Braze says marketing attribution can't see the moments that actually persuade buyers

The moment that actually sells a product may be a message in a group chat that no marketing platform will ever log. That is the premise of an analysis Braze published on January 27, 2026, written by Sally Wills, the company's senior content strategy manager, on why attribution is getting harder to trust.

Braze's argument starts with a definition. Attribution assigns credit to the touchpoints that influence a customer action, and it worked well enough when there were few channels and the path to a purchase was short. Braze says the modern journey looks more like a pinball machine than a funnel: unpredictable, spread across platforms, and stretched over days.

Its illustrative path runs through a blog post read on a phone, a product mention in a group chat, a social ad, a desktop browsing session, an app download and an exit. Days later the same person returns via search after an offline conversation and buys after a reminder message. Attribution tries to draw a straight line through all of that, Braze writes, while the most persuasive steps happen in private channels or the physical world and never become measurable signals.

For the marketing operations lead who has to defend a media budget to a CFO, that is an uncomfortable premise. The report that hands credit to "direct" or "search" may be recording the last visible step, not the cause. Braze's diagnosis rests on three factors: fragmented journeys, privacy that reduces what can be observed, and identity that breaks when a customer changes devices.

Two kinds of data gap, and only one of them is fixable in-house

Braze separates the missing signals into categories. Some gaps are expected: customers switch devices, privacy choices limit tracking, and store visits or word of mouth never produce a clean record. Conversions can also arrive late or aggregated in platform reporting, so the outcome lands days after the touchpoints that shaped it.

Other gaps are about setup. If customer identity and events are not consistent across web, app and messaging channels, Braze says a journey cannot be stitched together even when the data exists. Its example is a customer who clicks an ad on mobile, browses in the app and purchases on desktop; without strong first-party IDs and reliable event instrumentation, that purchase gets filed under "direct" or arrives stripped of context.

That second category is the one an operator controls. For teams whose web and app instrumentation grew up in separate projects with separate owners, the size of the "direct" bucket in the attribution report is arguably a diagnostic of identity plumbing rather than a finding about customer behavior. Braze does not put a number on how much of a typical report falls there, and nothing in its analysis supports one.

Every platform keeps its own score

Braze's next point is organizational. Each platform has its own definitions and its own incentives, one team optimizes for clicks, another for installs, another for revenue, and the resulting attribution reports reflect the org chart as much as the customer.

The consequence Braze draws is about time horizon. Most attribution reporting exists to explain one moment, the conversion. Retention and lifetime value build over weeks or months through a welcome series, an in-app message, a loyalty update, or a support interaction, and Braze says traditional attribution rarely connects those post-conversion touches to long-term outcomes.

So teams end up optimizing for whatever wins the first purchase rather than what keeps the customer. Braze also flags overconfidence in simplified attribution outputs as a challenge in its own right. For a subscription or app business where the second and third purchases carry the margin, this is the part of the analysis that matters most; for a one-time-purchase category, the conversion-moment bias costs less.

What Braze proposes instead: first-party IDs, lift tests, lifecycle analytics

Braze is careful not to declare attribution dead. Models still help, it says, but they are not a complete account of what worked. The practical shift it describes runs toward first-party data, journey-level measurement and experimentation designed to prove lift. According to Braze, measurement is also shifting toward analytics that focus on customer lifecycle value (CLV) and outcome-driven decisioning with the help of AI.

The analysis is a useful reframing of the buying question. The sharper thing to ask any measurement vendor is whether it can show lift against a holdout, not whether its model can distribute credit more elegantly than the last one.

The academic literature arrived at the same place a decade ago. A September 2016 special section of the International Journal of Research in Marketing, introduced by P.K. Kannan, Werner Reinartz and Peter C. Verhoef, described credit assignment across media, channels and devices as an important problem and argued attribution models could guide spending if they focused on the incremental value of a touchpoint and spillover effects between channels.

Read together, the two documents suggest what has changed is the reason for the difficulty rather than the difficulty itself. The 2016 framing was about channel count and cross-device paths. Braze's 2026 account adds privacy limits and identity breakage that remove signals the earlier models assumed were available, which would explain why the incrementality approach the academics proposed is now being pitched as the operational fallback.

More spend, more AI promises, more channels outside the dashboards

The measurement problem is landing on a growing budget. Marketing Dive reported on January 29, 2026 that global advertising revenue in 2025 exceeded initial projections from firms including WPP Media, and that the momentum is expected to carry into 2026 even as marketers work through a tense economic climate and accelerating AI adoption.

Two other threads in Marketing Dive's coverage bear directly on Braze's argument. Chris Kelly reported on January 20 that promises of one-stop, AI-powered advertising solutions are proliferating and that marketers will need to separate what is real from what is hype. Peter Adams reported on January 27 that retailers are experimenting with channels such as Substack as they pursue human-powered brand building.

Adding channels such as Substack to the mix would, on Braze's logic, make the attributable share of a journey smaller still, even as AI-driven tools promise to automate the allocation decision. For a brand whose 2026 plan leans on those channels, the honest expectation is that the dashboard explains less of the result, not more.

None of this is measured yet. Braze offers a mechanism and a direction, not a benchmark, and Marketing Dive's outlook is a forecast rather than a result. The first place the gap becomes visible is the reporting cycle where every channel dashboard claims credit for the same purchase and total revenue does not move to match.