There is a specific kind of meeting that anyone working on brand campaigns has sat through. The performance dashboard is open, the numbers are flat or worse, and someone senior asks what the money bought. The person who planned the campaign knows the answer is complicated. They also know that “it’s complicated” is not an answer anyone accepts on a Tuesday morning.
The uncomfortable truth is that most measurement systems were built to count conversions, and brand work does not produce conversions on a schedule. It produces conditions under which conversions become more likely later. Those are very different objects, and the tools that measure one tend to be actively misleading about the other.
Last-click was never designed for this
Default marketing attribution assigns credit to the final touchpoint before a purchase. For a discount code shared in an email, that logic is roughly fair. For a video someone watched three weeks earlier, it is nonsense dressed as rigor.
The distortion runs in one direction. Channels that intercept demand look efficient, because they are always the last thing standing between a decided buyer and a checkout page. Channels that create demand look wasteful, because their contribution is spread thin across a long window and multiple sessions. Optimize hard against last-click for two quarters and you will systematically defund the thing that fills your funnel, then wonder why acquisition costs keep climbing.
This is not an argument against measurement. It is an argument for measuring the right object.
What brand campaigns actually move
Before touching a model, decide what a win looks like. brand marketing tends to move a handful of things that show up in different places at different speeds. Unaided and aided recall. Consideration among people who were previously indifferent. Direct and organic traffic that arrives without an obvious referrer. Conversion rate on channels you did not touch, because a warmer audience converts better on the same landing page.
That last effect is the one most often missed. A well-run top of funnel marketing effort will make your paid search look better without changing a single keyword bid. If you are not watching for it, you will credit the search team and cut the video budget.
The case that breaks the dashboard
Sometimes the signals point in opposite directions, and that is where campaign analysis gets interesting. One striking example is a youtube marketing campaign by Adidas that grew site traffic by 68% while branded search interest dropped by half. Read through a search-first lens, that looks like failure. Read it through a behavioural one, and a different story appears: people who saw the video went to the site directly instead of routing through a search engine first. The demand did not shrink. It stopped passing through the measurement point everyone was watching.
This happens more often than most teams realize, particularly with video marketing on platforms where the brand name is visible on screen and the destination is easy to remember. Search volume is a proxy for intent, and proxies break when behaviour shifts.
Reading branded search properly
None of which means you should ignore search signals. branded search remains one of the cheapest and fastest indicators of whether a campaign has landed. It just needs interpreting rather than reading.
Pull the data from google trends alongside your own Search Console numbers and compare shapes, not absolute values. Look at whether the baseline before the campaign has shifted upward after it, rather than at the spike during the flight. Spikes tell you people saw something. A raised floor tells you they remembered it. Check seasonality across at least two prior years, because a lot of celebrated brand lift is a calendar effect wearing a costume.
Also watch the lag. Depending on category and purchase cycle, the search response to a brand campaign can arrive anywhere from a few days to several months after the media stops running. Reading week-one data as the final verdict is how good campaigns get killed.
Building the campaign so it can be measured
The most useful decisions happen before launch. A campaign structure designed for measurement holds back something to compare against. The cleanest version is a geographic holdout: run the campaign in most regions, withhold it in a matched set, and compare outcomes across both. It costs you some reach. It buys you the only clean read you will get.
Where holdouts are not possible, staggered launches across markets give you a rougher version of the same thing. Even a two-week delay between regions produces a comparison worth having.
Keep tracking parameters consistent across every asset from the start, and separate top-funnel and bottom-funnel activity into distinct campaign entities. Mixing them inside one line item produces blended numbers that mean nothing and cannot be unmixed afterwards.
Where the media budget conversation goes wrong
The argument about media budget usually collapses into a split between brand and performance, as though the two were competing for the same job. They are not. Performance spend harvests demand that already exists. Brand spend determines how much demand there is to harvest next quarter.
A practical way to reframe the conversation: instead of asking which line delivered better return, ask what happens to your cost per acquisition twelve months after you cut the top of the funnel. Most teams that have run that experiment involuntarily know the answer, and it is not comfortable.
Making analytics tell a fuller story
Mature marketing analytics for brand work does not rely on one number. It triangulates. Media mix modelling gives you a directional read across long horizons. Incrementality tests give you causal evidence on specific decisions. Survey-based brand lift gives you the perceptual shift that behavioural data cannot see. Direct traffic, assisted conversions, and post-campaign conversion rate on untouched channels fill in the rest.
None of these is precise on its own. Together they are considerably more honest than a dashboard that reports a confident number to two decimal places and quietly excludes everything it cannot see.
The goal is not perfect attribution. It is a decision you can defend when someone asks what the money bought, and a measurement plan that was built before the campaign ran rather than reverse-engineered after it.
