Brand and media measurement still sit in silos: four gaps and how to close them
Brand tracking, marketing mix modelling and creative testing are usually run by separate teams on separate data. Four gaps between brand and media measurement, a fix for each, and what AI changes.
Survey. Odyssey Beauty Signal, UK. Online panel interviews with consumers aged 16 to 65 who had bought beauty products in the past six months, collected every month and weighted. 4,538 interviews from September 2025 to August 2026. Awareness: "Which of the following brands do you know?" Consideration: "Which of the following brands would you consider buying?" Purchase: "Which of the following brands have you purchased from the last 6 months?" Each question covered 77 beauty brands.
Most companies measure their brand with surveys and their media with attribution and marketing mix models. Separate teams run the two sides on separate data, so results from one side rarely inform decisions on the other.
There are four gaps between brand and media measurement, and each can be closed. AI makes closing them more valuable, because AI tools can only analyse brand and media data together when the data can be joined.
I spent most of my career in media and marketing technology before joining Odyssey, where I now work with survey researchers. In my experience, each of these methods is usually done well by the team that owns it. The problems appear where one team's results should inform another team's decisions.
What brand measurement and media measurement cover
Brand measurement is usually split across four types of survey research:
Segmentation identifies the groups of buyers a brand should target.
Positioning research describes what the brand should stand for among those groups.
Communication evaluation measures whether people noticed a campaign, linked it to the brand and understood its message.
Brand tracking measures awareness, consideration, preference and brand associations over time.
Media measurement has two main parts:
Media planning and buying uses attribution, marketing mix modelling and incrementality tests to estimate what each channel contributed to sales.
In-channel creative performance measures attention and completion for each version of an ad on platforms such as TikTok, YouTube and Instagram.
Four gaps separate these activities, and each has a fix.
Gap 1: brand studies that use different samples and brand lists
Gap 1 concerns all four brand studies, and what is said about the brand outside the survey.
Segmentation, positioning, communication evaluation and brand tracking are often bought from different suppliers in different years. Each study then has its own sample, competitor list and question wording, so the results cannot be compared directly.
A positioning study may recommend that the brand should be associated with a particular territory. The tracker cannot report progress towards it if it did not ask the same question of the same people.
A survey also covers only the brands on its list, usually the 20 or so best-known brands when the study was designed. AI answers and social media cover every brand in the category. In our UK beauty data, several brands that lead those two sources are not among the 20 best-known brands in the survey:
Social media. Three of the ten brands with the most organic mentions are outside the top 20: The Ordinary, NYX and NARS.
AI answers. Four of the ten brands that AI assistants most often name first are outside the top 20: Redken, Glossier, The Ordinary and Kérastase.
Fix 1: one survey dataset for segmentation, positioning and brand tracking
Use one tracker that covers segmentation, positioning, communication evaluation and brand tracking in the same survey. The same respondents answer every part, with the same competitor list in every wave, so a question raised in one part can be answered from another without a new study.
Measure AI answers and social media on the same brand list, and review the list each wave. A brand that is often mentioned there but is missing from the survey is a candidate for the next wave.
This is the easiest of the four fixes, because it is a decision about which research to buy.
Gap 2: media buying is not measured against the brand
Gap 2 lies between brand tracking and media planning and buying.
Brand research sets the direction for the brand, but media buying is rarely evaluated against it.
In many companies, brand tracking results are presented once a quarter by a different team from the one that reviews the media model each month. Neither report uses figures from the other.
Fix 2: brand tracking data in the marketing mix model
A marketing mix model (MMM) relates weekly spend per channel, usually over about two years, to an outcome such as revenue. It estimates what each channel contributed and how the total would change if the budget were moved.
It needs no cookies or individual-level data, which is one reason many companies use it instead of last-click attribution. When revenue is the only outcome, the model estimates short-term sales effects and contains no information about the brand.
Brand tracking data can be added to the model. In an example with invented figures, a model with revenue as its only outcome estimates that 10 million spent on TV returned 6 million over the quarter, and the same spend on search returned 25 million. It recommends moving budget from TV to search.
When awareness and consideration from the tracker are added, the model can attribute part of the later search sales to demand that TV created. TV's estimated return becomes 16 million and search's 15 million, and the recommended allocation changes. Revenue is still the outcome the model optimises, with the brand measures as inputs.
Invented figures, with both panels on the same scale. With brand tracking data in the model, TV is also credited with the later sales that follow the awareness and consideration it built.
Tracker data can enter the model in three ways, from the simplest to the most complete:
As explanatory variables. Awareness, consideration and share of search sit alongside spend and help separate the effect of media from changes in the market.
As a brand component. The slow, long-term effect of upper-funnel media is attributed to the channels that caused it instead of to the baseline.
As a brand equity model. Upper-funnel spend affects the tracker's measures, which affect sales over a longer period. The example above uses this form.
The tracker has to be designed for this. It needs to report monthly in the markets the model covers, with samples large enough for a monthly result to be reliable, and the model and the brand measures need to be re-estimated together on a fixed schedule.
Most trackers were not designed this way, which makes this the hardest of the four fixes.
What brand tracking data can tell media planning today
A tracker can inform media planning before any model is built, because it shows the shape of each brand's funnel.
In our UK beauty data, CeraVe is known by 47% of beauty buyers, which places it 26th of 77 brands. It ranks 16th on consideration and 8th on purchase, so awareness is the measure where it ranks lowest.
The figures come from Beauty Signal, Odyssey's monthly online survey of UK beauty buyers aged 16 to 65, with 4,538 interviews between September 2025 and August 2026. The full method is at the end of this article.
A funnel like CeraVe's gives direction in three areas:
Media planning. Buy reach among people who do not know the brand yet, in the target groups the segmentation identifies.
Creative. Make ads that introduce the brand, its products and the associations the positioning is built on.
Evaluation. Judge the campaign on awareness and position in the tracker, and on reach and attention in the channels. Measured on conversion, a campaign that increased awareness could look unsuccessful.
Gap 3: creative and media buying are planned apart
Gap 3 lies inside media, between creative performance in the channels and media planning and buying.
Media teams are often divided as well. In-channel creative performance, such as which opening holds attention and which version people watch to the end, is often managed by the media buyer or a production partner. The media plan and the mix model are managed by someone else.
Each group optimises against its own metrics. Neither the mix model nor the creative reports connect an ad's performance in the channel with its effect on the brand.
Fix 3: buying, production and analysis in one team
Agency and in-house set-ups can both work, as long as the people who buy media, produce ads for the channels and analyse the data plan against the same target and use the same reports.
When the tracker shows a funnel like CeraVe's, all of them need to see it at the same time, because acting on it changes both the media plan and the ads.
Gap 4: one ad, two separate evaluations
Gap 4 runs across brand research and media, between communication evaluation and creative performance in the channels.
Communication evaluation in the survey and creative performance in the channels evaluate the same ads. Different teams own them, they report on different schedules, and their results are rarely compared.
A version with high completion rates may communicate little of the intended message. That is only visible when the two results are compared.
Fix 4: compare creative results in the channel and in the survey
Share in-channel creative results with the people who run communication evaluation, so that each ad is assessed both on its performance in the channel and on whether its message was understood.
The data usually exists already, in different teams and tools, so the change is mainly organisational:
One person is responsible for creative results in both the channels and the survey.
Both sets of results are reviewed in the same meeting.
How AI changes brand and media measurement
AI tools make data analysis fast and cheap. A marketing team can ask a question about its data in plain language and receive a table within seconds, and its competitors can do the same.
AI cannot combine studies that were collected separately, with different competitor lists and at different times. If the segmentation, the tracker and the campaign test cannot be linked at respondent level, an AI model cannot link them afterwards.
AI also lowers the cost of starting a company, so new competitors can enter a category faster than before. Some brands already have a large presence in social media and AI answers without being among the 20 best-known brands in a survey, as the UK beauty examples above show. Reviewing the competitor list each wave, as in the first fix, therefore becomes more important.
Where to start
Check whether your segmentation, positioning study, communication evaluation and brand tracker use one sample and one competitor list. If they do not, bring them into one survey dataset.
Measure AI answers and social media on the same brand list as the survey, and review the list each wave.
Use each brand's funnel in the tracker to guide media planning, creative and campaign evaluation.
Check whether your tracker can report monthly in the markets your mix model covers, with samples large enough for a monthly result, before adding brand measures to the model.
Let the people who buy media, produce ads for the channels and analyse the data plan against the same target and use the same reports.
Make one person responsible for creative results in both the channels and the survey, and review both in the same meeting.
Survey. Odyssey Beauty Signal, UK. Online panel interviews with consumers aged 16 to 65 who had bought beauty products in the past six months, collected every month and weighted. 4,538 interviews from September 2025 to August 2026. Awareness: "Which of the following brands do you know?" Consideration: "Which of the following brands would you consider buying?" Purchase: "Which of the following brands have you purchased from the last 6 months?" Each question covered 77 beauty brands.
Social media. All Ears, UK. 36,840 organic mentions of beauty brands on TikTok, YouTube, Instagram and podcasts from June to September 2026. Sponsored mentions are excluded.
AI answers. Odyssey AI visibility, UK. Answers from ChatGPT, Gemini and Microsoft Copilot to questions about beauty products, collected every month from January to September 2026. The ranking counts how often each brand is named first, 800 first places in total.
Marketing mix model example. The figures are invented to illustrate the method and do not describe a client.
AI use. The figures were computed with Odyssey's data engine. An AI assistant helped draft the text, which was checked against the data.
Matilda Rydow is CCO at Odyssey. She has spent her career in marketing technology and media measurement, and in recent years on how they connect to the brand, in the technology and in the organisation.
