across your industry see a higher conversion rate in products above 100$ or below? What is the purchase frequency for specific or groups of products? Which types of products sell the best in the fall vs. the spring? What would be the best product to promote in a newsletter, frontpage recommendation, or landing page?
  • Guiding business owners on which products they should ensure in the stock for better upsell opportunities. “We can see other relations in our system are seeing a better performance by selling shoelaces – You should add that to your stock !”
  • Lifetime value analysis to Identify the best possible product to sell as the next in line to nurture the highest possible lifetime value for newly acquired customers – Studies show that lifetime value varies a great deal based on what product the customer buys as the first one when seeing the shop for the first time – this is especially important in marketing scenarios – find the products to be promoted and be bought first to generate loyal customers with best possible lifetime value
  • Prediction tools – like customer behavior prediction around products – brand prediction for industries – pricing obstacles for campaigning, and stock handling 
  • Best in the business engines for Recommendations, Search, and Merchandising – available through APIs or via our Solution offering – all built on our Product Intelligence.

Last but not least, 

Our system evolves by learning from itself and is directed by continuous tests to become better and better. Just as we described previously – the more we learn as human beings, the better we get at making decisions in other scenarios.

As for human beings, knowledge is accumulated and skills are constantly practiced. At the heart of our product intelligence are our constantly running tests against real-world performance measures and the fitness of our customers’ needs. 

On the one hand, both the neurons and nerves are frequently updated according to the continuous flow of data from our customers, so it stays on top of the market trends. On the other hand, the internal strategies deployed by the Product Intelligence Engine to achieve various business goals by activating and orchestrating combinations of nerves are always optimized by monitoring the results of the tests.

All of this is, of course, being implemented in our product offerings like Product Recommendations, Search & Pages, but it will also be available as a standalone entry point via a well-described business API.

This is what we call Product Intelligence – This is what we believe will change the industry entirely in the future to give it holistic depth. We have it today in the core of our system to be used by both our products offered to the market, but also as an important building block in 3rd party systems and especially for business development in the global commerce market as a whole – If you want to learn more about our way – do reach out as soon as possible. 

We did this!!!

Thank you!