Consumer & Omnichannel · Growth problem
Digital investment can't be justified against total business revenue
Finance teams reasonably want to know what digital marketing spend is actually buying, but platform-reported return on ad spend answers a narrower question than the one being asked. It measures the online transactions a platform can claim, not the calls it generated, the store visits it influenced, or the brand search it created. Without a way to see the total effect, digital investment gets judged, in both directions, on incomplete evidence.
Symptoms
What this usually looks like
- Marketing is asked to justify budget purely on platform-reported ROAS
- Store leadership is sceptical that digital spend benefits their revenue
- No holdout or geo testing exists to separate incremental effect from correlation
- Halo effects on brand search and direct traffic aren't factored into channel evaluation
- Budget cuts to digital are proposed based on online-only performance during a downturn
- Assisted and call centre revenue isn't included in any marketing performance conversation
Diagnostic questions
What we would test first
- Design and run a geo holdout test on a defined channel or campaign over a set period
- Build a simple model comparing marketing spend trend against store traffic, call volume and brand search trend
- Review current marginal return by channel and spend level using available platform and finance data
- Interview finance on what evidence would change their view of digital's contribution
- Assess what data would be needed to build a credible total contribution model, and what's realistically available now
- Review any historical spend cuts or increases for a natural experiment that can be analysed retrospectively
Root causes
Why it happens
- 01
Measurement infrastructure only covers the online transaction
Standard platform and analytics reporting is built to track what happens within the digital funnel, so any effect that surfaces as a phone call, a store visit, or a delayed direct-traffic purchase is invisible by default, not because it doesn't exist.
- 02
Incrementality isn't tested, only correlated
Most performance reporting shows spend alongside revenue moving together over time, which is suggestive but not conclusive; without geo holdouts or similar experimental design, it's not possible to state confidently how much of that revenue was incremental.
- 03
Halo effects are assumed rather than quantified
Teams often believe brand campaigns support broader demand but rarely measure it, which leaves the belief vulnerable to being dismissed by finance during budget reviews when hard evidence isn't available to defend it.
- 04
Budget conversations happen in channel silos
Marketing, retail and finance frequently discuss digital investment without a shared model of total contribution, so the same evidence gap gets argued from both directions — marketing overclaiming influence, finance underclaiming it — without resolution.
Evidence
The numbers we would look at
These are the metrics that make the constraint visible, and the cuts that stop them being reassuring by accident.
| Metric | What it tells you | How we read it |
|---|---|---|
| Total revenue contribution per pound of marketing spend | A blended measure including estimated assisted, halo and store-influenced revenue, not just direct online attribution.Depends on the quality of underlying influence measurement; treat early estimates as directional, refine over time. | Depends on the quality of underlying influence measurement; treat early estimates as directional, refine over time. |
| Geo holdout test results by campaign or channel | The most defensible way to isolate incremental effect from correlation.Requires genuinely comparable control regions or stores; poorly matched controls will produce misleading conclusions. | Requires genuinely comparable control regions or stores; poorly matched controls will produce misleading conclusions. |
| Store traffic and call volume trend vs. marketing spend trend | A directional signal of assisted influence even before formal testing exists.Correlation only; useful for prioritising where to test, not for making final budget decisions. | Correlation only; useful for prioritising where to test, not for making final budget decisions. |
| Brand search volume trend relative to campaign activity | An indicator of halo and demand-generation effects beyond direct response.Search volume is noisy and affected by seasonality and competitor activity; interpret cautiously. | Search volume is noisy and affected by seasonality and competitor activity; interpret cautiously. |
| Marginal return by channel and spend level | Whether additional spend on a given channel is still producing proportionate incremental return.More useful for budget allocation decisions than average ROAS, which can mask steep diminishing returns. | More useful for budget allocation decisions than average ROAS, which can mask steep diminishing returns. |
Measurement traps
What can mislead you
- Looks finePlatform-reported ROAS has stayed strong, so budget should keep increasing
- Platform ROAS typically degrades in marginal value well before it shows in the platform's own reporting, since platforms are structurally biased towards claiming credit for conversions they can see, including ones that would have happened anyway.
- Looks fineCutting digital spend during a downturn didn't reduce online sales much
- A modest online sales effect from a spend cut doesn't rule out a larger, delayed effect on store visits, calls or brand search, which typically take longer to show up and are rarely tracked with the same rigour.
- Looks fineStore leadership says digital doesn't help store sales
- Anecdotal scepticism from store teams is a reasonable prompt to investigate, but it isn't evidence either way; the only way to settle it is a proper geo or holdout test, not opinion from either side.
Outcome
What better looks like
Not a promised number. A clearer basis for the next investment decision.
- Budget decisions reference incrementality evidence, not platform-reported ROAS alone
- Marketing, retail and finance share a common model of total contribution before debating allocation
- Halo and assisted effects are estimated with a stated method and confidence level, not asserted
- Geo testing becomes a routine part of evaluating new campaigns, not a one-off exercise
Where a Growth Diagnostic would start
A three to four week senior review across demand, discovery, acquisition, conversion, retention, measurement and capability — sequenced so this problem is either confirmed as the constraint or ruled out early. Read alongside the consumer & omnichannel model page for how we frame the wider system.
Related problems
Consumer & Omnichannel
Digital's influence on store sales isn't being measured
Search, store locator and calls are shaping footfall and phone enquiries, but digital only gets credited for sales it closes directly.
Consumer & Omnichannel
Growth has plateaued once single-channel gains are exhausted
Each channel has been optimised individually, but total growth has flattened because nobody owns the space between them.
Consumer & Omnichannel
Store and ecommerce are competing instead of compounding
Separate P&Ls, KPIs and stock pools put store and ecommerce teams in competition for the same customer.
Questions about this problem
- Do we need to increase marketing spend to test incrementality?
- No, a geo holdout test can be designed within existing budget by reallocating or pausing spend in selected regions rather than adding new spend overall.
- How long does a geo test need to run to be credible?
- It depends on typical purchase cycle length and regional demand variability, but most retail geo tests need at least several weeks to a full quarter to produce a reliable read.
- Will this settle the debate between marketing and finance permanently?
- It should materially improve the quality of the debate by replacing assumption with evidence, though ongoing testing is usually needed as channels and campaigns change.