For a while, digital marketing seemed to promise perfect measurement: every click tracked, every customer journey visible, every conversion neatly assigned to the ad that caused it. That promise was always overstated, and changes in privacy law, browser technology, and platform policy have made the gaps impossible to ignore.
That is not a reason for despair. Businesses made good marketing decisions long before user-level tracking existed. The task now is to understand what modern attribution can and cannot tell you, and to combine several imperfect sources of evidence into confident decisions.
What attribution is trying to answer
Attribution assigns credit for a conversion to the marketing touchpoints that preceded it. Different models distribute that credit differently: last-click gives everything to the final interaction, while data-driven models use observed patterns to share credit across several touchpoints.
The deeper question, however, is not "which ad was involved?" but "what would have happened without this marketing?" That is a question about incrementality, and no attribution model answers it fully on its own, because attribution only observes the people who converted — not what they would have done otherwise.
Why the data has become less complete
Several developments have reduced the share of the customer journey that can be observed directly:
- Consent requirements. Privacy laws in many jurisdictions, including the GDPR in Europe and privacy legislation in Canada and several US states, require or encourage consent before non-essential tracking. When visitors decline, their journeys are only partially visible.
- Browser and device restrictions. Several browsers limit third-party cookies and shorten the lifespan of some first-party storage, and mobile operating systems require apps to ask permission before tracking across other companies' apps and websites.
- Cross-device behaviour. People research on a phone and buy on a laptop, or see an ad on one device and search on another.
- Offline steps. Phone calls, sales meetings, and in-person purchases happen outside the browser entirely.
How platforms fill the gaps
Advertising platforms increasingly use statistical modelling to estimate conversions they cannot observe. Features such as consent-aware tag behaviour, server-side event integrations, and privacy-preserving matching of first-party data aim to recover some lost signal while respecting user choices.
These tools can be valuable, but it is important to understand that modelled conversions are estimates. They are generally designed to improve optimization and reporting in aggregate rather than to describe individual journeys precisely. Treat them as directional evidence, and make sure any data you share with platforms is collected and used lawfully, in line with your privacy notice and the consent you have obtained.
Build a measurement foundation you control
The more you rely on third parties to define your results, the more exposed you are to their changes. Strengthen the data you own:
- Consistent conversion definitions across your analytics tool and every ad platform.
- Clean, consent-aware tagging with documented events and no duplication.
- Campaign tagging discipline, using consistent UTM parameters so traffic sources are recorded reliably.
- CRM integration, so leads can be traced to qualification and revenue.
- Simple, durable signals such as asking customers how they heard about you — imperfect, but independent of tracking technology.
Triangulate with several methods
Rather than searching for one perfect model, combine methods that have different strengths:
- Platform and analytics attribution for fast, granular, day-to-day optimization.
- Blended efficiency metrics, such as total revenue relative to total marketing spend, which do not depend on assigning credit at all.
- Incrementality experiments, such as geographic holdout tests or platform lift studies, which compare outcomes with and without advertising to estimate true causal impact.
- Marketing mix modelling for larger budgets, which uses aggregate historical data to estimate the contribution of each channel while accounting for seasonality and other factors.
When these sources broadly agree, you can act with confidence. When they disagree, the disagreement itself is valuable: it tells you where to investigate or test next.
Interpret channel reports by role
Attribution consistently favours channels that appear near the end of the journey, such as brand search and retargeting, because they capture people who are already close to converting. Channels that create demand — video, prospecting social, and display — tend to look weaker in last-touch views even when they are doing important work.
Judge each channel by evidence appropriate to its role. For demand capture, attributed conversions are informative. For demand creation, look at incrementality tests, changes in branded search, and blended outcomes over time.
Making decisions with imperfect data
Perfect measurement is not coming back, and it never fully existed. Good decisions come from accepting uncertainty explicitly:
- State how confident you are in each conclusion and why.
- Make larger changes when several sources of evidence agree; test first when they do not.
- Review blended business results regularly to catch problems attribution may hide.
- Respect privacy choices — they are both a legal requirement and a matter of customer trust.
The goal of attribution isn't certainty. It's making better decisions than you would have made without it.
A Dune marketing audit applies these principles to your accounts, tracking, and funnel. See how the audit works or get in touch.
This article is general information, not advice for your specific situation. Platform features and policies change; confirm current details in each platform's official documentation.
