One platform can attribute a purchase to its campaign while another claims the same operation. Both readings can comply with their rules without two purchases existing. The economic problem appears when those credits justify expanded investment: distributing an observed outcome and determining the outcome advertising added are different questions.
Attribution remains useful for organizing routes and messages. Incremental impact requires a comparison approximating what would have happened without the intervention. Changing models can move credit between columns without changing a sale. Recognizing the difference protects budget and retains an unanswered causal question without dismissing all available information.
Market size does not establish a channel’s return
The Statista series based on IAB Spain and PwC records Spanish 2025 expenditure of EUR 1,992.6 million on search and EUR 1,832.5 million on social media. Display excluding social and branded content received EUR 1,225.9 million. The original report combines expenditure reported by a 78-company panel with estimates of the remaining market. These are market amounts rather than effectiveness or caused sales.
Spain 2025: three substantial digital advertising spending categories
Market expenditure estimates in nominal million euros. Results for 2025, published in the 2026 IAB Spain / PwC study.
million EUR
Tap or hover for details. All values are also available in the data table.
Collapse notes
IAB Spain / PwC / Statista. Original p. 13 confirms values; p. 4–8 explains a 78-company panel, reported data, calibration and econometric estimates of uncontrolled expenditure. Display includes video and excludes social and branded content; YouTube is counted in video rather than social. This is neither return, outcome share nor a 2026 forecast. Author report PDF hosted on a public mirror.
Explore the data table
| Category | Investment, million EUR |
|---|---|
| Search | 1,992.6 million EUR |
| Social media | 1,832.5 million EUR |
| Display excluding social and branded content | 1,225.9 million EUR |
The comparison places the resources needing a defined justification in context. Higher national search spending does not indicate its contribution to a particular offer. A buyer already searching for a brand might complete through another route without an ad. Discovery messaging can accompany a longer decision. Reporting needs to establish the scope of each situation rather than replace it with a spending ranking.
The commercial total remains separate from credits
The reference for how many operations occurred is the commercial record of unique valid operations, recognizing cancellations and adjustments. Each attribution report then states its rule, window, observed events and unit. An impression, a click and a final visit can receive credit for one purchase in different tools. Adding them does not create an additional total.
Wider windows or changed collection coverage can increase credit without increased commercial activity. The change retains a date and version when periods are compared. A missing link does not establish absent influence either: an unobserved route remains unknown rather than completing the journey with a personal story the data never supplied.
Choosing the highest report does not resolve incomparable data
In 2025, Nielsen asked 1,400 brand professionals with annual marketing budgets of at least one million dollars about their first or second obstacle to measuring digital return. Incomparable data received 19%, and insufficient granularity 18%. Statista records these priorities. The question measures neither how much return is overestimated nor which method would resolve each difficulty.
Nielsen 2025: four difficulties interpreting digital return
1,400 brand marketers at manager level or above, minimum marketing budget USD 1 million. Online survey February 25–March 6, 2025.
Tap or hover for details. All values are also available in the data table.
Collapse notes
Nielsen / Statista. 350 respondents per APAC, EMEA, North America and Latin America region. Original p. 22 and 29 confirms question, values and eligibility. IAB UK hosts the document rather than a UK-only study. Stated priorities, not every recognized problem or an assessment of causal attribution. Smaller budgets are not represented.
Explore the data table
| Category | First or second obstacle |
|---|---|
| Incomparable data | 19.0% |
| Insufficient granularity | 18.0% |
| Cannot measure newer channels | 17.0% |
| Inadequate measurement tools | 17.0% |
The operational consequence is preserving definitions before reconciliation. A matrix of operations appearing across sources can detect duplicate counting when linking is reliable and proportionate. It does not calculate each channel’s causal contribution. More granularity describes some facts better without itself removing interacting customers’ prior intent.
The additional-outcome question needs a bounded intervention
The Magenta Book on impact evaluation distinguishes outcome monitoring from evaluating changes caused by intervention using the counterfactual. Its setting is public policy. Applying it to advertising requires specifying offer, population, period and outcome before interpreting a comparison; the guide does not automatically validate a commercial report.
Comparing clickers with nonclickers mixes influence with selection: the first group might already have stronger purchase intent. Comparing before and after mixes intervention with calendar, price, availability and product changes. A descriptive difference remains useful, but the verb is changed afterward rather than changed because of the campaign without additional evidence.
Random allocation can help build a comparison when a design is suitable and feasible. It requires a unit, outcome and preventing conditions from mixing. Allocation and actual exposure remain distinct: later selecting only responders reintroduces selection. With existing groups, unobserved differences remain an assumption open to scrutiny.
A useful estimate retains magnitude, uncertainty and cost
An estimated incremental outcome states unit, population, window and assumptions alongside uncertainty. Imprecision does not establish no effect, and a positive estimate does not guarantee sufficient economic contribution. The decision compares plausible change with investment and delivery capacity, keeping unknown what the design does not identify.
Periods need comparable maturity: recent windows contain pending operations already consolidated in closed periods. Expanded investment can change audience and frequency, so an observed effect does not extrapolate indefinitely. Final reporting preserves three layers: operation totals, credit distribution and additional-impact evidence. Investment merits expansion when the layer answering its economic question is strong enough, rather than when more tools claim the same sale.
In this article
Marian Vasile
Marian Vasile is a MarTech and growth professional. His work spans product marketing, SEO, paid advertising and deep online tracking, including behavioral analytics and attribution.