
How Wolt runs global creative strategy: 10 webinar lessons
Wolt's Rumina Todorova on global creative strategy: local room in global concepts, quarterly cycles, spend-free benchmarks, AI tagging and personas.

"Most of the concepts that we make are global, but we always leave a little bit of space for local context and current trends as well."
Focal Point is our webinar series where we talk to creative strategists and other professionals working on performance creative at large scale. In the second session, Rumina Todorova, Global Brand & Creative Strategy Manager at Wolt, joined Focal CEO Andreas Stenman to talk about the global creative strategy behind Wolt's courier marketing:
Building a creative system on research and personas
Balancing global concepts with local context
Production cadence, and why Wolt plans fewer concepts and more iterations
Analyzing creative performance independent of spend, with AI tagging
Where AI helps in creative work, and where it doesn't yet
Working with UGC creators at scale
This is our overview of the session. You can watch the full session recording below as well.
The setup: one small team, one business line, many markets
Wolt is one of three brands under DoorDash, alongside DoorDash and Deliveroo. Rumina is based in Copenhagen and handles everything brand and creative for Wolt's courier business line. Her small team supports the courier marketing team with acquisition, which means delivering creative for every placement they need, from performance marketing to CRM, plus brand campaigns.
All courier creative production now sits under one umbrella. Content splits into two main workstreams:
BAU (business as usual): more basic, longer-lasting content that doesn't need refreshing as often. Some teams call it evergreen or always-on.
UGC: user-generated content, refreshed more often.
The team doesn't split people by workstream. Everyone works on everything, which only works because every production starts from the same global narratives.
10 global creative strategy lessons from the session
1. Build the system on research and personas
Everything in Wolt's creative system starts from brand foundations the team built over the past year. They ran in-depth market research to understand where they stand with courier partners, the competition and the general public, and above all what problem they're solving for people who want to become couriers.
That research produced a communication strategy with story pillars. From there, the team takes one of two routes:
Apply the story pillars to a global creative production, then localize it for each market.
Apply the story pillars to a specific persona, build a global concept for that persona, then localize it.
Either way, the cycle ends with launch, learn and optimize. (Watch: Wolt's creative system, 04:00)
The personas themselves are built almost entirely on interests, motivations and life goals, with demographics attached afterwards. Rumina's example: knowing that Johnny needs money for a new car tells you far more about what to show him than knowing he's 24. To build them, the team ran qualitative interviews with active couriers, potential couriers and people with no interest in the job, and kept going until the answers started repeating. That repetition is the signal you've found a real audience. (Watch: building personas from motivations, 46:01)
2. Make concepts global and leave room for local
Two years ago, Wolt's performance marketing needs were growing faster than anyone had bandwidth for. Rumina started with one freelancer making basic creatives, then added UGC agencies. The first UGC partner couldn't deliver hyper-local productions that performed, so the team switched to regional agencies, with in-house BAU assets also made per region.
That setup meant doing the same work three or four times, losing track of concepts and losing track of assets. Bringing in Focal gave the volume some structure. Once it was organized, the repetition was obvious, and the team started consolidating under global narratives and localizing from there.
The structure they landed on this year scales: "You can add 10 more countries or 50 more countries, and this thing will still make sense." Most concepts are global, with deliberate space left for local context and current trends. Rumina called that local space tricky to manage, but necessary at this scale. (Watch: from one freelancer to global concepts, 05:48)
3. Set your cadence by how fast creative fatigues
Wolt produces quarterly. That gives the team enough content for regular creative refreshes without running out, and enough time to get meaningful insights from what's running before the next refresh. The cadence is based on publisher advice, in-house testing and ongoing retesting to keep ads out of fatigue and CPA in check.
Rumina was clear that there's no universal answer. Cadence can differ between business lines. How often you refresh depends on how quickly your content fatigues. For Wolt, a longer cycle means they can produce more in bulk, and they keep cohesion across platforms and markets: someone traveling from Greece to Denmark should see the same pattern of ads. (Watch: why Wolt produces quarterly, 10:18)
4. Plan fewer concepts and more iterations
Next year, Wolt will make fewer net-new concepts and more variations of the ones that work. The production pipeline will leave room for more variations of each concept, with the goal of finding more winners and reducing CPA.
Her reasoning comes from experience. After long enough in an industry, there's rarely a concept you haven't heard of or thought of yourself, especially in a niche like courier recruitment where the possibilities are limited. Making the most of the good ones beats constantly reinventing. (Watch: fewer concepts, more iterations, 13:27)
5. Organize production around the global concept
Wolt runs 17 countries in the example Rumina gave, each with up to two languages, and each language with maybe two variations. That's a lot of content to track, and the board structure took several iterations to get right.
By status (briefing, in production, approved): the standard creative setup. An "in progress" view mixing every country and language was impossible to make sense of.
By country: easier to read, but Rumina had to approve the same global concept three times when it ran in three countries.
Global first, then broken down per country: one approval for the global concept, then the local versions underneath.
Different people also get different views. Whoever receives the content uses a status view to grab what's approved. The production team organizes the same work their own way. (Watch: organizing production across countries and languages, 14:29)
6. Benchmark creative independent of spend
Analyzing this volume of assets took the team a while to figure out. The first step was benchmarks, set in a creative context rather than a media one. CPA is heavily influenced by spend: too little spend can inflate it, and changes in spend move it in other ways too. Rumina wanted to see how creatives perform side by side, in isolation from spend.
With benchmarks in place, the process is:
Separate good content from bad against the benchmarks.
Work out what in each piece makes it good or bad.
Turn the patterns into best practices for the next production.
Rumina got the idea from a partner who presented a detailed, courier-specific creative analysis. The specificity is what convinced her: "I don't want to look at generic things. I want to look at my individual case scenario and make sense of it." (Watch: creative benchmarks, independent of spend, 17:05)
7. Tag deeper than your naming convention, but start with three
Wolt's naming convention captures the value proposition, but in a courier context that only goes as far as "earnings" or "flexibility". To get one level deeper, the team set up more than 25 attributes, applied to every asset with AI tagging in Focal. Examples include the mood of the person in the image, whether a face is visible and whether they're facing the camera or looking away.
Tagging thousands of assets with 25 tags each by hand would have taken a couple of years, by Rumina's estimate. With AI tagging it took a couple of days, including writing the prompts for each attribute. The team then uses AI to analyze the tagged data and look for patterns across countries and creatives, rather than comparing single tags in isolation.
Wolt went all in from the start, but Rumina doesn't recommend that for everyone. Start with three, or with your naming convention, and build from there. Her suggested first tags:
Brand elements: how much branding each asset needs. For couriers that's a jacket, bag, T-shirt, helmet or branded vehicle. The first round already showed the minimum needed and what's too much.
Gender: this produced an unexpected finding for Wolt.
Copy vs. visual: separate the two, since some visual elements may matter less to performance than the copy, or the other way around.
Those findings now feed hypotheses for an A/B testing framework, with an always-on A/B test planned to keep testing small elements. (Watch: AI tagging with 25+ attributes, 19:06)
8. Expect winners to differ by market, publisher and campaign type
One concept can do well in one market and poorly in another. The same concept can also perform differently on two publishers, or across two campaign types on the same publisher, such as web vs. app campaigns. Rumina sees more layers still untapped, like iOS vs. Android users.
Wolt handles this with enough variations in each quarterly production to rotate, and with personas that stay the same across markets so analysis can be cut by persona as well as country. How far you dig is a trade-off: it depends on your time and your goals.
One audience question asked for surprising patterns. In one Central European market, gray, gloomy creatives outperformed brighter ones, which nobody expected. Rumina's broader point: everyone has subjective bias, herself included, so when opinions clash the answer is to test and see. (Watch: why winners differ by market, publisher and platform, 25:00)
9. Use AI for the time-consuming work and keep a human on the creative
Rumina calls herself an early adopter, but she's selective: "The fact that we have it doesn't mean that we have to use it everywhere across every process." Where AI earns its place for Wolt:
Tagging and analysis: AI tagging on every asset, plus a dashboard that brings together tags, CPA trends, spend, benchmarks per campaign and country, and A/B test suggestions.
Tracking production: checking which creatives went live, how long they stayed live and what was left unused. After the performance team kept asking for more content, this showed where the team was producing too little or too much, which now feeds next year's planning.
Day-to-day work: Rumina runs AI tools on tasks that would otherwise eat her time.
Brainstorming from research: going back into existing research to find insights the team missed the first time.
Net-new concepts still start with people. The team brainstorms face to face and keeps a long backlog of ideas. (Watch: how Wolt uses AI day to day, 28:14)
10. Give UGC creators concepts and guardrails, then trust the agency to pre-review
Wolt now runs all UGC through one global agency with reach across many countries. Each creator gets the global concept and the creative freedom to make it their own, plus guidelines on what to do and what not to do, and any hard requirements, like a hook in the first three seconds. The team tried writing full scripts for creators to act out, and it didn't work as well.
Review works because of the agency relationship. The agency handles communication with creators and only sends near-finished drafts, so in almost every case Wolt just approves or asks for minor fixes. That took time: the team invested in teaching the agency the business, brand fundamentals and compliance. The agency knows, for example, that a wrinkled courier bag is an instant no.
Each concept has so far been made locally by a native-language creator. Wolt is now testing some concepts that only need a voiceover or subtitles, so one shoot can be reused across markets, without losing a local look and sound. (Watch: briefing UGC creators, 41:49)
The hardest part of running creative at scale
The last audience question asked what's most difficult about operating at this scale. Rumina named two things. The first is understanding and structuring all the processes and data so you can see the big picture and act on it. The second is trade-offs. AI translation is one example: some teams use it, but Rumina hasn't yet found a process that fits the approvals and compliance her translations go through.
The biggest trade-off is constant: more local concepts keep the local context, more global concepts scale better. "Maybe you magically find a way to balance them, which we are trying to do, but there is always a little bit of both worlds."
Looking ahead, Wolt's two priorities are learning more from what's already live to find more winning concepts, and, longer term, ads tailored to individual needs, built on the personas, tagging and data the team is putting in place now.
Watch the full session: How Wolt scales creative across markets on YouTube. Catch up on the first episode with Granola in our session overview, and find upcoming sessions at focal.inc/webinars.

