Stat of the Week. 15% to 20% of a marketing budget can be reinvested elsewhere or returned to the bottom line without losing marketing return on investment (ROI). That range comes from a sample of the 400 marketing ROI engagements McKinsey had run with clients published in November 2013. (McKinsey.)
Some numbers stick around long after their original proof fades, and this one has lasted longer than most. The 15% to 20% figure appears in board meetings, agency presentations, and budget talks when someone wants to cut the marketing spend. People quote it confidently, often without knowing its source, and it has been used this way for more than ten years.
The number itself holds up. What hasn’t lasted is the idea that you could actually identify which 15% to 20% to cut. That made sense in 2013, but it’s much harder in 2026 because the tools for finding waste have changed a lot since then. It’s still worth reviewing your budget, but you can’t start the same way as before.
The number is real, but it’s now 12 years old.
Credit where it’s due, since the estimate was never a marketing slogan. McKinsey drew it from a sample of the 400 times its teams had worked on marketing ROI with clients. The finding was a typical range of 15% to 20% of budget that could be reinvested or returned without losing return (McKinsey, November 2013).
McKinsey explained the idea more clearly than the number itself. People often repeat the old Ogilvy line about not knowing which half of the budget is wasted. McKinsey argued it’s more like a fifth, and if you understand your data, you should be able to identify which fifth.
That last point is key. The range is so broad that it’s hard to prove wrong, so the real value was in saying you could find the wasted spend. Seeing similar numbers across different clients suggests you probably have savings in your budget too, but actually finding them is a measurement challenge.
The instruments changed underneath it
Since then, the industry has changed how it measures results. New privacy rules, changes to platforms, and the gradual loss of tracking tools that used to connect ads to outcomes have made measurement more fragmented than it was when the original estimate was made.
In February 2026, the Interactive Advertising Bureau (IAB) surveyed 430 US planning and analytics leaders at brands and agencies (IAB, February 2026). Of those who use advanced measurement tools, 60% to 75% said these tools fall short in key areas like rigor, coverage, timeliness, trust, or efficiency. The report goes further, describing marketing and media measurement systems as fundamentally broken, with data spread across disconnected systems that make consistent cross-channel measurement impossible.
Keep in mind, this is what practitioners say about their own tools, not an independent audit. The AIB report is also sponsored by vendors and introduces an IAB initiative, so it’s not completely neutral. Still, it matters because the sources used are people in the roles who would handle your budget reallocation.
The IAB points out exactly where the problems are. Marketing mix models tend to underrepresent connected TV, retail media, creator-led formats, gaming, and commerce media, which makes the returns from these channels look different than they really are. Buyers see gaps in how every channel is measured, even traditional media.
Reallocating from a broken model shifts where the error occurs
That level of detail is more important than general complaints. When you reallocate budget based on a model’s output, any channel the model doesn’t measure well is at risk of being judged unfairly. If measurement isn’t even across channels, the one that looks cheapest might just be the one that’s measured best.
One insurance company learned this lesson over nine months. They did a thorough review, put every budget line into one model, and made the biggest cuts where the measured return seemed weakest—mainly in retail media and connected TV. These areas only had about a year of good data, compared to several years for search and direct mail, so the model’s estimates were less reliable than they appeared. The costly mistake was that the problem didn’t show up right away. The reallocation looked like a win at first, but the shortfall appeared in the second half of the next year, after the review had already been approved and filed.
The order of operations is what has to change
The solution is to change the order of steps, which costs less than the big measurement programs most teams try first. The old way was to measure everything, rank it, and then reallocate based on those rankings. But if the rankings aren’t equally reliable, this approach can lead to confident mistakes.
To flip the process, start by identifying the two or three key allocation decisions that really matter for the year, since most annual plans have fewer important choices than budget lines. These decisions are usually specific, like whether to keep or cut the brand budget, whether retail media deserves more funding, or if the always-on search budget is just capturing demand you already had. Then, focus your measurement efforts only on what’s needed to make those decisions. Running a geo holdout or an incrementality test on one debated line can answer a key question in a few months, and costs much less than rebuilding your whole measurement system. The evidence you get is the kind finance teams trust, because it’s built to answer their specific questions.
This new order also makes it easier to get internal funding, since you’re linking a test to a decision that already involves spending money.
Start smaller than the problem
What are the two or three decisions that really matter in this plan? First, write them out as questions with numbers, then remove any measurement requests that don’t help answer them.
Which channels in your model have less than two years of reliable data? First, label those lines as low-confidence before the reallocation meeting, so no one confuses a rough estimate with a poor result.
What can you test on one debated budget line this quarter? Start by picking the line with the most internal disagreement and set up a holdout test for it. Resolving a big argument is more valuable than just improving the model a little.
That 15% to 20% is likely still in most budgets, and the reason to pursue it is even stronger as budgets get tighter. What’s different now is that you have to earn the right to act on it one decision at a time, and each resolved debate makes the next budget change easier to justify.
Which decision would you test first? Let us know which budget line you’d choose.
Sources
McKinsey, “Smart analytics” can tap up to 20% of lost ROI,” November 2013 (the 15–20% is a typical range observed across a sample of roughly 400 McKinsey client marketing-ROI engagements, not a survey finding; vintage stated in-article. Quotes from Jonathan Gordon paraphrased, not quoted verbatim.)
IAB, “State of Data 2026: The AI-Powered Measurement Transformation,” February 2026 (n=430 US buy-side brand and agency planning and analytics decision-makers, fielded by BWG Strategy; the 60–75% is a range across several separate questions and sub-bases, and is self-reported perception of measurement performance, not a measured audit. Report sponsored by Dstillery and OptiMine; IAB states findings were not influenced by sponsors.)