Brand media gets cut first because most measurement only captures what happens in the few weeks after an ad runs. On that window, performance channels almost always look better. Marketing mix modeling can correct the picture by measuring two things separately: the short-term lift a campaign produces, and its slower contribution to base sales, the demand a brand would enjoy even with no media running. When both are on the table, brand and performance can be compared on equal terms.
Why brand loses the budget argument
Picture a consumer brand under pressure to trim spend mid-year. The team opens its reports. Paid search, retail media and social retargeting show strong returns within days. Brand video and TV show modest returns, or look unprofitable. The obvious cut is brand.
The problem is the measuring stick. Performance channels mostly harvest existing demand. A shopper who already knows and likes the brand searches for it, and the search ad gets the credit. Brand media builds that familiarity in the first place, but its payoff arrives slowly and spreads across every channel, including ones that aren’t paid media at all.
Cutting brand rarely causes a visible drop next week. Sales erode gradually as the brand’s pull weakens. By the time it shows up, the budget decision is months old and the link between cause and effect is hard to see.
Short-term effects: what most models capture
A standard mix model relates weekly media activity to weekly sales. It uses adstock to capture carryover, the fact that an ad keeps working for a few weeks after airing. Brand channels usually get a longer carryover than performance channels, which already helps.
But carryover measured this way typically spans weeks, not years. It captures the short-term lift: the extra sales a campaign drives while it’s fresh in people’s minds. It leaves out the lasting part.
Long-term effects: where base sales come in
Every mix model splits sales into two parts. Incremental sales are what the model attributes to media, price, promotion and other drivers. Base sales are what’s left: the volume the brand would sell anyway, from habit, distribution and reputation.
In many models, base is treated as a fixed or slowly drifting line and left unexplained. That’s where brand effects hide. If years of consistent brand investment raise the level of base sales, a model that treats base as a given will never credit brand for it.
There are several ways to bring that effect into view:
Let the base move, and explain it. A time-varying baseline allows base sales to rise and fall over time. The next step is to relate those movements back to sustained brand investment, with a long lag.
Use brand metrics as a bridge. Some models link media to awareness or consideration data from brand tracking, and then link those metrics to base sales. That two-stage approach makes the mechanism explicit: media builds the brand, the brand lifts base.
Look at price sensitivity. Strong brands often hold volume better when prices rise. If price elasticity changes as brand investment changes, that’s another measurable long-term effect, and one revenue and finance teams tend to care about.
Give the model enough history. Long-term effects need long data. Two years is a typical minimum for a mix model; separating lasting brand effects usually needs more, ideally including periods where brand spend varied meaningfully.
Approaches like these are part of what separates a reporting exercise from rigorous marketing mix analysis, because they model the channel interactions and incremental impact that a short-window view misses.
A hypothetical example
Consider a beverage brand comparing brand video with paid search. For example, a short-term model shows:
- Brand video: $0.80 back for every dollar within the carryover window.
- Paid search: $2.20 back for every dollar.
On those numbers, cutting video looks sensible. Now say the brand extends the model with a moving baseline linked to video investment over three years. Suppose it estimates that video also lifts base sales by an amount worth another $0.90 per dollar over the following year. Video’s total return becomes about $1.70.
Search still looks efficient. But the model also shows that part of search’s return depends on demand that brand activity created. If video is cut, search volume would likely soften too. The real comparison is no longer $0.80 against $2.20. It’s a combined system in which cutting one channel quietly weakens the other.
These numbers are illustrative. The shape of the result, a short-term view understating brand, is the reason to run the analysis.
Common mistakes
Judging brand on a performance window. Comparing channels on a 7- or 28-day return guarantees brand loses. Report short-term and long-term returns side by side.
Treating base sales as fixed. If base is a constant, the model can’t credit anything to brand investment. Ask how your model handles base, and whether it can move.
Over-claiming the long term. Long-term estimates are less certain than short-term ones. Present them with confidence ranges. Overselling them destroys credibility with finance faster than leaving them out.
Ignoring what performance depends on. Model interactions between channels, so the dependence of harvesting channels on brand-built demand becomes visible.
Taking it to finance
The goal isn’t to prove brand always wins. It’s to put both kinds of return on the same page. Show each channel’s short-term return, its estimated long-term contribution through base sales, and the uncertainty around both. Show what the model expects to happen to other channels if brand is cut.
That changes the conversation. Instead of “brand has a lower ROI,” the question becomes “what are we willing to give up over the next year to save money this quarter?” Consumer brands that can answer it clearly are far less likely to cut the media that keeps their base intact.
