Platform

Retailers need a holistic understanding of their consumers.

Today, retailers and e-commerce brands have millions of customers with often no real insight into who those customers really are and why they bought what they did. In order to truly market to customers, retailers and e-commerce brands need to understand what a customer is doing before, during and after a purchase, to be able to provide any fact-based attempt at marketing to them. If a retailer isn't capturing that data, the outreach is based on rule-automated and human-curated gut-instincts, triggered alerts based on rules, and other sub-optimal marketing attempts.

With the Mercanto Canvas platform marketers benefit from the power of artificial intelligence as it analyses consumer behaviour, aggregates thousands of unique predictive attributes on millions of customers in real-time, and maps those attributes to structured actionable data, building a comprehensive, holistic view of each customer.

Mercanto's personalisation platform learns and adapts to each customer's unique preferences via semantic attributes, taking the consumer experience to the next level and engaging every individual with unique and relevant experiences at scale.

  • DASHBOARD
  • DRAG & DROP UI
  • FILTERS
  • COPY HTML
  • PASTE INTO ESP

Mercanto's Learning-to-Rank Technology

Mercanto's “learning-to-rank” technology learns customer preferences from each customer’s interaction and dynamically adapts to each customer in real time. Mercanto tracks each piece of content it displays and ensures that no two customers see the same content, and no customer ever sees the same content twice.

Semantic understanding of content
Focus on the individual attributes of each piece of content, and each customer’s reaction to them.
Real-time scoring
Mercanto’s ranking engine automatically scores content for relevance based on every customer impression.
Adapt to each user's response

We learn from each customer’s reaction to displayed content, and adapt the following experiences accordingly; we amplify the content a customer expresses interest in, and mute the content they do not.

How it works

To help understand the concept, imagine a three-legged stool. The three legs of this stool are: (1) an auto-tagging or tokenisation system that understands everything about each of the retailer’s products, (2) a Spotify-style ‘taste profile’ for each consumer, and (3) the algorithms that take all this data and automatically rank products for each consumer.

1. Products

The first leg of the stool – product tagging – is based on a particular AI technology called Natural Language Processing (NLP). NLP automatically extracts each product's attributes - even for retailers with unstructured product data.

These could include single or multi-word attributes, such as ‘party shirt’, ‘slim fit jeans’, ‘Tommy Hilfiger’, ‘work to evening’ – there can be millions of attributes extracted across a retailer’s entire product catalogue. The goal is to look beyond the standard product data to understand each product’s most important features and attributes.

6
7-1

2. Consumers

As consumers interact with products, these interactions are married with each product’s attributes to construct a ‘taste profile’ for each consumer. This is the second leg of the three-legged stool. The taste profile is a collection of everything a consumer interacts with – akin to a word cloud of each person's tastes and interests.

Taste profiles are the foundation of personalisation: they aid retailers in recommending new products as they’re added to the catalogue. And if a consumer happens to like the most quirky and unusual product in your catalogue, something that only 20 people in your database would dig, you can find those 20 people and connect the dots between the product and the consumers.

At the end of the day, it’s about understanding each consumer’s personal tastes, and making recommendations accordingly.

3. Algorithms bring it all together

The connection between data from the retailer’s product catalogue and each consumer’s unique taste profile is the third and final leg of the stool:

  • One one side, we know all of the product attributes from the Natural Language Processing. 
  • On the other side, we have a taste profile for each consumer.

We continuously take these two things, do a little magic filtering, and rank products to match products to each consumer’s taste profile, while also incorporating new and popular products, together with an element of diversity.

11-1

Integrate Mercanto with your existing email provider

Connecting to your product feed and web analytics platform, Mercanto ranks catalogues, product collections, marketing promotions, and individual products. It then evaluates all your content types, allowing them to be blended into a single communication. For example, it could create a personalised email that includes the right items, categories, and suggested products for that individual consumer but also includes those that match a specific promotion.Using the platform, a retail marketer coordinates and combines different content types in relation to each other. The output is a code snippet that can be used in email templates in any leading ESP.

Integrate with ease

Mercanto completes comprehensive integrations in just a few weeks. Since we integrate directly with your existing web analytics platform and product feed, you’ll soon be up and running with tangible results.

Harness untapped data

Why should marketers care about a system that uses Deep Neural Networks to process consumer interests and product data in milliseconds? Because it (finally!) enables them to serve the right message to the right person at the right time. Marketers gain complete control, eliminate organizational bottlenecks, and drive more impact.

Launch at scale

Individualised content is key to unlocking ROI across digital channels, yet most marketers fall short due to organisational barriers, data silos and technological shortcomings. Mercanto’s artificial intelligence platform empowers marketers to launch 100% individualised campaigns in less than 10 minutes.