Same-Store Sales, But for Loyalty: Why "Active Members" Is the Wrong Headline Number

Every mature loyalty program eventually hits the same wall: enrollment growth slows, the database gets large, and "active members" stops telling you anything useful. Yet it's still the number chains lead with - in earnings calls, in press releases, in the slide a CMO puts in front of the board.

Chipotle Mexican Grill's Q2 2026 earnings call is a good example of why that's a problem, not because the number isn't real, but because it is meaningless on it's own. CEO Scott Boatwright touted 23 million active Chipotle Rewards members - up 9.5% since the program's April relaunch - in the same breath as this: only about 20% of in-restaurant transactions scan for rewards, versus nearly 90% of the chain's own digital orders. (Those are the same numbers as they reported for 1Q.) Twenty-three million is a headline. Twenty percent is the actual state of the business. Most public disclosure only gives you the first number.

To be fair, this isn't a knock on Chipotle's loyalty or data science team - and any VP of Loyalty reading this is probably nodding. They're not blind to any of this internally. They sit on sophisticated, multi-dimensional dashboards: micro-cohorts, LTV, churn risk. The problem isn't a lack of data. It's that when a brand publicly elevates "Active Members" as the headline metric, that framing filters down into internal KPIs whether the data team wants it to or not. Teams start optimizing for the number they're judged on - that might mean running margin-diluting national promos to trigger a wave of low-frequency digital trials, or deploying some other tactic which inflates a loosely-defined "active" count (often a 180-day rolling window — one visit in six months clears the bar) without moving the number that actually matters: whether your core customer is visiting more.

Metrics like "app downloads" and "active members" measure intent, not habit. Chipotle's own numbers make the gap explicit, and it bears repeating: 90% of digital orders are already tied to a loyalty account, but only 20% of in-restaurant transactions are. That's not a small optimization opportunity - it's the majority of a chain's physical traffic sitting in the "unidentified" bucket, invisible to personalization, invisible to churn modeling, invisible to everything except a same-store-sales top line.

Chipotle has recognized the problem, even as they spent the past quarter chasing acquisition instead. Boatwright didn't frame it as a marketing issue - he framed it as friction: "fidgeting with your phone, trying to get your loyalty app up, pay with your credit card... it's really a clunky experience." Their fix, piloting in this third quarter, isn't as simple as just another push notification - it's removing the scan step entirely, so paying automatically earns and as a secondary benefit, can prompt enrollment.

That's the tell. If your fix for low engagement is "make the app do less work at the point of purchase," you've implicitly admitted the app was never really an engagement engine for your most frequent customers - it functioned instead for many customers as a checkout tax for in-person transactions.

For a mature program like Chipotle's, the acquisition phase is over. The job now is identity capture and habit formation, not enrollment growth. Every transaction falls into one of four buckets:

• Unidentified: the anonymous cash/card diner

• New to Program: recently captured, unproven

• Existing – Recently Active: the habituated core

• Existing – Recently Inactive: leaking out the back

Two obvious trend lines fall out of this: the unidentified share of transactions should be shrinking, and the revenue share from recently-active members should be rising. Both are worth reporting quarter over quarter. But on their own, they don't tell you whether identification gains are actually converting into habit, or whether you're just relabeling the same transactions differently. You need a way to see whether the core is deepening, not just whether the label is shifting.

This is the same reason the industry already trusts "comparable sales" over raw revenue: comps blend traffic and average check into one number because either alone is misleading - traffic can rise while check collapses, or vice versa - and executives, analysts, and boards have spent decades learning to trust the blended read. Loyalty needs its own version of that: a number that can't be quietly inflated by moving one lever while the other deteriorates.

That's the case for an index. Let me suggest one that we can call Loyalty Retained Yield (LRY):

LRY = (% of total revenue from Existing-Active members) / (% of member database that's active)

Mechanically, this is average revenue per active member relative to average revenue across the whole database - how much more an active member is worth than the base rate. It's designed for comparison across time. A rising LRY means your active members are pulling further ahead of the average; a flat or falling LRY, even alongside a growing "active member" count, means you're accumulating members without deepening value.

To see why this matters in practice, look at what happens when you back into the numbers from Chipotle’s public disclosures across Q1 and Q2:

• Q1 Baseline: With an estimated total database of 54 million members and 21 million active, 38.9% of the database was active. Combining channel mix, digital penetration, and in-store scan rates yields roughly 42.9% of total revenue coming from active members. Q1 LRY then is 42.9% Revenue Share ÷ 38.9% Database Share = 1.10x.

• Q2 Relaunch Push: Active members rose to 23 million (roughly 41.1% of the database). But same-store sales growth slowed, and active-member revenue share dipped slightly to 42.7%. Q2 LRY is 42.7% Revenue Share ÷ 41.07% Database Share = 1.04x.

Three critical insights fall out of that shift:

1. Active members are outperforming, but barely: Anything above 1.0x means active members generate more than their baseline share of revenue. But an LRY hovering at 1.04x - 1.10x reveals that an "active" member is performing barely better than an unengaged account in their system, largely because 80% of physical traffic remains unidentified.

2. Growth diluted member yield: Headline active membership grew by nearly 10% quarter-over-quarter. On a board slide, that looks like momentum. In reality, adding those 2 million members drove LRY down from 1.10x to 1.04x. The program accumulated accounts, but average quality fell.

3. It exposes why public disclosures hide the denominator: Most brands don't report physical scan rates or total database size because a pure "Active Members" headline lets them claim growth even when yield is eroding. LRY forces transparency.

Here's another area where it could be useful for a CMO defending budget: this can happen even while the top line looks great. Same-store sales up 10%, active member count up, and LRY still falling - that combination doesn't mean the loyalty program is failing the business. It means the loyalty program's share of a growing pie is shrinking. The growth is coming from somewhere the identification and personalization engine isn't touching: price increases, unidentified walk-in traffic, general demand. The company is fine. The loyalty investment isn't the thing driving it. That's a very different, and much less comfortable, conversation than "comps are up, ship it" - and a single headline number will never force it. Only a ratio that can fall while everything else looks up will.

One important caveat: LRY should be expected to see a temporary structural drop when a company shifts from manual scanning to passive, payment-linked identification. Unmasking a large wave of previously anonymous, likely lower-frequency in-store diners swells the "active" database share overnight, diluting the average value of that bucket even though nothing about existing customers' behavior has changed. The health signal isn't the post-rollout snapshot - it's the slope of the LRY trendline over the following two quarters.

To keep the metric honest, internal teams need two guardrails:

Component reporting. Always report LRY alongside its two components - active-revenue share and active-database share - the same way comps are reported alongside traffic and check, not just as a blended number. That way a shift caused by redefinition rather than behavior is visible, not hidden inside a ratio.

The Habitual Share core. Pair it with a metric the active-window definition can't touch: Habitual Share. That's the percentage of transactions coming from members with, say, three or more visits in the trailing 90 days. This is a much higher bar than "active in the last six months," it can't be inflated by loosening a lookback window, and it's closer to what actually predicts durable revenue: not whether someone showed up once, but whether they've built a habit.

Together, unidentified-share, LRY (with its components), and Habitual Share give you a three-part scorecard: are you finding people, are your actives outperforming the base, and is that outperformance built on real frequency rather than a generous definition of "engaged."

Shifting the North Star from database size to LRY velocity changes what the marketing and product teams actually build. Four plays move the number instead of the headline count:

Predictive Daypart Expansion. Don't treat a three-times-a-week lunch regular as a closed case. If the data shows a customer exclusively orders at 12:30 PM, skip the lunch coupon. Instead, trigger a 2:30 PM push for a mid-afternoon snack or beverage. New frequency, no discount on the order that was already happening.

Passive In-Store Identity Capture. Eradicate the manual scan. By linking loyalty profiles passively to payment methods like Apple Pay at the physical register - a roadblock Chipotle's leadership explicitly validated on their call - casual regulars are seamlessly pulled out of the "unidentified" bucket without forcing them to change real-world checkout behavior.

Behavioral Streak Gamification. Slow points accumulation causes abandonment before it causes loyalty. A guest who visits every three weeks doesn't need a 1,200-point target - they need a two-week challenge with a visible progress bar and a near-term payoff. Chipotle's own "Summer of Extras" relaunch leans on exactly this: monthly streak challenges and local and national leaderboards built to shrink the purchase cycle, not just accumulate points that may not pay off for months.

Smart Re-engagement Triggers, timed to the individual, not a global window. A generic "we miss you" email at day 90 misses the point - by then the guest may have already formed a new habit elsewhere. If someone who orders every Friday skips two weeks, trigger an offer on day 15, built around their actual usual order. You're intercepting the drift from Recently Active to Recently Inactive before it compounds - protecting the LRY denominator instead of trying to win the customer back from scratch later.

If your loyalty database is growing but your unidentified-transaction share, your LRY, and your Habitual Share aren't moving with it, the personalization engine isn't broken - it's just not being asked to do the job the headline number implies it's doing. Chipotle's own numbers this quarter are a useful public data point for the industry to reckon with: a 70-point gap between digital and in-restaurant participation, at scale, is the whole argument for why enrollment count and engagement are not the same thing.

I'd be curious what other loyalty leads are seeing on their own unidentified rate - is 80% unidentified in-store closer to the norm than the exception right now? Start a discussion in the comments below.