00:00:00 – Introductions

Daria: Welcome everyone! Thank you so much for joining us today for the bonprix, Google, and Sellforte session on how bonprix built a modern marketing measurement system. My name is Daria, I'm from the Sellforte marketing team, and I'll be hosting today the chat and Q&A. We have around 40 minutes of presentation and also, in the end, 20 minutes of live Q&A. Friederike, over to you.

Friederike: Perfect. So most of us are currently living in a measurement nightmare. We have more dashboards than ever, but we still can't really answer that one question CEOs actually care about: did this spend actually drive the sale, or would the customer have bought from us anyway? Today we're going to show you how bonprix stopped guessing and started knowing.

00:01:48 – The Measurement Nightmare

Friederike: Take a look at these numbers. If you feel like you're constantly guessing which half of your marketing budget is actually wasted, you're in very good company. 76% of advertisers are essentially still in the toddler phase of measurement maturity. Only 9% have actually figured out how to make their data talk back to them in a way that makes sense. Most brands are stuck in this "emerging" bucket, which is really just a polite way of saying we have a lot of dashboards but just very few answers.

00:03:08 – The Webinar Roadmap

Friederike: I am Friederike from Google, and I spend my days helping partners navigate exactly that mess. I'm super happy to be joined today by Anika from bonprix and Juha from Sellforte. Before we open the hood, here's a roadmap: I will walk you through the strategy—the playbook for building a measurement stack that survives a post-cookie world. Then Anika will follow with the execution and the bonprix blueprint across 16 markets. Juh will then focus on the tech engine to show how MMM actually runs in the real world, integrating incrementality and attribution.

00:04:06 – The Measurement Triangle

Friederike: Can you give me a thumbs up if you have seen this triangle before? I see a few thumbs up coming in. Brilliant. Now, give me another thumbs up if you use all three: attribution, MMM, and incrementality. Anyone? Anika, thank you! And my final challenge: give me a thumbs up if you're using all three and these actually talk to each other. Any thumbs up still? It's not that easy, is it? That gap between those three pillars is exactly where your profit is leaking.

00:05:31 – The Football Analogy

Friederike: To future-proof your growth, you need these three on the same team. Attribution is your daily pulse for daily bidding. MMM is the strategic compass; it captures the untrackable stuff like physical catalogs or the weather. Incrementality is the truth serum; geo-lift experiments prove what happens to revenue when you turn a channel off. Think of a football analogy: a goal is scored after six players touch the ball. Attribution shares credit across the team—seeing the winger's assist and the midfielder's pass—but it sees the touch, not the cause. Incrementality asks: if we took that winger off the field, would the goal have happened anyway? MMM is the season review, looking at the whole year to allocate budget to win the league next year. The magic happens when they calibrate each other.

00:08:37 – Culture and Execution

Friederike: This isn't just theory. The leading 9% are delivering 3.4 times more growth. A study by BCG found that leaders share traits that boil down to culture and execution: elevate measurement to the C-level, link learning agendas to business objectives, run tests to isolate value, and integrate insights to calibrate solutions. I think this is quite a lot of theory, so Anika will talk to you a bit more about how they manage this in the real world. Anika, over to you.

00:09:55 – The bonprix Blueprint

Anika: Hi everyone, my name is Anika. For the past months, I've tried to help reshape how we want to measure marketing impact at bonprix, and how we want to turn data into decisions. When I reflect on this journey, three things mattered: first, having a clear strategic foundation; second, setting up the right organization with leadership buy-in; and third, driving real impact by turning measurement into business decisions.

00:11:00 – Why bonprix Implemented MMM

Anika: Building a strong foundation is one of my biggest learnings. We started last summer. For years, we heavily optimized toward the lower funnel. We had an in-house built attribution model working super well, but it meant we didn't really know the true value of mid and upper-funnel marketing. The core question was always: where should we invest next year? To make it more complex, we don't have just one country; we have 16 countries live. It was super difficult to say whether the next euro should be spent in lower funnel in Germany or upper funnel in France.

– Building a Unified System

Anika: Our setup was fragmented and silo-focused. We decided we wanted a unified, full-funnel view to answer exactly where to spend next year. We didn't replace our existing setup; we added MMM on top so all three pillars complement each other. Data from attribution goes into the MMM, and calibration factors from the MMM go back into attribution. The incrementality test results also become part of the MMM. This is the best fit for bonprix.

00:13:50 – Securing C-Level Priority

Anika: The real challenge is making measurement a true business priority. We had weekly meetings with our project group and VPs of Sales and Marketing. We had bi-weekly meetings with our CMO and CTO, and still have monthly meetings because they see this as one of our most important projects. Alignment with business and financial goals is crucial because we use this data for big budget decisions. I intensively talk with our VP of Finance—even this morning. At bonprix, this project was prioritized as the number one topic last year, and for next quarter, it is still a top-three topic.

00:16:07 – Cross-Functional Execution

Anika: For execution, we needed a setup that brings together relevant perspectives. We have a cross-functional project team with a centralized owner—myself—alongside senior people from data science, data engineering, IT, customer and sales departments, and marketing operations. We also work with our external partner, Sellforte, who provides the models and the UI interface to see data, make simulations, and run data optimizations. We are also supported by agile masters focusing on PMO and change management.

00:17:25 – Driving Change and Trust

Anika: For change management, involving key stakeholders early is essential. We use structured stakeholder mapping to define who needs to be involved, from the board to the operational teams. I feel like I've been on a roadshow for the past months because I've been telling this story so often internally. Trust in data doesn't come automatically. We built an in-house MMM a few years ago, but we didn't use it because people didn't trust the data and we couldn't answer questions internally. This time, building trust in both input and output data is a top priority.

00:19:27 – Translating Insights into Decisions

Anika: The best model is useless if it doesn't change decisions. We are exactly in the stage where we want to translate insights into decisions across planning, budgeting, steering, and reporting. To build trust, we answer every question in every FAQ session. Another key learning is to move fast and iterate. We chose a specialized partner because we wanted to speed up instead of waiting for perfection. We started with an MVP at 90% certainty, aligned it with the board, and they loved it. Now we are speeding up and rolling out the next countries. I will hand over to Juha to show how it works.

00:21:52 – The Tech Engine

Juha: Thank you, Anika. My name is Juha Ludinen, CEO and co-founder of Sellforte. We are extremely thankful to Google for their thought leadership around the measurement triangle, and so happy to be part of this journey with bonprix. The implementation of the measurement triangle is a very simple yet powerful concept. Typically, incrementality tests, MMM, and attribution operate entirely separately, which is a massive problem. If your experiment says PMax ROAS is 5, MMM says it's 3, and attribution says it's 10, which one do you believe? They need to tell the same story and be consistent.

00:23:41 – The Three Implementation Steps

Juha: To achieve this, we follow three steps. First, incrementality experiments find the causal truth. Second, MMM acts as the integration layer between experiments and attribution. Third, attribution executes these insights at scale. Everything begins with experiments because they give the real causal truth: you show ads to one group, don't show them to a similar control group, and measure the incremental sales. We have done tens of conversion lift tests, catalog experiments, and geo-lift tests at bonprix to build an experiment library.

00:26:36 – Why MMM is the Layer

Juha: However, an experiment library alone doesn't tell you how to make optimal daily decisions on a channel. That's where MMM comes in. First, you cannot test everything—channels like TV, radio, out-of-home, and influencers are notoriously hard to test. Even for digital channels, it is unrealistic to run experiments on every channel in every country every year. MMM gives you the incremental ROAS for all channels, including those you haven't tested. Second, it gives you the marginal return for the next euro spent, which is what you actually want to optimize for.

00:27:51 – Executing with Calibrated Attribution

Juha: Finally, you need attribution for daily campaign optimization, whether it's bonprix's own model, GA4, or platform pixels. The challenge is converting causal truth into concrete bid recommendations. If an experiment says PMax ROAS is 5 and attribution says it's 10, you multiply the conversion value by a calibration multiplier of 50%. You must be very careful, as different campaigns have wildly different multipliers—Google Brand Search might need 30%, whereas YouTube might need 300%.

00:29:31 – Actionable Campaign Bidding

Juha: You need to make it actionable. If you need to scale PMax, which of the 20 campaigns do you scale? We translate high-level incrementality into campaign and ad-set level bid changes—for example, recommending changing a target ROAS from 9 to 8 to help a PMax campaign scale. We want to make activating these changes as easy as a green "Apply" button that flows directly to the ad platform. Finally, you constantly monitor whether the bid change achieved the forecasted outcome, keeping what works and reverting what doesn't to create a self-correcting, closed-loop optimization system.

00:32:03 – Q&A: Moving from Report to Action

Friederike: Many marketers struggle with the "so what" of MMM, viewing it as a strategic report that just sits on a shelf. For bonprix, how did you turn MMM into an active translation layer for day-to-day tactical decisions?

Anika: First, the data must be available. We implement Sellforte's data directly into our existing dashboards alongside our high-level tables. Second, people must understand it. We are onboarding 40 to 50 people, explaining what they see, and we have running FAQ documents. Third, getting senior leadership backing helps the complete organization feel confident and want to work with the data.

00:35:15 – Q&A: Offline Data and Store Visits

Friederike: Tim asks: "Is it possible to feed MMM with offline data and provide an impact assessment of offline advertising materials and their relation to store visits?"

Juha: Yes, it's possible, and we did this with bonprix. We select a primary KPI, typically revenue before returns, but we can include store and web traffic as secondary target KPIs. We also model customer lifetime value, as new and returning customer values differ. Depending on data maturity, you can input direct spend and circulation metrics, or go deeper by geographic area to confirm direct mail effectiveness via customer-based A/B tests or geo-lift experiments.

00:37:32 – Q&A: Building Trust with Finance

Friederike: Kevin asks: "How do you build trust with finance for the project?"

Anika: It's super important to understand where finance is at. Do they see that we need new answers to old questions? In my case, they totally understood the problem. We built an internal business case to prove how onboarding an MMM drives growth and budget efficiency. If you give finance a good business case and prove you can uphold the numbers you promised, it works.

00:38:50 – Q&A: Handling Unscalable Channels

Friederike: Martin asks: "How do you handle marketing channels that are not directly scalable or easily testable, such as SEO or AI-driven search?"

Juha: Email, SMS, CRM, and SEO are perfect examples. We look at history to understand their long-term value and incrementality—if we stopped them altogether, how much demand would we lose? It's really important to include them so we don't accidentally give their credit to paid channels. For future optimization, we can set channel options to "locked." We forecast their demand and spend correctly, but don't allow the optimizer to scale them up or down, keeping them safely in the mix.

Friederike: From a Google perspective, a good model uses organic volume as a control variable to ensure paid search isn't stealing credit from SEO. You can also input activity metrics like organic impressions or AI share of voice instead of budget inputs.

00:42:34 – Q&A: Managing Calibration Complexity

Friederike: George asks: "How has this triangle approach changed day-to-day decisions based on your MTA concretely? Do you calibrate results with incrementality on a regular basis?"

Anika: We are currently in the process of calibrating our dynamic attribution using the factors provided by Sellforte. This changes our marketing campaign steering and our daily budget shifting across departments. To the second question: yes, every incrementality test we do is fed into the Sellforte MMM. The UI makes it super easy to see where the data goes, and it all makes sense for us now.

00:45:21 – Q&A: The Biggest Bottlenecks

Friederike: Ingrida asks: "What became the biggest bottleneck—the modeling itself, or maintaining clean and reliable input data?"

Anika: Reliable input data is super important, but I must admit our internal team maintained an amazing data structure. The better and cleaner your data is, the easier your onboarding process with an MMM vendor becomes. If you trust the input data, you trust the output data more.

Juha: I can confirm the bonprix team has amazing data maturity. You never start from a perfect position, but starting the journey with marketing mix modeling helps justify the work that goes into continuously maintaining and improving your data streams.

00:48:22 – Q&A: Incorporating Brand Equity

Friederike: Eduardo asks: "How do you incorporate brand equity in the triangulation?"

: We look at different effects. There are short-term immediate effects seen in attribution windows. There are mid-term effects where a geo-lift test shows you keep losing sales for 30 to 60 days in a region even after you turn ads back on. Then there are long-term effects. We typically create a brand equity index out of constant survey data tracking awareness and consideration, and model that separately or tie it back to the demand model to show up to a 360-day effect on sales.

Friederike: From Google's side, we advise using brand lift studies to monitor direct traffic spikes, geo-testing, or inputting branded search volume via Google Query Volume into your MMM as early proxies for brand equity.

00:51:56 – Q&A: Measuring AI and LLM Traffic

Friederike: The last question is for Anika: "For bonprix, how are you looking into LLMs—ChatGPT, Gemini, etc.—as a relevant channel in the customer journey and measuring it in MMM?"

Anika: No, it is not part of our MMM today. We are closely monitoring and observing LLM traffic, but it is currently at a very low level for us. It might change in the future, as things change quickly, but it's not a super big traffic source for us right now.

Juha: Agentic e-commerce is growing for everyone. It's just a question of time before we add it into the model for each customer, including bonprix, once the moment is right.

Friederike: From a Google perspective, you can potentially put weekly traffic metrics from sources like Perplexity AI, ChatGPT, or Gemini into the MMM. Let's follow up afterward to answer how to integrate that for the future. I see we have answered all your questions. Perfect timing! Daria, do you want to take over?

Daria: It's extremely perfect timing, five minutes till the full hour. Yeah, thank you very much Anika, Juha, and Friederike...