01
Data quality
Clean, consistent data across channels, with tracking checked against orders and enquiries.
UK marketing science service
Every channel claims credit for the same sale. We test the channels against each other, measure what each one adds, and reallocate budget to the combinations that grow revenue.
Select a node to see its links
Last click hides the channels that build demand. Tests and models show what each one really adds.
Attribution, incrementality tests and marketing mix modelling combined, so the contribution of every channel is measured against the sales it helped create.
How marketing science works
No single method shows the whole picture. Attribution shows the paths people take, tests show what would have happened without a channel, and models show how channels combine. Used together, they show where growth comes from.
Attribution shows the touchpoints people use before they buy. It is useful for understanding journeys, but it over-credits the last click, so it is checked against tests.
Holdout groups and geo tests show the sales a channel adds beyond what would have happened anyway. They are the strongest evidence for where budget should go.
Marketing mix models estimate how each channel, including offline media, contributes over time. They guide budget across the year and show where returns start to fall.
Marketing science is judged by the budget it moves and the growth that follows. We report the contribution of each channel, the cost per incremental sale and the reallocation expected to add most revenue.
What moves results
Measurement only helps when it is designed well, uses clean data and feeds decisions. These levers decide whether the numbers change where money goes.
01
Clean, consistent data across channels, with tracking checked against orders and enquiries.
02
Holdouts and geo tests designed with the right size, duration and control before spend starts.
03
Clear rules for crediting touchpoints, checked against tests so they do not over-credit the last click.
04
The sales each channel adds, measured separately from the sales it would have won anyway.
05
Models that include offline media and seasonality, refreshed as the data changes.
06
The point where another pound of spend adds less, found so budget stops before the returns fall.
07
Budget options compared, with the expected growth and risk for each.
08
A regular review where the numbers change budget, not just reports.
What we run
Data, testing, modelling and planning in one programme, so every channel's contribution is measured and budget follows it.
01
A review of tracking, data and current attribution, with the gaps sized in pounds.
02
Attribution rules and data joins built so the channels can be compared on one basis.
03
Holdout and geo tests designed, run and read, with the results reported plainly.
04
Models built and refreshed to show the contribution of each channel over time.
05
Budget scenarios that show where the next pound adds most, with the risks stated.
06
Expected sales and enquiries by channel and season, updated as results come in.
07
Consistent events, offline conversions and first-party data joined across the mix.
08
One view of channel contribution, cost and incremental sales, reviewed each month.
09
Recommendations taken to the budget owners, with the evidence and the expected effect.
Beyond this channel
Marketing science tells every other channel what it is worth. It stops credit going to the wrong place and moves budget to the channels that grow revenue.
BOOK A DISCOVERY CALLTests show how much paid search adds beyond organic and brand demand, so bids are set on true return.
Holdouts show which social audiences add sales, so budget moves from cold reach to what converts.
Holdout groups show how much email adds beyond sales that would have happened, so the list is valued fairly.
Incrementality stops retargeting taking credit for sales other channels created.
Geo tests show how brand media lifts search and sales, so awareness budget is judged on its effect.
Tests of conversion changes are read as revenue per visitor and sales, so wins are real and counted once.
How it connects
Data, insight and customer behaviour flow across channels, so every interaction works harder and drives more revenue.
01
Set bids on the return paid search truly adds, using incrementality tests.
Marketing science contributes
Marketing science learns
02
Decide which audiences and creative get budget, based on holdout results.
Marketing science contributes
Marketing science learns
03
Show what brand media lifts, so awareness spend is judged on sales.
Marketing science contributes
Marketing science learns
04
Measure conversion wins as revenue, so the lift is real and counted once.
Marketing science contributes
Marketing science learns
Common mistakes
Measurement that is wrong does more harm than no measurement. These habits send budget to the channels that look best rather than the ones that grow revenue.
Last-click reports credit the final touch and starve the channels that create demand earlier. Budget decisions based on them favour retargeting and branded search.
A campaign compared with last month cannot show its effect, because the season and the market moved too. Controls are the difference between a guess and a result.
A dashboard that nobody uses changes nothing. Measurement earns its cost only when it moves budget at a regular review.
Each mistake sends budget to the wrong channels and leaves growth on the table. Better measurement pays back quickly when it changes where money goes.
How we work
We start with the data and the current picture, test the channels that matter most, then model the mix and agree where budget should move.
The engagement · six steps
We review data, tracking and current attribution, and size the gaps in pounds.
Tests and models designed with the right controls, sizes and durations.
Holdout and geo tests run and read, with results reported plainly.
Mix models built and refreshed to show each channel's contribution over time.
Budget scenarios agreed and moved towards the channels that add most revenue.
Contribution, cost and incremental sales reviewed monthly at a decision meeting.
| Segment | Attributed | Incremental | Cost efficiency | Status |
|---|---|---|---|---|
| Paid search | 72% | 46% | 54% | Overcredited |
| Retargeting | 64% | 22% | 30% | Overcredited |
| Brand media | 18% | 58% | 46% | Undervalued |
| 44% | 52% | 70% | Working | |
| Organic and content | 38% | 40% | 62% | Testing |
Relative reach by segment · illustrative
Gaps become the roadmap: retargeting and brand media first.
Reviewed every monthWhat we report
MARKETING SCIENCE · Measurement
Sales credited to each channel on the agreed rules
Sales each channel adds above its holdout or control
How well the mix model fits the observed sales
Share of spend covered by a current test or model
COMMERCIAL · Budget
Spend reallocated towards channels with higher incremental return
Expected and realised revenue from the budget changes
Spend divided by the sales each channel adds
Months until measurement work is repaid from budget gains
A revenue plan for this channel, not a media checklist. The metrics are how we measure it.
BOOK A DISCOVERY CALLTiming
The first tests can show which channels are overcredited within weeks. Models become reliable as data builds, and the biggest gains come from acting on them consistently.
The audit and first tests show the largest mismatches between attributed and incremental sales. Those findings can change budget before the models are complete.
Data quality, the volume of sales and spend, the length of the buying cycle and how often the decision meeting acts on the findings decide how quickly the picture becomes reliable.
We agree the test design, the model scope and the decision cadence before work starts, so each review judges the same evidence.
Marketing science pays back through the budget it moves. Agreeing the decision cadence first means the measurement is used, not just reported, and its value is counted in revenue.
Who it suits
Marketing science works best where there is meaningful spend across several channels and enough sales to test. Pick a model to see how it does the work.
Ecommerce with several channels
When a budget runs across search, social, email and reach, the models show where the next pound adds most. Channel holdouts and budget scenarios test what happens when money moves, before the money moves.
B2B with long cycles
Tests and models show which early channels create pipeline, which attribution alone cannot see. Long cycles need a view across months, so the model tracks pipeline from first touch to signed deal.
Brands with offline media
Mix models bring offline media into the same view as digital, so brand spend is judged alongside everything else. Regional tests give a clean read on offline spend, so the budget follows the lift it creates.
Subscription businesses
Lifetime value and retention are modelled by channel, so acquisition spend follows the revenue it brings over time. Cohorts are tracked from first purchase, so the payback period for each channel is known before the budget is set.
Using tests, models and data to measure how each channel contributes to sales, and to move budget to the channels that grow revenue.
Attribution assigns credit to touchpoints on the paths people take. Incrementality measures the sales a channel adds beyond what would have happened anyway. We use both, and check attribution against tests.
A statistical model that estimates how each channel, including offline media, contributes to sales over time. It guides budget across the year.
No. Models estimate effects and tests read them, but markets change. We commit to the method and report the expected and realised effect honestly.
Enough sales and spend to detect a difference. We check this in the audit and recommend the tests and models that the data can support.
Not usually. We work with the tools you have and add tests and models where they are needed, and recommend changes only where the data requires them.
Monthly, or as the data changes. Models are refreshed so that budget decisions use current evidence.
It tests and models every channel in the mix, and its results return to each channel as budget and bid decisions.
It can be, if the budget is spread across channels and the tests are designed to fit. We recommend the simplest method that answers the budget question.
From the blog