<img height="1" width="1" style="display:none" src="https://q.quora.com/_/ad/ad9bc6b5b2de42fb9d7cd993ebb80066/pixel?tag=ViewContent&amp;noscript=1">

A new era for product discovery: Loop54 is now part of FactFinder. For the most up-to-date product and company newsvisit the FactFinder website

How it works

We've pioneered machine learning for eCommerce that overcomes data, personalisation, and merchandising problems - making Loop54 unique.


In-store vs. online


Online e-tailers struggle to replicate in-store customer experience and sales interactions.


Brick-and-mortar store

Sales assistants can direct the shopper to the exact department or item they’re looking for. They can interact, ask questions, and offer that personal touch.



E-tailers lack the ‘human-like’, one-to-one personalisation of product discovery. No virtual assistant is waiting to react to each customer need. The shopper must navigate the website alone.

Relevance is key


It’s difficult for online retailers to automatically merchandise thousands of Stock Keeping Units (SKUs) in a relevant way.


Real-time relevance

Changes to inventory or promotions can make it hard to achieve widespread relevance across all products (general relevance) in real-time when searching or browsing through an online store.


User experience

As general relevance is still unattainable, e-tailers can’t create a personalised customer experience across key user touchpoints – including site search and navigation.

To solve these problems, e-businesses must utilise the benefits of machine learning.

Machine learning for eCommerce

The key to creating the sought-after relevance and personalisation in eCommerce is to understand people’s intent and preferences in real-time. Machine learning can, but the eCommerce industry has so far lacked the vital technology for it.



There has been a lot of hype around machine learning technology in eCommerce. The rapid demand for artificial intelligence (AI) powered solutions has become the norm. Yet, these machine learning solutions rely on big data models.


As the name suggests, big data models need huge amounts of quality data – and eCommerce data is extremely sparse. As a retailer, you may have users who visit your store at a lower frequency. More so, you’ll have a large catalogue of products that frequently change, per season, campaign, or by stock-levels.


The constant changes and scarcity of eCommerce data means that big data models can’t learn any meaningful patterns, especially for long-tail search queries. The result is eCommerce machine learning models that only work with large amounts of data and fail to deliver personalisation on a smaller scale.

How can e-tailers solve this problem?

That’s where we come in. We created Loop54 as an innovative solution to these problems. Our easy-to-implement machine learning algorithm personalises user experience without relying on big data.

To better understand why Loop54 is unique, let’s look at how an example of classical machine learning – AI-based image classification.

How machines work

A machine typically takes an input, processes it, and outputs something relevant such as a classification of the input. In many modern machine learning approaches the input to train the machine needs to be labeled, meaning you have to tell the machine what you are inputting, e.g. this a picture of a dog. In the beginning, the machine is likely to get things wrong, and might understand the image is a cat, not a dog. But as the machine learns, it will adjust the parameters until it has created a model that can accurately differentiate between the two.

With increasing complexity of the input data and the model, the number of parameters that the machine has to learn explodes.

What this means:

  • The model requires input of lots of repetitive data to work
  • Results are only relevant after a certain amount of time
  • Product attributes need to remain the same to achieve the same answer

How humans work

In contrast, humans can associate and deduce information between objects quickly. It’s a natural function that helps us understand the world around us. Simply put, when we see the image of a dog, the image stimulates parts of the brain that stores information about the dog, such as size, colour, or friendliness.

As you encounter more images, you start to learn layers of complexity behind the concept of dogs. For example you learn that things like their color, size and friendliness will vary greatly, which is why you wouldn’t be surprised if you see a dog with a different color than you’ve previously seen.

What this means:

  • Only needs small amounts of data to understand product attributes
  • Gradually learns the complexity of pattern detection
  • Understands attributes even if they frequently change

How Loop54 Works

When combined, human and machine methods are a powerful solution to the personalisation problem.

Loop54 has merged machine learning with the human-like ability to learn attributes of individual eCommerce products with only a small amount of data. And it continues understanding, even if product attributes vary.

Let’s take a closer look at exactly how with the latest model of Loop54: GOLEM



The machine learning model starts in a clean slate – a blank mind



We expose the model your product catalogue, one product at a time



The Loop54 algorithm creates ‘neurons’ that are independent of your products. A neuron is where information about each new product type is stored



If a product appears which is very different from the previously seen products, Loop54 creates and stimulates a new ‘neuron’ somewhere else in the model



When the same neuron is activated several times by several products, the algorithm intuitively learns the complexity of the neurons, meaning it learns to differentiate sub-types of the neuron's product type



Products such as a ‘t-shirt’ and a ‘tank top’ may stimulate the same neuron ‘top’. As Loop54 is exposed to more of these products, it can differentiate them from each other



It learns that ‘t-shirt’ products can have varying values for attributes like ‘brand’, ‘size’ or ‘colour’, while other attributes such as the category path ‘top>t-shirt’ do not change


As the model learns to differentiate product sub-types, a second layer of neurons begins its work. An additional network of interconnected neurons creates a deeper layer of complexity. This layer starts building a hierarchal model that explains a neuron’s complexity step-wise in layers


We continue exposing the model to products until we have been exposed to all products of the feed. This creates a relatively large model with 5-8 layers of complexity, for example the neuron hierarchy ‘clothing > top > t-shirt > cotton t-shirt > organic cotton t-shirt > black organic cotton t-shirt’.

The model with its interconnected neurons and layers is unique for your eCommerce store, depending on the breadth and depth of your product catalogue

This forms the brain which powers all the products within your site and Loop54!

How is Loop54 unique?

Instead of relying on products, Loop54 creates product-independent 'neurons'.

Here is why Loop54 is the only true personalisation solution for eCommerce:


Create powerful, personalised customer experiences

Loop54’s algorithm applies to Search and Navigation to present highly relevant, personalised results for each user session.

Site Search Navigation Recommendations

“Loop54 are an important piece of the puzzle to implement our vision of 'total commerce' and serve the customer in the best way."

Large children’s e-business

“It was obvious to us product search was going to be a key component of our success. Loop54 has really delivered on our expectations and helped us achieve our goals.”

Major fashion e-tailer

“Loop54 gives us valuable data about how customers behave on our site. It’s a gauge of what our customers want – insights we can use to connect the digital and physical environment in a relevant way.”

Established online DIY chain

Try Loop54 today

Help your shoppers find exactly what they want, fast.

Let's Go