BrIM — Brand Incrementality Measurement answers two questions every brand planner cares about: how much attention is my brand winning, and did my campaign actually move it?
It reads consumer search demand — how often people look for your brand versus its competitors — from Google Trends, and turns it into two views. The Share of Search tab tracks where your brand stands over time. The Incrementality tab measures whether a specific campaign caused a real lift, over and above what would have happened anyway.
Search interest is a fast, unbiased read on brand demand: it's what real people do when no one is watching. It tends to move ahead of market share, which makes it a useful early signal for planning and a credible yardstick for campaign impact.
How the two work together
Share of Searchis your always-on scoreboard — it tells you where the brand stands and where it's heading. Incrementality is the campaign lens — it takes a moment on that scoreboard and asks whether a specific investment caused the move. Both draw on the same underlying signal: what people are searching for. The data refreshes nightly, so the picture keeps up with the market.
Share of Search
Of all the category search interest going to your brand and the competitors you track, how much is yours — and which way is it trending?
What it's based on
For your brand and each named competitor, we pull search interest from Google Trendsin your chosen market. On any given day we add up the interest across the whole tracked set and express each brand as a percentage of that total. Because it's a share, the tracked brands always add up to 100% — so a competitor gaining means someone else is giving ground.
How to read the tab
- The headlineshows your brand's average share across the period you've selected, and the change versus the equal-length period just before it — so you can see momentum at a glance.
- The chart plots daily share for every brand over time. Toggle brands on and off to declutter, switch between stacked and line views, and export the exact picture as PNG or CSV.
- The tablelists each brand's average share, its change versus the prior period, and a small trend sparkline (peaks in green, troughs in red).
Good for
- Spotting momentum and competitive shifts early
- Seeing seasonality and category-level demand
- Tracking the drag of always-on brand activity
Keep in mind
- It's a share, not volume — yours can rise while the whole category shrinks.
- It's relative to the competitors you pick — add or remove one and every share re-bases.
- Trends data is sampled and rounded, so very small brands can round to zero on quiet days.
Where the numbers come from.Google Trends doesn't report how many people searched for a brand — it reports a relative index: the busiest day for the busiest brand in the set is called 100, and every other value is scaled against it. An index of 50 means "half as much search interest as the peak", not "50 searches". BrIM always requests the whole brand set together, so every brand is measured on the same ruler, in the client's market only.
From index to share.On each day we add up the whole set's index values and ask: what slice of that total belongs to each brand? That slice is the Share of Search. Because it's a percentage of the day's total, the set always sums to 100% — if one brand rises, the others must give something up, exactly like market share.
When there are more than five brands. Google Trends compares at most five things at a time, so bigger sets are split into batches of five — and your own brand rides in every batch as a common reference point (we call it the anchor). Think of two group photos taken separately with the same person standing in both: because that person's height is known, the photos can be scaled to match each other. Numerically, we compare the anchor's values across batches day by day and take the median ratio (the middle value — resistant to the odd weird day) as the rescale factor. Checked on live multi-batch data: the factor is stable to about ±3% across a full 92-day overlap. If your own brand had very little search volume, this bridge would get noisier — something we watch for.
Why old history looks slightly different. Google only provides day-by-day data for recent windows (roughly the last 90 days). Anything older arrives as weekly totals, which we spread across days — statisticians call this interpolation, and the charts mark it as "stitched". A side effect: a week of searches added together can register a small brand that any single day would round down to zero, so small brands can appear to step up or down right at the boundary. That's a property of the source data, not a real market movement — read long-range charts for the trend, not for any single day, and let the shaded region plus the caveat chip tell you where the stitching lives.
Freshness.Feeds refresh every morning and the numbers rebuild nightly — and immediately after a client's feeds finish rebuilding. The "Data through" label shows exactly how current the chart is; amber means the pipeline is running behind, and pipeline failures email the team automatically.
FAQs
Why don't these numbers match the Google Trends website?
The Trends website re-scales every search you do to its own 0–100 scale, so two visits rarely show the same numbers. BrIM always fetches the full brand set together — one shared scale — then converts to shares and keeps a consistent daily history. That's what makes brands comparable to each other and trends stable over time.
Why did older shares change slightly after a refresh?
Every daily fetch re-scales its whole window against the newest peak day, and the batch rescale factors are recomputed on every pipeline run. Tiny retroactive shifts are normal and expected. Large ones are not — the pipeline's built-in checks fail the nightly run and email the team if the numbers stop adding up.
Why does a small brand's share look different in older history?
Old history is stitched from weekly data. A week of searches added together can register a small brand that any single day would round to zero — so at the boundary between weekly and daily data, small brands can appear to step up or down. The shaded region on the chart marks exactly where that happens.
How fresh is the data?
Feeds refresh every morning and the numbers rebuild nightly — plus immediately whenever a client's feeds finish rebuilding after an edit. The "Data through" label shows recency; amber means the pipeline is behind, and failures email the team automatically.
Why do brand suggestions show global entities?
Google Trends entities are global topics — there is no per-market list to search from. The market matters at measurement time instead: every data point is collected for the client's market only.
A different signal, the same share math. Google Trends measures revealed behaviour — what people actually typed into a search box. YouGov BrandIndex measures stated perception— what a representative panel says about a brand when surveyed (for example, "Have you seen an ad for this brand recently?" for Ad Awareness). BrIM turns either signal into a sharethe same way: on each day it adds up the brand set's values and gives each brand its slice of the total, so the set sums to 100%. Only the underlying number changes — the share arithmetic is identical.
Scoring bases. Every YouGov metric is calculated over a chosen base — the group of respondents it's measured on. Total uses everyone surveyed; Aware narrows to people who already know the brand; Opinion narrows further to people who hold a view of it. A narrower base usually lifts the raw score and can reorder how brands compare, so the base is fixed per client and kept consistent over time.
Net metrics and the zero floor. Most YouGov metrics run 0–100, but several are netscores — positive mentions minus negative ones (Buzz, Impression, Index, Quality, Value, Reputation, Satisfaction, Recommend). Those can go negative when criticism outweighs praise. A share of a daily total can't be negative, so any negative day is floored to 0 for the share calculation (the raw score is kept for reference). This keeps every share in the 0–100% range and summing to 100, but it means a floored brand simply contributes nothing that day rather than dragging the total down.
Share of Ad Awareness ≠ Share of Search.Because the two sources measure different things — stated perception versus revealed search — their shares will not match, and neither is "more correct". A brand can be highly aware but lightly searched, or heavily searched with modest awareness. Always read a YouGov client's headline as "Share of {Metric}" (the card and workspace label it), not as Share of Search.
Licensing. YouGov BrandIndex data is licensed. Client-facing display follows the YouGov licence held for the account — treat these figures as licensed YouGov data and confirm the terms before publishing them externally.
FAQs
Is Share of Ad Awareness the same as Share of Search?
No. Share of Search measures what people actually searched for (revealed behaviour); Share of Ad Awareness measures what people say in a survey (stated perception). They answer different questions and won't match — a brand can be widely known yet rarely searched, or the reverse. Read each on its own terms rather than expecting them to line up.
What do the scoring bases (Total / Aware / Opinion) change?
They change who the metric is calculated over. Total is everyone surveyed; Aware narrows to people who know the brand; Opinion narrows further to people who hold a view of it. A narrower base usually raises the score and changes how brands rank, so keep the base consistent when you compare across time or brands.
Why can a YouGov score be negative?
Some metrics are net scores — positive mentions minus negative ones (Buzz, Impression, Index, Quality, Value, Reputation, Satisfaction, Recommend). On a bad day the negatives can outweigh the positives and the value drops below zero. Because a share of a total can't be negative, those days are floored to 0 before the share is computed — so a floored day contributes nothing rather than pulling the total down.
Can I show these numbers in a client deck?
Client-facing display follows the YouGov licence for the account. Treat the figures as licensed YouGov BrandIndex data and confirm the licence terms before publishing them externally.
Incrementality
Did a specific campaign actually cause a lift in your Share of Search — over and above what would have happened without it?
The idea: a counterfactual
The honest question isn't “did search go up during the campaign?” — plenty moves for reasons that have nothing to do with you. It's how much of the movement the campaign caused. To answer that, BrIM builds a counterfactual: a best estimate of what your Share of Search would have been during the campaign if it had never run.
The model learns your normal behaviour from a baseline period before the campaign, using the competitor brands as a control for whatever was happening in the category anyway. The gap between what actually happened and that counterfactual is the estimated lift. (Under the hood this is a Bayesian causal-impact model — but you never have to touch that; you give it a campaign, it gives you an answer.)
What you provide
A campaign name, its start and end dates, and total spend. That's it. Advanced users can change the baseline window (it defaults to the 3 months before the start date); most planners never need to.
What you get back
- A verdict — Strong, Moderate, or No significant effect — plus a plain-English read of what it means.
- Posterior probability— how confident the model is that a real causal effect exists (e.g. 91% means “very likely real”).
- Average daily incremental change — the size of the lift, in share-of-search points versus the counterfactual.
- Cost per unit change— spend divided by the lift, so you can compare efficiency across campaigns. It's shown only when there's a meaningful positive effect (dividing spend by a null or negative lift isn't meaningful).
- A chart — your actual Share of Search against the dashed counterfactual, with a shaded confidence band and the campaign window highlighted. Exportable as PNG or CSV.
Good for
- Proving whether a burst actually moved demand
- Comparing campaign efficiency on a like-for-like basis
- Separating your effect from category noise
Keep in mind
- It measures lift in search demand, not sales — a strong signal for attention, not a direct revenue number.
- It needs enough clean pre-campaign history to learn a stable baseline.
- If several things ran in the same window, treat the size directionally — the model can't untangle overlapping activity.
The question this answers."Did the campaign cause extra search demand — or would it have happened anyway?" Comparing before/after averages can't answer that, because demand moves for lots of reasons (seasonality, competitors, news). A causal read needs a picture of what would have happened without the campaign, to hold your actual results against.
How the model learns. During the baseline window (the months before the campaign), a Bayesian structural time-series model — a statistical model that learns patterns and how certain it is about them — studies how your share normally moves and how it tracks your competitors. Competitors are useful because they experienced the same season, the same market events, the same news — everything except your campaign.
The counterfactual.When the campaign starts, the model stops learning and starts predicting: "given everything I learned, and given what the competitors are doing right now, here is the path your share should have taken without a campaign." That predicted path is the counterfactual— the dashed line on the chart. The measured effect ("lift") is simply the gap between the solid line (what actually happened) and the dashed one.
How sure is it?The model doesn't pretend the counterfactual is exact — it draws a band around it, the 95% credible interval: the range it believes the no-campaign path plausibly lived in. If your actual line sits comfortably outside that band, the campaign very likely caused the difference. That confidence is summarized as the posterior probability of a causal effect— in plain words, "how convinced the model is that the campaign, not chance, moved the number."
The verdict tiers translate that probability into words: at 95% or higher we call the evidence strong; between 90% and 95%, moderate — treat the size directionally; below 90%, not significant— the data can't rule out chance, so the cost-per-unit-change is withheld rather than dressed up as precision.
How we know the engine works.We injected known, artificial lifts (+5%, +10%, +20%) into real client data and asked the engine to find them: every one was recovered within about 1.1 percentage points, and a zero-lift control run was correctly called "not significant". An independent implementation of the same method agreed within ~1.2 points on every case. The engine is also deterministic — the same inputs always produce the same answer.
What to say in a deck. Effects are measured on search demand, not sales. Concurrent events — a competitor's burst, a PR moment — share credit with your campaign. And short or thin campaign windows widen the uncertainty band. Stating these limits alongside the result is what makes the number defensible.
FAQs
What is "posterior probability" in plain terms?
It's the model's confidence — after seeing your data — that the campaign, and not random chance, caused the change. A posterior probability of 97% means: given how your share and your competitors actually behaved, in 97 of 100 plausible no-campaign worlds the model considered, your real result came out ahead. "Posterior" is the statistical word for "after seeing the evidence" (its opposite, "prior", is what was believed before looking).
What does "not significant" actually mean?
It means the model couldn't rule out that the movement would have happened anyway. It is NOT proof the campaign did nothing — it means the data doesn't support claiming lift. That's also why the cost-per-unit-change is withheld: dividing spend by an unestablished effect produces a precise-looking but meaningless number.
Can I put the cost-per-unit-change in a client deck?
With a strong verdict, yes — describe it as "spend per point of incremental share of search per day". With a moderate verdict, present it as directional. And remember it measures search demand, not sales.
What can bias an incrementality read?
Anything else that moved search demand in the same window: a competitor's campaign, a PR event, a seasonal shift the baseline didn't learn, or a baseline window that overlaps another of your own campaigns. Pick clean baselines, and take the overlapping-campaign warning seriously when the form shows it.
Why does re-running the same campaign give the same numbers?
The engine is deterministic — the same inputs produce the same posterior probability and effect every time. That's a feature for defensibility: a number you present today can be reproduced exactly next quarter.