The intersection of marketing and product by Jean de Bressy, Director of Martech at Delivery Hero

Originally, I moved to Düsseldorf to take German classes. Then I heard that trivago was actively hiring and thought my marketing experience could help land a job. When I joined in 2013, trivago was in hyper-growth and went from three hundred people to a thousand very rapidly. I did SEO for a year or two and eventually moved internally to a product position because it sounded like the coolest job in tech.

Since then I’ve been loving the career. Over time I moved to working on trivago’s travel magazine, the main tool our content marketing department was using to improve the site’s natural search ranking. This led to an opportunity in Zalando, the largest e-commerce for Fashion in Europe. Zalando was automating their entire performance marketing department and since I had prior experience with Display ads, there was a fit between my skills, and what they were looking for. Martech happens to be the mix of my two main domains of expertise.

Oh, it went brutally. I knew product management was a mix of different things that I liked doing: communication, tech and business. In the beginning I had no team and piggybacked on other teams resources to build features. I quickly learned that product managers really do not like to share their engineers. Then we started hiring a team of front end developers and built a few applications for trivago before moving to martech. We really started with the bare minimum, just an engineer sitting with me. I was reading articles and books on product management and applying new techniques as they were needed. The learning curve was extremely steep, something I still enjoy today.

My best guess would be to say it’s hubris! What can happen in scale ups is you start to think that everything is possible. So why not ditch Salesforce and build a copy of it “tailored for our needs”? And later down the road you realize that maybe it’s not such a good strategy for everything and you still need the occasional Salesforce or Mailchimp to take care of very important aspects of your tech stack.

At trivago it began with doing some inhouse bidding for SEM, since search is the main marketing channel in travel. It’s easier to focus on the key differentiators of a business when making these “build” decisions. Later on the logic was extended to almost all the tools — to the point where some decisions had to be reverted to “buy”, a process which is not easy since you have to shut off a tool you built and go for an external provider instead. But as it turns out, you can’t easily replace Salesforce.

A positive side effect of building internal tools is keeping a “can-do” attitude. Back at trivago, the general thinking was that you could solve your business challenges by building something. This mindset had the advantage of having us look for creative solutions to many problems raised during hypergrowth.

Oh, definitely. We tend to believe the workload on a new tool decreases drastically once it is delivered. But it doesn’t work that way: if we take the trivago magazine for instance, we originally had two engineers working on it, and even after its first and second redesigns, the team kept growing. The infrastructure needed to be updated, new marketing features were frequently requested, and so on. The workload even after a successful launch never went down.

My advice is to be extremely attentive to this drag along effect — teams in martech end up maintaining dozens of products and pipelines, some of which simply get outdated over time. I’m a big proponent of letting things go and focus on that part of your stack which produces most of the outcomes.

For context, Zalando is the largest e-commerce for Fashion in Europe. With operations in sixteen markets when I joined, it was twenty times bigger than my previous company, with well over ten thousand employees and a campus of several thousand tech people sitting in Berlin. They decided to move to radical automation in performance marketing, which was unique at that time.

I joined their newly founded team as a product manager for Facebook marketing. In the beginning, the product team juggled between building automation products and doing marketing manually — there was really no one to do any performance marketing but us. The structure was a classic matrixed organisation, with product managers, engineers and data scientists, and absolutely no marketing managers in the first years. And as odd as this seems (it was a fairly crazy bet) we ended up running a leaner, more efficient marketing team. We cut the cost of acquisition by focusing on just a few channels with highly automated ad formats such as dynamic ads; a better approach to budget management based on CLV and ROI targets; and a relentless focus on measuring the incrementality of our investment.

I’m happy that with the most recent changes in the marketing landscape the team at Zalando is recruiting marketing managers again — we were really good at optimisation problems but not having a marketing team slowed our innovation speed down.

When we thought about what to build first, we simply looked at what manual process took the most time from people and what value did each of these processes bring. We built a two-dimensional matrix showing in the top right corner what elements were both time consuming and very important.

Bidding ended up in this corner. In many companies, people come to work and export Excel sheets and run calculations to understand what the performance of their marketing was yesterday or last week. When done, they go and apply a new bid or a new budget manually to a set of campaigns, just to repeat the same thing the following day. This is clearly a very important process and a perfect target for an in-house solution. So we got our bidding automated for several channels.

Next, we took over the product feed on Facebook which is so important for e-commerce companies. It became our key product, and we kept adding information to the items sent to our partners. Finally we worked on automating simple campaign changes to run A/B tests faster, and change creatives through code rather than using the partners interfaces. So looking back the approach was simple — saving time on repetitive, yet important tasks.

Another layer of prioritization was the cost of building and maintaining tools. We tended to build simple proof of concepts to deliver value on a small scale, before incrementally investing more over time as the project succeeded.

The customer journey has less of an impact on the actual automation work than the type of business. Food delivery isn’t as much of a search driven industry as e-commerce and travel are. Display campaigns matter a lot, and within display there are less opportunities to work with dynamic product ads. This means creatives play an essential role in the performance of our marketing activities and we have to focus a part of our effort in understanding how to simplify the life of our different design teams and use the best performing assets possible in every campaign. Other than that the main topics are pretty much aligned across verticals at the moment, with incrementality, attribution and measurement being right left and center in 2022.

It confirmed what I saw in the past years, namely that there is an increasingly rapid shift towards supporting performance marketing with automation. Companies are hiring data scientists to understand the causal impact of marketing as it is the only way to really see the value of display campaigns in 2022. We also need data engineers to work with all the data flowing to and from our partners, since cookies are becoming less useful over time.

It actually feels like I get to build so much more by working with a series of teams. Every product manager contribute in their area of expertise and they have skills I actually do not possess, between data, engineering and applied science. Since I have always been interested in initiating projects and driving them in the right direction, it’s a perfect fit for what I like doing.

We work on data exchange topics with partners, feed management, which we also did a lot of at Zalando. There are also great experiments with geo-testing which we will hopefully scale this year, and systems related to creatives or campaign creation. The company structure is quite interesting, very distributed, with a performance team working for many local entities who also have a say on the development of their market.

The one thing I try to do is talk to our stakeholders as frequently as possible and make sure our teams know exactly what the organisation needs and where it is going. We are building a vision for the years to come so my role is to shape what the optimal portfolio of products looks like for us in the mid to long term.

I believe that our mission as martech builders is more in empowering marketers than replacing them. In other industries like car manufacturing, it’s easy to say “if we put a machine there, we can replace a human being”. In marketing it’s a bit different.

On the one hand it’s true that the increasing automation of platforms such as Google and Facebook means requiring less people to handle the same amount of work. On the other hand, there’s a lot of new work in marketing, new channels, new ad formats and endless possibilities. It’s now commonplace knowledge that creatives drive a large part of the performance of our campaigns, and no amount of automation can replace human creativity.

What automation did was mostly enable companies to rely more heavily on advertisers to make a lot of simple decisions. For example, dynamic product ads from Facebook and Google Shopping do a great job of showing the right product to the right customer. These ads leverage advanced machine learning products to deliver value at a scale otherwise impossible with manual tools. It is almost impossible to beat these solutions with custom engines built in-house (we tried!). Another example: on Facebook, we used to set bids for all of our ads. Now with features like Campaign Budget Optimisation, this layer of decision making is less relevant.

I believe that the time saved with in-house solutions unlocks more potential for marketing activities themselves, like competitor analysis, customer interviews, market research. Domains which require a lot more human intelligence than what we automate. A fun example for instance: trivago had some kind of an open mic culture for advertising ideas. Anyone could pitch their ad idea for Brand marketing and the best ones got filmed. There was also a chance to get featured in the ads.

One of these projects was the now very famous “Hotel? trivago” ad. Its success is one of brand marketing mysteries. The whole ad was just explaining how to use the website to compare prices, and hammered “Hotel? Trivago” at the end. It can’t get simpler than this. Yet this was the most efficient ad trivago ever shot; and the concept was scaled to all of the markets with local ambassadors. The Indian ad even starred our Indian country manager. When it came out on TV, he became a real star!

Recent history shows that performance marketers don’t get to be entirely left out. First, as marketing evolves, people migrate from an old, automated domain to a newer one. You do a great job automating Google and all of a sudden TikTok appears and you need someone who understands that platform and can build your strategy there.

Second, the market itself is quite competitive and it is difficult to find great marketers. Overall it looks like we are not seeing tech replacing marketing people — the current status quo is that marketing simply requires a lot more tech to be done effectively. Companies need to own their first party data and do advanced measurement to fend off the loss of signals the industry is confronted with lately. All things which require engineers.

Last but not least, a fully replaceable marketer is an elusive goal, because any automation solution is getting outdated fast. You can’t build it and forget it. Let’s say for now you built a system to run campaigns without human input. Next thing you know, GDPR appears and you need to make adjustments to your data sources. Then video ads and Youtube take over the entire industry. Then AR, and so on. Saving as much time as possible and being efficient is an amazing goal for building advertising tech, but hoping to use a solution on the long run without it requiring many changes is unrealistic.

am proof that it doesn’t hurt, at least. My replacement at trivago was also a marketing manager turned product manager. The domain matter expertise is really helpful to understand the challenges of our customers, and the project management side can be learned on the job.

At Zalando, I have seen a series of people switch from marketing to product management successfully. They were working on Facebook or Google, and started focusing on automation products for these channels. The experience they had as marketeers made the switch easier as they had to learn a new job but in the same domain. As the company evolved, we also needed people with more specific statistical or technical skills, in fields like attribution or incrementality.

So the marketing experience helps to solve specific problems, but the skill set required to solve complex data problems is extremely technical, meaning quite a few martech positions are going to be out of reach.

What I know for a fact is that you can understand the basics of marketing fast, given the correct foundation. Even though I initially studied political science I switched to marketing and then product and things went almost smoothly. So I am not certain this specific transition requires more or less preparation than any other.

However there’s a basic set of skills I believe are really useful to have. I generally look for them in candidates when I’m hiring. It’s a mix of behavioral economics, or macro / micro economics; and advanced data analysis and ideally programming — not just SQL but a bit of python. Because when you assign a performance marketing challenge to a person trained in statistics and behavioural economics, they can repurpose their skills in an environment with a lot of data which also requires understanding customer behaviour at scale. That makes the transition rather simple.

But then again, if you know advanced stats, python and behavioural economics, you shouldn’t have to worry about finding a job — ever. The trick is to get these people interested in Martech.