You walk in with ROAS on the rise. Your CFO asks how many of those sales would have happened without advertising. You make the case for video as a brand builder. The CFO counters with search ROI. You propose a new channel. The CFO wants proof before signing off.
Those are fair questions. But dashboards don't always answer them. We ran six scenarios by France Hassenforder of marketing science consultancy Ekimetrics and Jonathan Dupasquier of data-driven marketing consultancy M13h to separate the arguments that hold up from the ones that don't.
Use this to build your case. But as both point out, these fights are won long before they start, by bringing finance up to speed.
"The best thing is to start these conversations as early as possible and get them on board with your vision, rather than clashing with them the moment they tell you to cut," Hassenforder said. Dupasquier agreed that the groundwork comes first. "Explaining how platforms count their conversions, what a model measures and why the numbers don't always tell the same story is a prerequisite," he said.
1. "I'm being asked to deliver the same results on a 15% smaller budget"
An across-the-board cut has one advantage: it's easy to explain. Everyone loses 15%, so nobody feels singled out. Whether it makes economic sense is another matter.
The other instinct is to rank channels by ROI and drop the one at the bottom. That isn't much better. "The best approach isn't to cut the channel with the lowest ROI. That's where incrementality really comes into play," Dupasquier said.
That's because a channel's returns change with how much you put into it. The first euros reach the easiest buyers. The next ones chase harder conversions, or hit the same people again. A great average ROI can hide the fact that the last euros do very little.
That's where marginal ROI comes in: what one more euro brings in, or what you'd lose by pulling it. A channel can top the year-end review and still be the first place to look for savings. And a channel with a weaker average return may still deserve to keep part of its budget.
Marketing mix modeling (MMM) helps with that call. It ties sales to media spend while accounting for other factors such as pricing, promotions and seasonality, and it doesn't depend on cookies. Its response curves estimate what different spending levels are likely to produce.
Dupasquier said he regularly runs scenarios with clients ranging from a 25% cut to a 25% increase in the marketing budget. The goal is "to help them get comfortable with their choices and shape the budget around their growth goals." In some simulations, a better mix keeps revenue intact on lower spend. In others, it doesn't. Marketers need to be able to show that, too.
So a good case doesn't promise 15% savings through some mysterious "optimization." It compares several options: cutting across the board, trimming the most saturated channels, dropping certain campaigns. Each one comes with an estimate of what it saves and what it puts at risk.
Hassenforder singled out brand spend. It's "easy to attack because it has less of a direct effect on short-term sales," she said. Cutting it can improve the near-term numbers while hurting future results. And that's assuming the analysis properly credits brand's contribution to begin with.
A brand that stops advertising for a year or more sees sales fall 16% on average after one year, and 25% after two years.
The cost of going dark is well documented. According to the Ehrenberg-Bass Institute, a brand that stops advertising for a year or more sees sales fall 16% on average after one year, and 25% after two. Binet & Field and Warc, two other go-to references for marketers, have written extensively on the subject.
Finally, beware of campaigns kept on life support to make everyone happy but starved of the money they need to work. A line item is pointless if what's left of the budget can no longer reach enough of the audience.
2. "My CFO wants proof that extra spend will pay off"
"It works, so let's put more into it." That logic holds up, until the channel starts to saturate.
A good ROI on the current budget doesn't guarantee the same return on a bigger one. That's the real question: will the next euro win one more buyer, or mostly retarget someone you already reach often?
Hassenforder recommends going "back to the MMM response curves to identify the channels that still have headroom. Then, if the request is about a specific channel, test the increase on a small scale."
The idea behind a geo test is fairly simple: increase media weight in some regions and compare the change in sales with similar control regions. Running one is harder. It takes enough volume and the right regions, and a local promotion or a change in distribution can't be allowed to skew the comparison.
"There's nothing more powerful than a well-calibrated test," Dupasquier said. Instead of signing off on a projection alone, finance can look at the results of an experiment.
It also lets marketers ask for the money in two stages. The first tranche funds the test. The next one is released if the result clears a threshold agreed in advance. Everyone knows what they're looking for, how much they're willing to spend to find out and what will justify the next step.
Then there's the question of what counts as a result. Revenue? Margin? New customers? Dupasquier pointed to "the importance of setting a clear goal and of tying ad optimization more closely to the company's actual profitability." A €100 order on a heavily discounted product isn't worth as much as a €100 order on a high-margin one.
This goes all the way down to platform settings. If the algorithm optimizes for revenue, it will do what it's told, which isn't necessarily what the CFO wants. Dupasquier mentioned "the option of changing the value shared with platforms to give more weight to the most valuable conversions, without necessarily passing along the exact margin on each sale."
Meta does this with its "Value optimization for profit." Google takes a similar approach. By cross-referencing cart data against product costs entered in Merchant Center, Performance Max and Shopping campaigns can optimize for gross margin rather than revenue. Exact costs aren't required.
The same logic applies to acquisition. Boosting orders from existing customers and winning new buyers both grow revenue, but they serve different goals. Before asking for more money, marketers need to know what kind of growth they want to buy, and at what price.
3. "I want to test a new channel, but I have no track record"
You have to start somewhere. Influencer marketing and retail media would never have made it into media plans if every advertiser had waited for its own proof before spending a euro. That still doesn't justify funding a campaign because "everyone's doing it."
Hassenforder recommends "putting 90% of the budget into known channels and strategies and keeping 10% for new approaches." The goal is to protect the core while making room to learn. The exact split will vary by company. What matters most is setting the money aside before the opportunity shows up.
We recommend keeping 10% of the marketing budget for new approaches.
Public benchmarks on other advertisers' results can help frame the test. "Your vendor may have already worked on that channel, in your sector or elsewhere," Hassenforder said. That gives you a ballpark. It doesn't guarantee that your brand, with your product and your creative, will get the same results.
The real work starts with the question the test should answer. Are you looking for incremental customers? An audience other channels don't reach well? A lift in consideration? Without a precise answer, you risk running a campaign, getting a few encouraging metrics and still not knowing whether to keep going.
Next comes sizing the test. To measure an effect on sales, you need to estimate the smallest lift the test can detect amid normal swings in the business. "A budget that's too small can leave the experiment unable to reach a conclusion, even if the channel works," Dupasquier warned.
That's the trap of the "let's just try it" test. You spend little, learn little, and then the result becomes company lore: "We already tried that. It doesn't work."
Dupasquier argued for a more structured approach: "Part of your media plan is designed to test things and figure out what works for the company." The idea of a "learning agenda" is having a moment (we'll dig into it in an upcoming report). It organizes experiments around the questions the company needs to settle, instead of stacking up pilots whenever a vendor pitches one.
Marketers also need to be clear about where the money comes from. Taking budget from another channel tests a reallocation. Adding money tests incremental spend. If sales stay flat, the first case means something very different from the second.
Finally, map out the next decision before launch. What result would justify a bigger budget? Which one would justify another try? When should you stop? "A test doesn't need to succeed to be useful. It needs to let you make a decision," Hassenforder said.
4. "My CFO wants to cut a channel with poor ROI in the MMM, but I think it builds my brand"
You championed MMM. Then the model says to cut video, and now you're the one explaining why it shouldn't be taken literally. It's an awkward position, but not necessarily an inconsistent one.
A model can capture a campaign's immediate effects well but have a harder time capturing its contribution to the brand. It can also struggle to tell apart two channels that always run together. So before challenging the conclusion, look at what drives it: the available data, the lag effects the model assumes and the uncertainty around the estimated ROI.
Dupasquier pointed to the baseline, the share of sales the MMM doesn't attribute to any of the marketing activity measured during the period. "The risk is reading it as revenue that would come in no matter what," he said.
But that baseline can benefit from brand equity built up over the years, just as it benefits from distribution, product experience and buying habits. Just because a model doesn't attribute a sale to a recent campaign doesn't mean advertising played no part in it.
Hassenforder said part of the answer lies in brand models, "models that analyze how investments contribute to brand strength, on top of their effect on sales."
The goal is to understand which channels drive awareness, consideration or preference, then how those factors support the business. A growing number of advertisers are rolling out multi-layer MMMs, with one layer dedicated to the strategic question of brand building.
Here, actually cutting the channel to see what happens is a hard sell. "A short experiment can miss the gradual erosion of the brand. An experiment long enough to see it can leave damage behind," Hassenforder said.
So a holdout test, which works well for some conversion channels, isn't the default answer when the question is brand.
Without a dedicated model, marketers can cross-reference brand studies from Kantar, YouGov and the like. They can also use proxies such as Google search trends for the brand name or follower growth on social accounts. None of this will magically produce an unassailable ROI. But it will at least keep a poorly measured contribution from being mistaken for a nonexistent one.
Hassenforder also made an argument that should resonate with finance: a strong brand can raise prices with less risk of losing customers. "Demand for a strong brand is less sensitive to price. That's well documented," she said. So branding helps defend margin, not just an awareness score.
Still, the brand argument shouldn't become a catch-all defense. Weak creative, too much frequency or poor targeting can drag down a useful channel. You can defend its role while admitting its execution needs work.
5. "My results are good, but my CFO thinks some of those sales would have happened without advertising"
The CFO is probably right. Better to concede that up front before defending the numbers.
A sale attributed to a campaign isn't necessarily a sale caused by that campaign. Someone can click an ad after already deciding to buy. Or get retargeted just as they were coming back to the site on their own. In both cases, the platform can take credit for a conversion it didn't really create.
The problem is especially visible in retargeting, branded search and campaigns aimed at existing customers. These channels reach people who are close to buying. That's what makes them useful, and what makes their stellar ROAS sometimes misleading.
The platforms themselves are starting to act on this. Google has made its Conversion Lift incrementality tests more accessible, in part by lowering minimum budgets.
Meta has rolled out an "incremental attribution" option, which aims to optimize campaigns toward conversions its models estimate wouldn't happen without ad exposure. The pitch is shifting toward incremental sales rather than the ones that are easiest to claim.
Still, the two approaches differ. Optimizing on an estimate of incrementality is not the same as verifying it with an experiment. A test compares results from a group exposed to the campaign with a control group that isn't, or compares matched geographic regions.
Hassenforder sees geo tests as especially useful here. "A geo test lets you show, on a small scale, what changes when the channel is switched on or off. It's crystal clear," she said.
But don't turn that into a hard rule. In branded search, cutting ads can be offset by organic clicks. It can also open the door to a competitor bidding on your name. The result depends on the conditions the test runs under.
So the conclusion isn't necessarily "this channel is useless." It may be that some of the spend works and some works much less. You can lower frequency, suppress certain audiences or reinvest in acquisition.
A media plan can become more useful to the business while looking less flattering in platform dashboards.
There's a consequence worth flagging up front: reported ROAS is likely to drop. The conversions that are easy to claim disappear from the reporting, while the money goes after customers who are further from buying.
That's the paradox. A media plan can become more useful to the business while looking less flattering in platform dashboards. If marketing and finance haven't agreed on how to evaluate it, whoever did the cleanup may end up having to justify it.
6. "Finance wants a fixed annual budget, but I don't know where the best opportunities will be in six months"
Committing to an annual budget shouldn't mean locking in how it's spent. In November, nobody knows the best way to spend every euro through the following December. Pretending otherwise doesn't protect the investment. It mostly risks tying the company to choices that have stopped making sense.
That's the distinction finance needs to accept: an annual budget commits the company to an amount, not to a fixed allocation. The split should be able to shift with results, demand and business priorities.
Hassenforder pointed to weather-sensitive businesses. "An ice cream brand shouldn't advertise exactly the same way whether spring is hot or rainy," she said. A windshield replacement company, on the other hand, may need to step up when conditions drive demand. The budget calendar doesn't control the weather, so don't let it stop you from reacting to it.
Flexibility also lets marketers scale up a successful test, jump on a sales opportunity or ease off a saturated channel. But it's easier to get if the rules are set before it's needed: how much money can move, on what criteria and with whose sign-off?
"A quarterly review can be a way to update the simulations and revisit upcoming investments with finance and the agency," Hassenforder said. The point is to agree in advance on what will justify a reallocation, so flexibility itself doesn't have to be renegotiated with every request.
The 10% exploration budget Ekimetrics recommends covers some of that. But testing an unknown channel and shifting a proven campaign are two different things. The first takes a learning budget. The second takes room to maneuver within the existing plan. You can plan for one and still have no room for the other.
Reviewing the plan regularly also shouldn't put the whole budget up for grabs based on quarterly ROAS. Play that game, and brand spend will be put on trial again and again, asked to deliver right away what it's meant to deliver over time. The criteria for reallocating money need to respect each investment's time horizon. Otherwise, agility becomes an excuse to sacrifice the long term.
That freedom also has to exist outside the spreadsheet. Hassenforder said the media agency must be able to shift spend. Contracts, buying commitments and decision timelines all need to allow for it. A budget that's labeled flexible but already tied up in firm commitments leaves little real choice.
So before promising the CFO that money will move to the best opportunities, make sure you'll still have control over it. Budget agility gets negotiated when the plan is built, not on the day it needs to change.

