By Allan Tinkler, Commercial Director, Anonymised
The year 2025 has brought back onto the agenda the legacy systems of measurement and decreasing usefulness under the weight of accelerating signal loss. For too long, marketers have relied on what has become an increasingly flawed and misleading model: the idea that we could connect each impression to a conversion. Overambitious at the best of times, this approach still carries the weight of supporting performance measurement across the industry and it is now starting to creak. Some rebalancing is necessary.
The causes of this renewed focus on measurement are not new. The removal of third-party cookies, once the backbone of cross-site tracking, is now compounded by the growth of IPv6, the new IP address standard that can no longer sustain alternative ID graphs the way the old IPv4 could. On top of that, increased channel and identity fragmentation (with each DSP provider picking its own identifier) is leaving a gaping hole in our ability to follow the user journey.
Lower funnel measurement is being hijacked
In this environment, relying on the traditional last-touch attribution (LTA) model is damaging to advertisers. LTA, by its very nature, gives all the credit for a conversion to the final touchpoint, ignoring the complex, multi-channel journey that truly influenced the consumer. This model systematically undervalues channels that drive awareness and consideration, leading to distorted budget allocation. Today more than ever, cross-channel LTA is more likely than ever to attribute a conversion to impressions associated to persistent identifiers – typically the hashed emails that abound in the walled gardens but are rarer than a white rhino in the large prairies of the open web.
The situation is further complicated by the fact that walled gardens keep marking their own homework, unchallenged despite ample evidence that they overestimate the impact of their own activity. Walled gardens claim conversions that may have been influenced by your other marketing efforts, creating siloes where every platform is a winner, and the marketer is left to untangle the mess of overlapping, self-attributed data. The end result is that a large portion of your marketing budget may be spent on actions that would have occurred anyway.
Upper funnel metrics optimise towards vanity
Beyond the structural flaws of LTA, we also face the challenge of prioritising the wrong metrics altogether. Upper funnel campaigns are still using metrics like click-through rate (CTR) or even view-based conversions. These metrics can be easily inflated and provide a false sense of success. A high CTR doesn’t guarantee a sale, and a conversion claimed by LTA might have happened regardless of the last click. The focus has been on “did they see it?” or “did they click it?” instead of the only question that truly matters to the bottom line: “did they buy it?” The inherent bias in these metrics can lead to significant overspending on channels that appear to perform well on paper but fail to drive any real business growth. A famous 2013 Adobe ad encapsulated this problem perfectly, but 12 years later the market is still using CTR as a metric.
This problem is further compounded when we consider the impact on specific audiences, particularly those on Apple devices. Apple’s focus on user privacy, driven by features like Intelligent Tracking Prevention (ITP), has made huge audiences effectively untargetable and unmeasurable for marketers. As tracking becomes more difficult and unreliable, performance campaigns gone extinct on Safari, iOS and other environments. This can cause marketers to deprioritise or completely ignore a valuable and high-spending demographic, a blind spot that reinforces the for better forms of measurement.
Outcome versus incremental value: multi-modal measurement
The opportunity we have is to evolve the measurement system: from complete reliance on single-channel attribution to measuring incremental sales, from chasing clicks to understanding causality. Campaign measurement is not bean-counting, it’s an ongoing investigation into reality, it’s police work on a crime scene. Every good investigation needs multiple clues to find the most plausible answer. Advertisers that care about “outcomes” should ban vanity metrics and stop relying on a single measurement signal. User-base attribution can and should be complemented by signals like brand uplift studies and geo-incrementality tests.
We call this approach “multi-modal measurement”. Multi-modal measurement is the analysis of multiple data points to understand the effectiveness of a campaign. For a branding campaign, a good multi-modal mix can include well-known KPIs like viewability, inventory quality, attention and brand-lift studies. For a performance campaign, advertisers can combine post-click and post-view metrics with geo-incrementality data and easy-to-use attribution frameworks that duplicate self-attributed conversions of the various channels.
Examples of multi-modal measurement combos
| Branding campaign | Performance campaign |
| Viewability (source: DSP) | Post-click engagement (source: analytics) |
| Attention (source: vendor) | Post-view sales (source: DSP or vendor) |
| Inventory quality (source: DSP or vendor) | Geo-incrementality (source: vendor) |
| Brand-lift (source: DSP or vendor) | Attribution (source: vendor) |
The good news is that multi-modal measurement is far less complicated and expensive than it sounds. The data is easily available and relatively inexpensive, and its interpretation can be automated using AI. Critics will say that some of this data will only be available at the end of a campaign, undermining in-flight optimisation efforts. But advertisers are better off understanding what strategies actually work than optimising towards the wrong metrics, the wrong data and the wrong channels.
The sooner we stop treating campaign measurement as bean-counting and we look at it as detective work, the sooner we will start making informed marketing decisions.
About the Author
Allan Tinkler is Commercial Director at Anonymised. For more information visit: anonymised.io
