As the cookieless world is getting closer to reality, marketers face scenarios they can only crack with a creative mindset. There’s no one-size-fits-all solution, so businesses are having to look inwards to identify the best approach for them. That’s why first-party data is invaluable — and typically the starting point.
This is certainly true for Elgiganten, a Danish electronics retailer, who has become increasingly data-driven ever since swapping printed leaflets for digital ones showed them the benefits of digital transformation. Together with their agency Carat, part of Dentsu, they’ve continually been working on becoming more digitally mature.
“A lot of this is about partnership,” says Michael Laursen, digital marketing team lead at Elgiganten. “Carat always approach us with ideas of what we should try out next even if we don’t know the answer.”
Most recently, the collaboration led to a cloud-based machine learning model that uses first-party data to predict the value of Elgiganten’s customers.
Calculating real customer value
To get things off the ground as quickly as possible, the team started by looking at the first-party data they already had available. And they realised the website data they’d been collecting in Google Analytics 360 (GA 360) had more potential than they initially thought.
While electronics retailers are best known for selling big-ticket items, such as TVs and laptops, they actually make most of their profit from add-on products, such as insurance, cables, and other accessories. To calculate the true value of a single customer, Elgiganten and Carat needed to look at user behaviour across all product pages instead of just focusing on conversion value.
To get things off the ground as quickly as possible, the team started by looking at the first-party data they already had available.
“Most algorithms base conversions on a single buy,” explains Stefan Hansen, chief product officer, media at Dentsu. “They don’t take into account what you actually make from that product. But if someone buys a washing machine, why would you leave out the accessories they purchase to go with it? That’s what shows the real value of individual customers.”
Identifying buyers in the cloud
Elgiganten had already linked their GA360 data to BigQuery, Google’s serverless data warehouse. This meant that Carat’s data team could easily export all raw website data into the cloud-based platform to start their data analyses.
The ultimate goal was to predict whether a website user could be placed in one of two customer segments:
- Those who were purchasing (Buyers);
- Those who were just browsing (Explorers).
To achieve this, they sifted through the vast pool of data and dimensions (filters) tracked in GA360 and narrowed it down to four key criteria: ‘time spent’, ‘product group views’, ‘previous purchases’, and ‘type of products previously browsed’. An example of what they found here is that ‘Explorers’, on average, viewed more products and a wider range of categories than ‘Buyers’.
Using this insight, they then used BigQuery’s machine learning model to predict a particular outcome based on the relationship between variables. In this case, the four key criteria were used to predict whether someone would make a purchase or not.
They established that someone browsing over 100 products while spending under 40 seconds on the website only has a 3% chance of being a buyer. For someone who browses 16 products but spends 20 minutes doing so, that likelihood jumps up to 80% – even if they don’t make a purchase during that particular session.
In addition to categorising website users based on value, the model also helped Elgiganten reach new, lookalike audiences by adding the data back into Google Analytics and using it to optimise ad campaigns on Search and YouTube. This approach led to a 7X better performance in retargeting audiences year over year.
Scale with a future-proof foundation
While the retailer wanted to start small when it came to finding a bespoke, future-proof solution based on first-party data, their plans for the future are big.
“We wanted to build a smaller concept to show that you can have all the first-party data in the world, but if you don’t use it properly, it’s not going to be valuable. It’s the small version of what we are planning to do on a much bigger scale,” confirms Laursen.
The next step will be to start using Elgiganten’s membership club data. Currently counting 1.2 million online and offline customers, this will instantly boost the level of insights the model can provide, such as physical store data. Elgiganten and Carat both recognise this is where you can get the most useful information, so until they’re ready to take this step, the retailer is focusing on growing its customer club audience by expanding on their value exchange.
We wanted to build a smaller concept to show that you can have all the first-party data in the world, but if you don’t use it properly, it’s not going to be valuable.
But what they really took away from building their first cloud-based machine learning model is that it’s key not to overcomplicate things. Now that they have a blueprint, Elgiganten is well positioned to start scaling the approach and connect the wealth of data they’ve been collecting across their business. And that’s what will give them a competitive advantage in a world without cookies.
