Marketing Science: 15% Budget for 2026 Growth

In digital advertising, we’re drowning in data but still struggle to prove our marketing actually works. Incrementality is the solution, it’s how you get past simple correlation to see what’s actually causing people to buy. It’s the difference between watching sales go up while an ad is running and knowing, for a fact, that the ad *made* those sales happen. Knowing that ‘why’ is how you stop wasting money and build a growth engine that actually grows the business. So how do you prove your campaigns are the cause and not just a coincidence?

The Limitations of Correlation in Marketing Measurement

For too long, marketing has used metrics that feel right but give a totally misleading picture of success. We see sales spike after launching an ad campaign and just assume the campaign did it. This completely ignores other factors, like a holiday season that was going to drive sales anyway or a competitor’s screw-up that sent customers our way. A rise in sales during Christmas might look like it’s from your ad push, when really, it’s just Christmas.

Take last-click attribution. This model, which is somehow still everywhere, gives 100% of the credit for a conversion to whatever the customer touched last. It’s simple, but it’s also dead wrong because it says nothing about whether that customer would have bought anyway and completely ignores the top-of-funnel work that got them there. A customer might see a display ad, research your product for a week, and then finally click a branded paid search ad right before buying. Last-click only credits the search ad, ignoring the display ad that did the heavy lifting. This broken thinking leads to terrible budget decisions, like pouring money into bottom-funnel channels that just harvest existing demand instead of creating it.

This problem goes way beyond last-click. Even sophisticated multi-touch attribution (MTA) models can fall into the same trap if they’re only looking at observed correlations. They use stats to find patterns, but a pattern isn’t proof. A strong correlation between your ad impressions and conversions might just mean you’re really good at showing ads to people who were already about to buy your product, not that you’re persuading anyone. If you don’t actively test for cause-and-effect, you end up optimizing for an audience that was going to convert anyway, which is basically just paying to watch a parade you didn’t organize. This is why moving to incrementality is a practical necessity if you want to make smart decisions.

Defining Incrementality: What Truly Drives Action?

Incrementality measures the true causal impact of a marketing activity by answering one simple question: “How many more sales did we get because we ran this ad, that wouldn’t have happened otherwise?” It isolates the net new conversions or revenue that are directly because of your marketing, filtering out all the noise from organic traffic, existing brand loyalty, and other factors. It’s a huge departure from just watching two lines on a graph go up at the same time.

The whole idea behind measuring incrementality is to run a proper controlled experiment. You have one group that gets the marketing message (the “test” group) and a similar group that doesn’t (the “control” group). Any statistically significant difference in their behavior, like a higher conversion rate in the test group, is the incremental lift you can attribute directly to your marketing. This scientific method gives you a much more defensible read on cause and effect than just looking at observational data ever could.

For instance, in a simple A/B test, you might show an ad to one random group of people and a public service announcement to an identical group. The difference in conversion rate is your lift. A more complex method for bigger campaigns is a geo-lift study. You could run a big TV campaign in specific test markets (say, Denver and Phoenix) while leaving similar markets (like Salt Lake City and Austin) as controls. By comparing sales data and foot traffic between the test and control cities, you can quantify exactly how much impact the campaign had. This approach gives you a solid framework for proving the real value of your marketing, which lets you do things like confidently say, “The campaign in Ohio drove a 5% incremental sales lift, so let’s roll it out to these five other similar markets.”

Methods for Measuring Causal Impact

Getting from correlation to causation means changing your measurement strategy to one that’s designed to isolate variables. There are a few solid ways to measure incrementality, and the right one depends on your campaign’s scale, your data setup, and what you’re trying to figure out. A massive brand awareness campaign needs a different approach than a direct-response email push.

1. A/B Testing (Randomized Control Trials): This is the gold standard because randomization is the cleanest way to isolate a single variable. Users are randomly put into a test group (sees the ad) or a control group (doesn’t). Because it’s random, you can assume all other factors are balanced between the groups, so any difference in outcome must be from your ad. Most major platforms, like Google Ads and Meta Ads Manager, have built-in tools that make it easy to run these experiments and compare strategies.
2. Geo-Lift Studies: As I mentioned, these are great for big campaigns that are hard to track at a user level, like TV or radio ads. You pick test and control cities that are demographically and economically similar, then run the campaign only in the test areas. The sales lift in the test cities compared to the controls is your incremental impact. You have to be super careful in picking your markets and doing the analysis to make sure you get a valid result and account for things like people from a control city driving into a test city.
3. Matched Market Testing: This is a lot like a geo-lift study but used when you can’t randomize at the user level and the campaign hits a whole market. You find pairs of markets that have historically tracked very closely in terms of sales, then apply your marketing to one market in each pair. The big challenge is making sure the markets are truly matched and that nothing else (like a new competitor opening up) messes with one market and not the other during the test.
4. Synthetic Control Methods: What if you have a single, huge intervention, like a national campaign, and no good control market exists? For these situations, you can use synthetic controls. A statistician basically builds a “ghost” control group by taking a weighted average of a bunch of other unexposed markets or units that, when combined, perfectly matched the pre-campaign trend of your test market. It requires a lot of historical data and serious stats knowledge but can give you causal answers when a normal A/B test is impossible.

On top of these direct experiments, even big analytical tools like Marketing Mix Modeling (MMM) are getting smarter about incrementality. Old-school MMM just looked at historical correlations, but modern MMM can ingest the results of these kinds of experiments to calibrate its models, giving you a much truer read on each channel’s contribution. Companies like Nielsen are pushing these hybrid models, combining their econometric analysis with real experimental data to provide a much clearer picture of marketing effectiveness, which they cover in their latest reports. The point is to stop just watching what happens and start designing your measurement to find out why it’s happening.

Building a Culture of Incrementality

Adopting an incrementality mindset is a cultural and operational shift, not just a matter of buying a new tool. It means you stop “setting and forgetting” your campaigns and start embracing a constant cycle of testing and learning. This requires real investment in your technology, your people, and your processes.

First, your leadership has to be on board. If your execs aren’t championing this, your team will never get the political cover they need to run tests that might prove a “successful” campaign actually had zero incremental impact. That news can be uncomfortable, but you have to be willing to hear it to get better. Marketing teams need the room to run experiments where the outcome is uncertain. A test that “fails” to show lift is still a win because it teaches you what doesn’t work and stops you from wasting more money on it.

Second, your data infrastructure has to be rock solid. You can’t measure incrementality if your data is a mess. You need clean, consistent data across all your touchpoints, which means things like customer data platforms (CDPs) that can stitch together a user’s journey, good analytics tools, and the ability to link an ad exposure to an actual sale. It’s no surprise that a 2023 IAB report found companies with mature data strategies are 2.5 times more likely to see big ROI improvements from their marketing. Investing in data governance isn’t sexy, but it’s mandatory.

Third, get everyone talking. Data scientists, marketers, product managers, and the finance team all have to be in the same room (or Slack channel). When marketing finds a 2% incremental lift from an ad highlighting a new feature, the product team can use that data to help prioritize their roadmap, and finance can project the real revenue impact. This way, incrementality insights actually drive business-wide decisions instead of just living in a marketing dashboard.

Finally, this is a loop, not a project. You’re never “done.” The whole thing is a constant cycle of generating hypotheses (What do we think will work?), designing tests to see if you’re right, and then using what you learned to make the next campaign better. For example, you test three ad creatives, find one has a 10% higher incremental lift, kill the other two, and reallocate that budget to the winner. That’s the optimization loop. This “experimentation culture” is where real competitive advantage comes from, because you’re always making your marketing spend smarter.

Overcoming Challenges in Incremental Measurement

While measuring incrementality is powerful, getting it right has some real hurdles. The biggest one is just the sheer complexity of running a statistically valid experiment. We’re all running fast and have tight budgets, so it’s hard to carve out the time and resources for a proper test. You need someone who actually understands statistical significance, p-values, and how to spot problems like contamination between your test and control groups. That expertise isn’t always sitting around in the marketing department.

Another huge pain point is fragmented and messy data. To see the incremental impact, you need a single view of the customer journey, which means pulling together data from your CRM, ad platforms, web analytics, and offline sales systems. In my experience, you spend half your time just cleaning and connecting data before you can even start the analysis. If your data is garbage, even the fanciest incrementality model will give you garbage results. Clean data means your test results are trustworthy, which means the decisions you make based on them (like shifting a million-dollar budget) are sound. That’s the dividend from investing in data engineering.

Also, the results can be hard to stomach. An experiment might show that your big, splashy campaign had a negative incremental lift, meaning you actually spent money to lose sales or that the cost of each new conversion was sky-high. It’s tempting to try and explain away bad results, especially if you’re the one who championed the campaign, but a real incrementality-driven team has to accept the findings as a lesson and move on. It’s about being disciplined.

Lastly, you have the “holdout problem.” To run a test, you have to intentionally not show your ads to a control group, which means you’re willingly giving up potential sales from that group in the short term. Trying to convince a sales director to let you hold out a group of potential customers is a tough conversation. You have to be able to clearly explain the long-term benefit of learning what really works, framing the discussion around maximizing total business value, not just hitting a short-term campaign goal.

Getting to a full incrementality measurement framework is a journey. It demands ongoing investment in your tech and talent, and a serious commitment to the scientific method. You won’t get it perfect on day one, but by building a system for continuous improvement, you can start making sure every marketing dollar is being judged by its true, causal impact.

If you can prove your marketing causes growth while your competitors are still just guessing, you win. Embracing incrementality is how you do it. By moving past simple correlation and digging for the causal impact of every dollar you spend, you can find huge efficiencies and drive real growth. To see how this gets even more powerful, check out how AI attribution can achieve 90% accuracy by 2026, which will only sharpen this understanding.