How Salomon cut audience build time by 14x and built the foundation for AI Decisioning

Results at a glance

Some results reflect Salomon's initial AI Decisioning (AID) pilot. Performance will continue to evolve as the pilot expands to additional customer segments and markets.

The challenge

For Salomon's marketing teams, launching a campaign often depended on getting the right audience data at the right time.

An email coordinator would have a campaign ready: a targeted push for a specific program timed to reach families at the right moment. The creative was done, the brief was approved, and the campaign was ready to launch. The only thing missing was the audience — and getting it meant submitting a request to another team and waiting for the segment to be built.

Days would pass, and by the time the segment arrived, the window had closed. The campaign went out after the peak opportunity had passed, and the results suffered.

Agency partners ran into the same wall. A brief would land Friday at lunchtime with a campaign that needed to be live before the weekend. Two or three days to build an audience simply wasn't cutting it.

Meanwhile, the paid media team was running into similar issues. They wanted to push a suppression list to Meta before a paid campaign went live, but getting data out of the system and into the platform requires cross-functional dependencies — and what should take minutes took days.

Platform bottlenecks

Salomon wasn't starting from zero. Sophisticated ML models were already in place, including propensity scores, CLV segmentation, and loyalty tiers, but the infrastructure wasn't built to activate them.

As Head of Data Platform, Nicolas Bélie's job was to make the company's data work for the business. The existing DMP setup was creating friction, limiting how quickly teams could turn data into action.

The platform was a bottleneck by design. Every audience had to be extracted by the data team, and insights were slow to surface. High-value use cases like advanced segmentation (e.g. trail running shoe enthusiasts), loyalty modeling, product recommendations, and personalized journeys were theoretically possible but extremely difficult to execute.

The consequences compounded quietly: campaigns that missed their moment, budgets that weren't optimized, and customers who received generic messages instead of relevant ones. Salomon sells across a complex global ecosystem, and the gap between what the marketing teams could imagine and what the data infrastructure could deliver was growing. Every month without self-serve access to their own data was another month of missed conversions, underperforming campaigns, and a data team stretched thin.

The solution

The team made the call to decommission the DMP entirely. They replaced it with Snowflake as the single source of truth and Hightouch to connect that data directly to every system that needed it. The principle was simple: Salomon's data shouldn't be locked in a system. It should automatically and reliably flow to every platform — email, Meta, Pinterest, on-site personalization, customer support, and the point of sale.

When the right data reaches every touchpoint, every team has the ability to deliver more relevant, thoughtful experiences for customers.

The team was surprised how fast the new reality took hold. Email campaign specialists were building their own audiences within the first week — some within the first day — and Nicolas and his team no longer needed to train new hires on the tool. The product is intuitive enough that they could learn from their peers rather than going through a formal onboarding.

The impact

CRM activation

Where Salomon's email coordinator used to submit a request and wait, she now can build her own audiences — pulling from purchase history, behavioral data, demographics, and offline retail data simultaneously. The first time she ran a campaign by herself, she targeted families in a campaign with a segment she created in Hightouch's Customer Studio.

The results were significant: open rates increased 30% and click-through rates increased 85% compared to average 2024 campaigns.

Today, Salomon's 30+ self-serve users have created hundreds of audiences on the platform and can move quickly ahead of each major seasonal launch. Email marketers can build two or three new segments targeting the customers most likely to respond to their next launch and adjust the scope themselves in minutes, moving from a specific product version to a broader category.

Workflow showing how to build in audience in Customer Studio

Paid media activation

The EMEA Global Media Buying team can now push audiences directly from Snowflake to Google Products, Meta, Pinterest, Bing Ads, and TikTok themselves in minutes, without IT. Paid campaigns that used to take days to configure across channels now go live in twenty minutes. And because Conversion API signals flow directly from Snowflake to Meta and Google, the loop between spend and signal has tightened in ways that weren't previously possible.

Distributing ML models across the Salomon ecosystem

Behind the scenes, the data team finally got to do data team work.

Salomon's data and analytics team had always built sophisticated ML models in Snowflake. The problem was getting those outputs into the hands of the teams who needed them was a manual, fragile process.

Now, Hightouch propagates those models automatically across every system that needs them: CRM, media platforms, on-site personalization, customer support, and point of sale. Attributes defined in Snowflake are distributed across Salomon's services within minutes, ensuring every team works from the same up-to-date customer data.

Schema Builder in Hightouch

Across all three teams, the day-to-day shifted measurably:

That was the composable CDP at work — the data warehouse as the source of truth, Hightouch as the activation layer. With the foundation in place, the team set their sights on what AI Decisioning could add on top.

The next chapter with AI Decisioning

Most brands are still figuring out segmentation, but Salomon is already asking the next question. With its data foundation in place, the team piloted Hightouch AI Decisioning across approximately 350,000 English-speaking customers, starting with a "first to second purchase" use case. The approach was a departure from how lifecycle marketing had worked before.

Rather than building a journey, writing message variations, and waiting weeks to see what worked, an AI agent selects the content, product recommendation, subject line, and send time for each individual — running experiments continuously across every customer interaction, learning with each one, and getting sharper over time. Marketers define the campaign structure and goals; the AI handles execution and optimization.

The team can see which messages were selected, why they were chosen, and how they performed. The model doesn't start from scratch either — Salomon's existing ML models (the propensity scores, CLV segmentation, and loyalty tier calculations built and maintained in Snowflake) feed directly into the agent's decisions at runtime. The AI is drawing on everything Salomon already knows about each customer, and the early results are promising.

Hightouch AI Decisioning (AID) interface where marketers create agents and personalize customer experiences

With Snowflake and Hightouch in place, Salomon moved from a platform that gatekept data to one that activates it — and from segmentation as a bottleneck to AI Decisioning as the next frontier. When the right data reaches every touchpoint, marketing stops chasing the moment and starts meeting it.