Omnichannel performance measurement: why your dashboards are making you allocate budget backwards

Advertising investment is shifting toward digital while, in many sectors, final conversion still happens in person. Unsurprisingly, performance measurement struggles to keep up… A tour of the blind spots weighing on media allocation, and the solutions to address them.

Digital captures the budgets, the store rings up the sales Dentsu’s forecasts for 2026 put global advertising investment above $1 trillion, of which 68.7% goes to digital (Magna and GroupM arrive at similar orders of magnitude). Growth is concentrated in retail media (+14.1%), online video (+11.5%), and social (+11.4%), without traditional channels collapsing for all that. Purchasing behavior, however, doesn’t follow the same geography. McKinsey observes that more than half of B2C customers use between three and five channels for a single purchase or service interaction. In retail, the vast majority of journeys start with an online search and often conclude in a store, over the phone, or at an agency. This gap creates a piloting problem: budgets move toward the environments that produce the most data, while part of the value materializes in systems the platforms can’t see. The result is that measurement has been built around whatever was easy to instrument.

Three blind spots weighing on allocation decisions

1. Last-click favors the end of the journey For about fifteen years, budget optimization has relied on the last touchpoint before conversion. The shortcut is still widespread, and its bias always leans the same way: it credits the channels closest to purchase (search, retargeting, transactional email) and leaves in the dark those that create demand upstream.
2. Identifier loss undermines journey reconstruction Since the launch of App Tracking Transparency in 2021, advertising tracking opt-in rates have sat around 15 to 25% globally. Meta has publicly put the impact on its 2022 revenue at around $10 billion. On the advertiser side, industry estimates put the share of acquisition cost affected by this signal degradation at between 20% and 50%.
3. Offline conversion doesn’t show up in native reports An in-store sale, an appointment booked by phone, a contract signed at an agency: these events exist in the point-of-sale system or the CRM, rarely in advertising interfaces. Campaigns that contribute to them show only partial performance, with nothing flagging the gap.

Blind spot	Effect on allocation
Last-click model	Over-weighting of conversion channels, under-funding of demand-generation channels
Identifier loss	Fragmented journeys, inflated reported acquisition cost, degraded targeting
Missing offline conversion	ROAS understated on campaigns that contribute to in-person sales

A layered measurement approach rather than a single model The opposition between multi-touch attribution (MTA) and Marketing Mix Modeling (MMM) has dominated the debate for years, even though the two approaches answer different questions. MTA examines sequences of touchpoints at the individual level: which sequence preceded this conversion? MMM works on aggregated series of spend and sales: what is the marginal return of each channel at the current spend level, offline channels included? The former requires user identifiers; the latter doesn’t, which makes it more resilient to tracking restrictions. A third instrument has taken hold among the most advanced organizations: incrementality testing. Using geo-tests, control groups, and geographic A/B tests, it establishes causality where the other two produce contribution estimates.

The framework emerging today combines these approaches, assigning each a distinct role: incrementality as the causal reference, MMM as the portfolio-level decision engine, attribution as the tactical signal. Each runs on its own cadence, producing nested feedback loops — quarterly for strategic arbitration, weekly for tactical optimization, periodic but regular for causal validation. An organization that only has the tactical loop is flying by short-term instinct; one that relies solely on the strategic loop reacts late to market shifts.

This combination long remained the preserve of the biggest budgets, for cost reasons. The release of industrial-grade open source libraries (Meridian at Google, Robyn at Meta, PyMC-Marketing at PyMC Labs) has now lowered that barrier.

Conversion data: the common prerequisite for every method These loops nonetheless share the same dependency: the completeness of conversion data. A statistical model applied to a signal that’s missing its physical half will reproduce, with more method, the very bias it was meant to correct.

This is where offline Conversion APIs bring their value. Deployed since 2021 by Meta, Google, and TikTok, they follow the same principle: transmitting offline conversion events from the advertiser’s server to the platform’s server. These events come with SHA-256-hashed identifiers, so they can be matched back to the original ad exposure.

The remaining challenge is actually building these events, and that’s where the difficulty lies. Their reliability rests on a technical foundation: a stable first-party identifier over time, identity resolution capable of linking an in-store purchase (via a loyalty card, for example) to a prior digital exposure, and consent management propagated all the way to the transmitted event. It’s on this layer of collection, identity resolution, and consent — upstream of activation — that a platform like Commanders Act positions itself.

The quality of the matching keys matters just as much as the architecture: several keys rather than a single one, plus a transaction identifier to rule out duplicates. A clean signal upstream produces far better results than a sophisticated model downstream.

Where does your own measurement setup stand? Three questions give a sense of the effort needed to build this kind of omnichannel measurement. What share of conversions is completed offline, and what proportion of that currently flows back to activation platforms? Do budget decisions rest on a single model, or on several instruments that cross-check one another? Does consent data travel all the way to the transmitted events, or does it stop at web collection?

These questions (and their answers) are at the heart of our guide “Omnichannel Piloting and Performance.” In it you’ll find a comparative overview of Meta, Google, and TikTok’s Conversion APIs, possible implementations, the five pillars of an infrastructure that reconciles online and offline, and more detail on the opportunity that Meta’s offline CAPI represents.

Invest, Measure, Allocate Your Media Budget in the Omnichannel Era