How one home improvement retailer uses machine learning to build smarter ads at speed

Sonja Horelli, Emmy Gustavson October 2019 Digital Transformation, Retail, Case Studies

The evolving digital landscape comes with its share of both challenges and opportunities. While we are more empowered to reach relevant audiences, developing and maintaining efficient campaigns is increasingly demanding.

In the retail industry one of the struggles is keeping up with the constantly changing promotions. Managing these manually is not only tedious, but an ineffective use of marketers' time.

This was the case for Bygghemma Group, the leading home improvement retailer in the Nordics and parent company of both taloon.com and netrauta.fi. The team in Finland needed to create text ads for a vast number of varying sales promotions across search, a key channel for sales generation. This required the in-house team to spend a large part of each day manually updating assets and bidding.

“We only had enough resources to create ad copy for about 10% of the most important promotions at any given time,” reveals Juha Saarinen, Head of Traffic and Analytics for Bygghemma Finland. “We needed a data-driven method to help automate the creation of relevant ads for our constantly changing sales.”

To answer the challenge, Bygghemma developed an approach that combined the Google Ads API, ad customisers, ad rotation, and Smart Bidding.

An automated solution

The team set up the Google Ads API to automate the data flow between Bygghemma’s Google Ads account and their custom in-house machine learning tool.

“This made it possible to match every ad slot with the most relevant sales promotion and landing page, and to include information about the promotion together with the appropriate link in text ads,” Saarinen explains.

Ad customisers enabled the team to adapt Bygghemma’s ads to the date, time of day, and day of the week. As a way of encouraging consumers to take action, they inserted a countdown showing the time left before a sales promotion ended. The team also used the ad rotation setting in Google Ads to specify how often the ads in each ad group would be served in relation to others.

By activating the Smart Bidding strategy target return-on-ad-spend (tROAS), bids were automatically optimised and tailored for each and every auction to produce more conversion value at the target set by the Bygghemma team.

Growing revenue and efficiency

Black Friday provided an ideal moment to test the automated solutions’ power. Bygghemma switched on thousands of landing pages with promotions at the same time. The system immediately started the countdown and created highly relevant ads, resulting in impressive year over year revenue growth. Ads containing ad customisers produced 115% more revenue than the manual approach from the previous year.

Now when there’s a new sales promotion in the store, all the relevant ads are automatically activated and the countdown is triggered. The last days of sales promotions are the biggest revenue drivers for the business, with better coverage, volumes, and conversion rates. During the last day, Bygghemma sees conversion rates increase by 80% on average.

Further discoveries

Today the ad copy coverage for active sales promotions is close to 100%, and so the team is able to use their time more productively and focus on strategic priorities.

For instance, their analysis has revealed that automated ads generate more newsletter subscribers than standard ads. They hypothesise that due to the positive user experiences, people feel compelled to sign up in order to take advantage of future sales. The team’s research shows that the value of a newsletter subscriber is much higher than a new visitor, and this value increases much faster over time.

“Now we’re more confident to continue developing the machine learning to match our products and sales database with our ads, and to automate our extensions and ad copy even further,” affirms Saarinen.

Looking ahead, Bygghemma’s goal is to develop their machine learning tool to map any keyword to business level data and customer behaviour, giving them the ability to optimise their ad customisers, promotion extensions, and sitelinks to produce even better performance.

How nostalgia proved to be a success for men’s fashion retailer Stayhard