Advertising Identifiers: What Marketers Get Wrong About Identity Resolution

Advertising identifiers are the technical signals, cookies, device IDs, and hashed data points that allow marketers to recognise users across sessions, platforms, and channels. They sit at the centre of targeting, attribution, frequency capping, and personalisation. Without them, most of what digital advertising claims to do becomes significantly harder to prove and considerably harder to execute.

The problem is not that marketers use them. The problem is that most marketers have built entire growth strategies on top of identifier infrastructure they do not fully understand, and they are now discovering how fragile that foundation is.

What Are Advertising Identifiers and Why Do They Matter?

At their simplest, advertising identifiers are persistent labels attached to a device, browser session, or user profile. Third-party cookies have been the dominant form in web advertising for decades. Mobile equivalents include Apple’s IDFA (Identifier for Advertisers) and Google’s GAID (Google Advertising ID). These identifiers allow ad platforms to build profiles, serve targeted ads, cap frequency, and attribute conversions to specific exposures.

The commercial logic is straightforward. If you can identify that the same person saw your ad on Tuesday and converted on Thursday, you can make an argument about causation. You can also avoid showing the same person the same ad seventeen times in a week, which is both wasteful and corrosive to brand perception. Identifiers make these things possible, at least in theory.

In practice, the picture has always been messier. Cross-device matching has never been fully reliable. Consent rates have varied wildly by market and implementation. And the attribution models built on top of identifier data have often been more self-serving than accurate. I spent years managing hundreds of millions in ad spend across multiple industries, and the honest truth is that the confidence people placed in identifier-based reporting was rarely matched by the rigour of the underlying methodology.

If you want to understand where advertising identifiers sit in a broader commercial growth context, the Go-To-Market and Growth Strategy hub covers the full strategic picture, including how measurement, audience development, and channel selection interact at a business level.

Why Identifier Infrastructure Is Breaking Down

The deprecation of third-party cookies in Chrome has been delayed repeatedly, but the direction of travel is not in doubt. Safari and Firefox already block them by default. Apple’s App Tracking Transparency framework, introduced in 2021, requires explicit opt-in for IDFA use, and opt-in rates in most markets settled well below 50 percent. The regulatory environment, shaped by GDPR in Europe and a growing number of state-level privacy laws in the US, has added legal complexity to what was already a technical problem.

The result is a fragmented identity landscape. Some users are trackable with reasonable fidelity. Others are essentially invisible to standard identifier-based systems. And the proportion of invisible users is growing, not shrinking.

This creates a specific kind of measurement distortion. If your attribution model can only see a subset of the users who were exposed to your advertising, the conversions it credits to paid channels are systematically biased toward people who consented to tracking. That is not a random sample. It tends to skew toward certain demographics, certain device types, and certain behavioural profiles. The aggregate numbers look plausible. The underlying reality is more complicated.

I have seen this play out in practice. When I was running agency operations and we started pressure-testing attribution models against incrementality tests, the gaps were often uncomfortable. Channels that looked like strong performers in last-click or even data-driven attribution sometimes showed much weaker incremental lift when you isolated the effect properly. The identifier-based data was not lying exactly, but it was telling a selective version of the truth.

The Walled Garden Problem

The major platforms, Google, Meta, Amazon, and Apple, have responded to identifier deprecation by consolidating measurement inside their own ecosystems. Conversion API integrations, privacy-preserving attribution APIs, and on-platform reporting tools all share a common characteristic: they are controlled by the same companies that sell the advertising. The conflict of interest is structural, not incidental.

This is not a conspiracy. It is just incentive alignment working exactly as you would expect. Platforms have every reason to build measurement systems that show their advertising in a favourable light. And because the alternative, independent cross-platform measurement, is technically harder in a cookieless environment, many advertisers have drifted toward accepting platform-reported metrics as ground truth.

The problem with that approach is not just commercial. It shapes strategy. If your understanding of what is working is filtered through platform-controlled measurement, your decisions about where to invest, which audiences to target, and which channels to scale will be shaped by data that has a built-in bias. That is a significant strategic risk, particularly for businesses operating across multiple channels where cross-platform attribution matters.

Forrester’s work on go-to-market complexity in regulated industries illustrates how measurement gaps compound when you add platform fragmentation to an already complicated channel environment. The identifier problem does not exist in isolation. It interacts with everything else in your go-to-market architecture.

What First-Party Data Actually Solves (and What It Does Not)

The industry consensus response to identifier deprecation has been “invest in first-party data.” That is broadly correct but often oversimplified in ways that lead to poor investment decisions.

First-party data, the information users directly share with you through account creation, purchase history, email sign-ups, and on-site behaviour, does give you a durable identity signal that does not depend on third-party cookies. If someone logs in to your platform or provides an email address, you have a persistent identifier that is yours to use within the bounds of consent. That is genuinely valuable.

What it does not solve is reach. Your first-party data is, by definition, limited to people who have already engaged with your brand. It is excellent for retention, personalisation, and loyalty marketing. It is a weak tool for reaching the audiences you have never spoken to before, which is where most business growth actually comes from.

Earlier in my career, I overvalued lower-funnel performance signals. The numbers looked clean, the attribution was tidy, and the CPAs were defensible in a board presentation. What I was slower to recognise was how much of that performance was capturing demand that already existed rather than creating new demand. Investing in first-party data strategies without also investing in broad-reach brand activity is a version of the same mistake. You are optimising the bottom of a funnel that is not being adequately filled at the top.

The market penetration research from Semrush reinforces a point that Byron Sharp’s work has made for years: growth comes primarily from reaching light buyers and non-buyers, not from extracting more value from your existing customer base. First-party data is not a growth strategy on its own. It is one component of a broader identity and measurement architecture.

Identity Resolution: The Technical Reality

Identity resolution is the process of connecting disparate data points, a cookie here, a device ID there, an email address somewhere else, into a unified view of a person. It sounds straightforward. In practice, it is one of the harder problems in data engineering, and the commercial solutions vary enormously in quality.

Deterministic matching, where you link records based on a shared definitive identifier like an email address, is accurate but requires users to authenticate. Probabilistic matching, where you infer connections based on shared signals like IP address, device type, and behavioural patterns, has wider coverage but introduces error rates that most vendors are not transparent about.

The identity resolution vendor market has grown significantly as cookies have declined, and the quality of what is being sold ranges from genuinely useful to essentially speculative. I have sat through enough vendor pitches to know that match rate claims deserve scepticism. A high match rate is meaningless if the underlying matches are inaccurate. The relevant question is not how many records you can connect, but how confident you should be in each connection, and what the downstream consequences of incorrect matches are for your targeting and measurement.

For most mid-market businesses, building sophisticated identity resolution infrastructure in-house is not realistic. The engineering complexity is significant, and the data volumes required to make probabilistic matching work at acceptable accuracy levels are substantial. This is an area where the gap between what large platform-native businesses can do and what everyone else can do is genuinely wide.

Privacy Sandbox and the Alternatives

Google’s Privacy Sandbox initiative proposed a set of browser-level APIs designed to enable some of the targeting and measurement functions of third-party cookies without exposing individual-level data. The most discussed of these is the Topics API, which assigns users to broad interest categories based on browsing history, without sharing that history with advertisers directly.

The reception from the advertising industry has been mixed, and from privacy advocates, largely critical. The practical concern for marketers is that Privacy Sandbox APIs offer significantly less targeting precision than third-party cookies did. The privacy concern is that even coarse-grained interest signals can be used to infer sensitive characteristics about users when combined with other data.

Google’s decision to retain third-party cookies in Chrome while offering users a choice mechanism, rather than deprecating them outright, has added ambiguity to an already complicated transition. The direction of travel is still toward less identifier-based targeting. The timeline and the specific technical mechanisms remain uncertain.

Other approaches gaining traction include contextual targeting, which does not require user-level identifiers at all, clean room technologies that allow data collaboration without raw data sharing, and cohort-based modelling that works at an aggregate level rather than individually. None of these is a perfect substitute for what third-party cookies enabled. Each involves different trade-offs between reach, precision, and privacy.

BCG’s work on scaling agile capabilities is worth reading in this context, not because it is directly about advertising identifiers, but because the organisational challenge of adapting to a changed technical environment is fundamentally an agility problem. Businesses that have built rigid processes around specific identifier types will find the transition harder than those that have maintained flexibility in their measurement and targeting architecture.

What Good Measurement Looks Like Without Perfect Identifiers

The honest answer is that good measurement has never required perfect identifiers. It has required honest approximation, triangulation across multiple signals, and a willingness to hold uncertainty without either ignoring it or being paralysed by it.

When I was judging the Effie Awards, the campaigns that impressed me most were not the ones with the cleanest attribution stories. They were the ones where the teams understood what they could and could not know, made reasonable assumptions, and designed measurement frameworks that were honest about their limitations. That discipline is harder to maintain when you have a dashboard full of apparently precise numbers, but it is exactly what the current identifier environment demands.

Media mix modelling, which uses statistical regression to estimate the contribution of different channels to business outcomes, is experiencing a significant revival. It does not require individual-level identifier data. It works at an aggregate level and can account for channels that are difficult to track directly, including offline media, organic search, and brand-driven direct traffic. It is not perfect, and it requires meaningful data volumes to produce reliable outputs, but it is a more honest framework than last-click attribution built on incomplete identifier data.

Incrementality testing, where you deliberately withhold advertising from a holdout group and measure the difference in outcomes, is the closest thing to a ground truth measurement in digital advertising. It is also underused, partly because it requires accepting short-term inefficiency in the service of long-term measurement accuracy. Most businesses find that uncomfortable. The ones that do it consistently tend to make better investment decisions.

Tools like Hotjar’s feedback and behaviour analytics offer a different kind of signal, qualitative and behavioural rather than identifier-based, that can complement quantitative attribution models. Understanding why users behave as they do on your site is a different question from tracking which ad they saw before arriving. Both questions matter. The identifier conversation tends to focus entirely on the second.

Vidyard’s research on pipeline and revenue potential for go-to-market teams points to a broader issue: the gap between what teams think they know about buyer behaviour and what is actually happening. Identifier-based measurement has given marketers a false sense of visibility. The transition to a less identifier-dependent environment is uncomfortable partly because it reveals how much of that visibility was illusory.

The Strategic Implication for Go-To-Market Planning

The advertising identifier transition is not primarily a technical problem. It is a strategic one. Businesses that have built their go-to-market models around hyper-targeted, identifier-driven performance marketing are facing a structural challenge that cannot be solved by finding a better technical substitute for the third-party cookie.

The businesses that will handle this most effectively are the ones that rebalance their investment toward brand and reach, invest in genuine first-party data infrastructure rather than treating it as a checkbox, adopt measurement frameworks that are honest about uncertainty, and resist the temptation to accept platform-reported metrics as an adequate substitute for independent measurement.

That is not a comfortable set of recommendations for teams that have spent years optimising toward clean, attributable performance metrics. But the alternative, continuing to optimise a measurement system that is increasingly disconnected from commercial reality, is worse.

Creator-led content and community-based distribution, areas explored in Later’s go-to-market with creators resource, represent one direction that does not depend on identifier infrastructure in the same way that programmatic targeting does. Reach built through trusted voices and organic distribution is harder to measure precisely, but it is also more durable and less exposed to the identifier deprecation risk.

The broader principles that connect advertising identifiers to go-to-market strategy, how you reach new audiences, how you measure what matters, and how you build durable commercial advantage, are covered in depth across the Go-To-Market and Growth Strategy hub. If you are rethinking your measurement architecture, that is the right context to do it in.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions