Insights

Why your marketing science team isn’t getting traction

Your models can be strong, and your recommendations can still go nowhere. Why marketing science teams struggle to influence decisions, and how to change that.

A see-saw: a large grey block labeled “Great model. No decision.” is outweighed by a small blue block labeled “Decent model. Tied to a decision.” Caption: A decent model tied to a decision is a budget lever.
Sundar Swaminathan — Author

Sundar Swaminathan

Marketing Science Advisor

In this article

Stay sharp, sign up
to our newsletter

Why your marketing science team isn’t getting traction

Marketing science teams can invest heavily in their analytical skills while failing to give the same attention to influencing decisions. But the value of their work isn’t determined only by how sophisticated the models are — it’s mostly determined by what the business can do differently because of them.

Data science is supposed to make people’s lives easier, or better, or both. The technical work is only half the job. The other half is organizational: getting buy-in, speaking the right language with different functions, and building credibility.

Below, we’ll talk about why marketing science teams fail to get the credit they deserve and, more importantly, how to fix that.

4 ways marketing science teams lose credibility

Strong analytical teams often lose credibility because of small mistakes in the way they work with stakeholders. Here are the four most common mistakes I see:

1. Leading with method instead of the decision

Marketing science teams get stuck because they sell methodology to people who buy decisions. They focus on explaining the MMM approach, the data sources, the model architecture, and all the assumptions behind it.

But what do stakeholders care about? What that means for their business.

Compare these two ways of presenting the same work:

  • “Here’s our MMM approach, how we built the model, and the results we got.”

  • “Here’s where we’d move $2 million in budget, why we’d move it, and what we expect to happen.”

The first conversation puts the cognitive burden on the stakeholder. You’ve explained the analysis, but left them to work out what to do with it. Should they move budget? From where to where? And why is that the right move?

In the second version, the decision on the table is moving $2 million. You’ve given them a why and an expected outcome. From there, they can push on the assumptions, challenge the logic, and poke holes in the recommendation. It’s a much healthier debate.

2. Delivering insights too late

Many data scientists want to deliver perfect insights. But while you’re pushing a model’s R-squared from .85 to .99, stakeholders are managing expectations and dealing with organizational pressure.

Stop chasing perfection. That doesn’t mean cutting corners. It means asking whether another week of analysis is likely to change what you recommend, or just make you more comfortable recommending it.

Stakeholders don’t want precision or certainty, because that’s not possible anyway. What they want is confidence and a solid understanding of risk.

.85 gives them that.

You just have to be okay with it.

3. Failing to take ownership

One of the values Uber had, and one I see a lot of companies replicating now, is “act like an owner.” Many of us, especially in tech, are equity holders in our companies, which makes us literal owners of the company’s future.

However, many data scientists don’t carry that ownership through to the end. They forget the job isn’t to deliver an amazing model, build a great pipeline, or write a beautifully clean SQL query. The job is to add profit to the business — simple as that.

If you deliver a study and walk away, you haven’t finished the job. You need to understand if the team acted on the recommendation and, if not, why that’s the case.

Think about it: if you’ve done all that work, but nobody is using it, the business isn’t making better decisions, and nothing is improving, then all that effort has gone to waste. That’s not a great feeling.

4. Treating stakeholders as customers instead of partners

The last place data scientists can lose traction is relationship management. In many companies, data science gets filed under “support function”, whereas it should be positioned as a partner instead. Your marketing stakeholder needs you just as much as you need them.

A partnership means you understand each other’s priorities and can challenge each other without undermining the relationship.

A few things that tend to get in the way:

  • Surprising stakeholders in the room - The first time a channel owner hears a negative result should never be in the readout itself. Have that conversation before.

  • Overselling early - Walking back a result you presented too confidently undermines trust in your judgment. Be clear about what you know, what you don’t know, and how confident you are in the findings.

  • Being submissive - It’s your job to challenge the strategy with what the data says and be a champion of truth. But that’s not the same as challenging everything or turning every disagreement into a confrontation. It’s still a relationship. Sometimes you’ll need to push harder; other times, you’ll need to listen and find common ground.

The common thread is that good analytical work isn’t enough. You also need to build the kind of relationship where people trust you, involve you early, and are willing to hear what the data has to say.

Align, Communicate, Deliver

So how do you build that kind of relationship? I recommend getting into the habit of doing three things: Align, Communicate, and Deliver.

Align: Speak the same language

Think of your stakeholders as an audience you’re telling a story to. To keep them interested, bear these two questions in mind:

  • What do they care about?

  • What do they need to understand to make a decision?

Different stakeholders look at the same analysis through different points of view. Here’s roughly what each one cares about:

  • CMO: budget, growth, and risk. What they can defend upward, and what they should stop funding.

  • Finance: incremental revenue, payback, marginal ROI, and how confident they can be in the numbers.

  • Channel owners: what this means for their budget, their targets, and the path forward.

Imagine you’re recommending a budget change. The CMO needs to understand how it supports the growth plan. Finance wants to know the expected return and the assumptions behind it. The channel owner is much more interested in hearing what changes in their daily work and what they’re expected to do next.

This is not changing the findings according to the audience. It’s making clear what those findings mean for the decisions each person has to make.

Communicate: Meet regularly

I don’t understand teams whose only interaction with stakeholders happens through a ticketing system. It makes the relationship purely transactional. I always aim for at least a weekly meeting with stakeholders.

Get in a room and talk through priorities, updates, challenges, and blockers. This puts some humanness back into the relationship and opens the door to richer conversations.

I’m not advocating for another pointless status meeting. I’m saying you should aim to stay close enough to the business so that your work stays useful.

Deliver: Building credibility incrementally

Like in any relationship, you can’t promise the world from day one. You’ll probably fail if you try. Instead, build credibility through a series of small actions:

  • Start with a small, winnable decision where the test is cheap and the answer will change something visible. Early wins fund the harder conversations later.

  • Make a prediction before the result comes in, then show how it landed. A track record of called shots matters more than any methodology doc.

  • Be the first to report your own bad news, a test that failed, or a model that got revised. It’s the fastest way to earn trust with Finance and skeptics alike.

  • Agree on the measurement framework and decision thresholds before results exist, so no one can renegotiate the rules once they see the answer.

Influence is built by being right in ways people can verify, being useful before being asked, and making it easy for others to act on what you found.

So if you’re struggling to get traction, take a step back. Don’t just ask, “How do I improve the analysis?” Ask, “What would help someone act on it?” Sometimes you need more evidence. Other times, you need a clearer recommendation, an earlier conversation, or better follow-through.

Stay experiMENTAL,
Sundar

Stay sharp, sign up
to our newsletter

Find your edge

Book a demo now and you'll get a live, expert-run walk through of how Paramark can help you with incrementality measurement and experimentation.