If sales growth is stuck at 0.0% to 1.0%, the easiest thing to do is stare harder at ROAS. More dashboards. More “optimization.” More arguments about attribution windows.
Albertsons is taking a different step: its retail media business, the Collective, is adding lifetime value (LTV) to its ad platform analytics. That’s not a cosmetic metric swap. It’s a choice about what gets funded when the business can’t count on topline momentum. (And yes, Albertsons’ 2026 guidance included identical sales growth of 0.0%–1.0%.)
Here’s the tension: ROAS is great at telling you what already happened. LTV is an attempt to price what happens next.
Why this matters now: retail media needs a longer memory
Retail media has been expanding because retailers sit on the thing most advertisers are losing: first-party data. A 2026 analysis describes Albertsons’ strategic shift toward digital—especially its retail media network—as a catalyst, using loyalty and digital properties to sell targeted advertising (typically higher-margin than core grocery retail). That business model pushes measurement in a specific direction: toward who a shopper becomes, not just what they bought after an ad.
And Albertsons has the prerequisite ingredient. The same 2026 analysis flags increased loyalty engagement as a major catalyst—exactly the kind of behavioral signal you need for segmentation, churn reduction, and LTV modeling, whether or not the company brands it as an “LTV program.”
Put those together and the move makes sense. If retail media is going to be a real profit center, the retailer has to defend the idea that some conversions are simply worth more than others. Not emotionally. Economically.
Albertsons’ actual point: ROAS is a snapshot, not a relationship
The Collective’s LTV addition is also a plainspoken critique of how retail media has been sold. Liz Roche, VP of media and measurement at Albertsons, told AdExchanger:
“If we only are looking at ROAS, we’re being fairly myopic about how we understand the actual relationship between the retailer and the brand.”
That’s the pattern interrupt. A retailer—who could happily keep reporting short-window returns—chooses to complicate its own story. Why? Because advertisers are asking for what Roche called
“longitudinal metrics”that quantify value over time, instead of campaign performance as a quick “snapshot.”
There’s another implication hiding in the quote: when LTV becomes the shared language, the conversation shifts from “did this campaign work?” to “did this campaign acquire the right kind of customer?” That’s a much harder question. It’s also the one that actually changes budget allocation.
The hard part: LTV is modeling, not counting
LTV sounds clean until it hits an ops team’s reality. Roche put it directly: building LTV requires large graphs of historical, person-level data that can model future behavior. That’s identity resolution, cohorting, and time-series behavior—plus the political work of agreeing on definitions.
Albertsons is already working in cohorts. The Collective groups shoppers into sets like “lapsed,” “new-to-brand,” “repeat,” and “loyal”. With shopper history going back a year (and many years for some), Roche said the grocer can analyze retention, changes in average order value, and purchase frequency, then see how marketing tactics affect those over time. From there, it can project the value of cohorts over the next year.
That projection piece is where teams get burned. Because LTV is not a receipt. It’s a forecast.
In retail/e-commerce, a common baseline formula is: CLV = Average Order Value × Purchase Frequency × Customer Lifespan. Profit-based CLV comes from multiplying revenue CLV by gross margin percentage. (That margin step matters; finance operators regularly warn that revenue doesn’t pay back CAC—gross profit does.)
So yes, LTV can be directionally useful without being perfect. But it has to be treated like a model with assumptions, not a “truth metric.”
If you only change one thing: add LTV as a holdout readout, not a targeting input
Roche said LTV might eventually be incorporated into ad targeting. Maybe. But trying to buy media directly to modeled LTV too early is how teams create a black box they can’t audit.
Here’s the move that’s more practical for most ops teams (and it maps to what Albertsons is doing by starting in analytics): use LTV as the readout for incrementality experiments, not as the optimization lever inside the platform.
The hypothesis (make it falsifiable): If we run retail media campaigns with cohort-based measurement and a holdout, then the exposed group’s 90-day gross-margin LTV will increase versus holdout because we’re shifting spend toward shoppers more likely to repeat (not just convert once).
Success = lift in gross-margin LTV for the cohort over a fixed window (pick 60–120 days; don’t pretend you can “prove lifetime” in a quarter). Guardrails = same-period revenue and purchase frequency. Stop-loss = if incremental gross profit doesn’t cover incremental media cost by an agreed payback window.
Why be this strict? Because LTV is famous for telling comforting stories. Roche even called the new LTV metric
“built really for storytelling”—a tool to justify investment. That’s not evil. It’s just a warning label.
Run it this week: the ops setup for an LTV-ready measurement spine
This is the version that a Marketing Ops Pro can actually implement without rewriting the whole data warehouse.
- Setup: Define 4 cohorts that match how The Collective already thinks: new-to-brand, repeat, loyal, lapsed. Pick one product/category where purchase cycles are short enough to observe repeat within 60–90 days.
- Owners: Marketing Ops (definitions + instrumentation), RevOps/Finance (margin assumptions + payback rules), Channel owner (campaign execution).
- Tools: Whatever you already use for cohorting and analysis (warehouse + BI is enough). The requirement is identity consistency, not a new platform.
- Launch: Run a holdout (or geo split) with a fixed budget. Keep creative and offer stable to reduce noise from creative fatigue.
- Readout: Report two numbers: incremental gross profit and incremental cohort LTV over the window. Treat attribution as directional; treat holdout lift as the decision signal.
Trade-off (say it out loud): this will reduce short-term ROAS “wins” before it improves decision quality. Teams used to optimizing to last-click will feel like they’re losing control. They’re not. They’re switching scoreboards.
Albertsons isn’t pretending LTV replaces ROAS overnight. It’s starting where it can: analytics, cohorts, and a longer time horizon. In a year where guidance sits at 0.0%–1.0%, that’s the honest bet—less obsession with the snapshot, more discipline about the relationship.
