Three cells, one clear directional read.
The Q3 Test & Learn framework was designed to isolate the role of value in driving incremental transactions and visits versus an equity-only baseline. Across 861 completed guest conversations, both targeted value moves broke through. The Las Vegas $1.99 iced coffee drove the strongest behavioral lift; the Miami $6 combo reframed price expectations; the Denver control held steady on satisfaction but volunteered "more deals" as the single biggest improvement ask.
Headline findings
1 · Value offers visibly worked, unaided.
27 Miami guests volunteered the $6 combo when asked about deals; 21 Vegas guests volunteered $1.99. Deal-noticing in Miami (~51%) was more than double the control market (~20%).
2 · Vegas drove the strongest behavioral lift.
Iced-coffee attach hit ~27% in Vegas vs. ~17% in Miami and ~15% in Denver. The hot-vs-iced preference split also flipped to near-parity (hot 80 / iced 77 / either 68). That is a pattern not seen in the other two cells.
3 · The $6 anchor pulled price expectations down.
Median expected combo price in Miami was $7.00, a full dollar below Denver's $8.00 and the Vegas $7.99. The promotion is reshaping what a combo "should" cost, not just selling a single offer.
4 · The equity baseline is genuinely healthy.
"Fair" dominated value language in all three cells (Denver 93, Miami 173, Vegas 172) with very few "expensive" mentions. Vegas return intent was strongly skewed positive (124 very likely, 89 likely, 0 unlikely).
5 · Even the control market asked for value.
With no active offer, Denver guests still volunteered pricing/affordability (35 mentions) and more deals/combos (26) as the top changes that would improve the experience. Value cues are a baseline expectation, not just a promotional tool.
6 · Iced-coffee procedure improvements need a reason-to-try.
In Denver, only ~4 guests spontaneously noted any difference in their iced coffee. In Vegas, the same procedural improvements registered as fixed pain points (chocolate flavor properly mixed) because the offer pulled guests into the iced occasion to evaluate them.
Behavioral lift across the three cells
Deal noticing and iced-coffee attach by cell
Three cells. Three different questions.
Each test cell was deployed simultaneously to isolate the role of value in driving incremental transactions and visits, alongside an equity baseline. Denver controls for the value moves entirely. Miami tests the role of price-pointed combos. Las Vegas tests the role of cold beverages as a visit driver.
Test & learn framework
Cell construct
How to read each cell
Denver, CO
The $8 B&B combo was pulled from all channels (in-bakery, app, web, 3PD). Iced-coffee procedural improvements were also deployed. Equity-only media. The reference point for measuring the lift in Miami and Las Vegas.
Miami, FL
Adds a new lower-tier combo at $6 alongside the existing $8 Breakfast & Brew. Iced coffee included with an "add flavor" upcharge. Price-pointed value media in market.
Las Vegas, NV
Any-size iced coffee at $1.99, all four flavors included (Classic, Caramel, Chocolate, Vanilla). No upgrade option to refresher or cold brew. Iced-coffee specific media.
The equity baseline holds up.
Removal of $8 Breakfast & Brew. Improved iced-coffee procedures. Equity media only. With no active price-pointed offer, this is the cell that tells you what value perception and behavior look like on the brand alone.
Value perception held even without a promoted offer
Value language used by Denver guests
With no active offer, Denver was the least likely cell to recall a deal, and yet "fair" dominated the value language. Of 207 complete responses, 93 used the word fair, 17 used great or good value, and only 16 used anything in the expensive / too much family. The brand equity holds on its own. The Q3 question is what's incrementally available on top of this baseline, not whether the baseline itself is broken.
Even without a promo, guests volunteered "more deals"
Top spontaneous improvement themes
When asked what one change would improve the experience, Denver guests led with pricing/affordability (35 mentions) and more deals/combos (26 mentions). The hierarchy is consistent with what appears in the test cells. The asks come up without prompting; the value cues that the other cells deliver are filling a real expectation, not creating an artificial one.
Iced-coffee improvements: limited unaided signal
The Denver cell was the place to read whether procedural improvements to iced coffee register on their own, without a promoted iced-coffee offer to pull guests in. Of the 30 Denver guests who ordered iced coffee, only about four spontaneously commented on a difference, and a similar number described the drink as "same as usual." The remaining majority commented neutrally or did not engage on the procedural dimension at all. The improvements look necessary but not sufficient; they need a reason-to-try paired with them to register.
Return intent in the control
Denver was the only cell where the questionnaire explicitly asked likelihood to return. Responses skewed strongly positive, with multiple guests describing weekly or twice-weekly routines ("I visit Einsteins 4-6 times a month"). The control looks like a healthy loyal base. Q3 is being asked how to add incremental visits on top of that.
The $6 anchor broke through.
$6 Bagel + Shmear + LG Coffee combo, alongside the existing $8 Build-Your-Own Egg Sandwich + LG Coffee. Iced-coffee procedure improvements. Price-pointed value media.
The offer broke through, unaided
Unaided mentions of the test offers
Roughly half of Miami guests noticed a deal of some kind, more than twice the Denver rate. 27 volunteered the $6 combo specifically, and 16 cited the $8 combo. Importantly, the median expected combo price in Miami was $7.00, a full dollar below Denver's $8.00. The $6 anchor did not just sell a single combo. It lowered the price the segment thought a bagel-and-coffee combo "should" cost, which is exactly the elasticity signal a value test is meant to reveal.
Value perception was strongest here
Value language used by Miami guests
Of the 319 complete Miami responses, 173 used the word fair (the highest absolute count of any cell), 33 used great or good value, and 16 used expensive or too much. Even guests who said "fair but slightly expensive" tended to attribute it to general inflation rather than to the brand. This is the cell where the value message and the operational delivery look aligned.
Combo composition: a real split between simpler and upgraded
Preference for the $6-tier vs the $8-tier construct
Roughly 90 Miami responses leaned simpler/cheaper, and roughly 106 leaned toward a sandwich or premium pairing. Q3 should not collapse the tiers. Both are doing work for different segments, and the $6 entry and $8 upgrade together cover demand that one alone would leave behind.
Some guests converted; others were going to buy anyway
Among guests who saw the offer, the open-ends fell into two groups. Some converted directly because of the deal ("It changed my mind"); others noticed the signage but said it didn't change their order because they ordered through the app, came in for something specific, or were already loyal. Both groups matter for sizing the move. The first is incremental volume, the second is margin captured at a lower price point. Sizing the tradeoff is a Q3 behavioral-measurement question, not a survey question.
Coffee preference: still hot-skewed in Miami
Hot vs iced vs either
In Miami, hot coffee remained dominant in the combo conversation over iced/cold, with another 67 saying they want a choice. The $6 combo is being received as a primarily hot-coffee occasion in Miami despite the climate, consistent with the morning-routine framing of the offer rather than an afternoon refresher.
$1.99 drove the strongest behavioral lift.
Any-size iced coffee at $1.99, all four flavors included. $8 Breakfast & Brew removed. Iced-coffee specific media. This is the cell that tells you how responsive the cold-beverage occasion is to a sharply-priced offer.
Iced attach reached parity with hot
Coffee preference flipped to near-parity in Vegas
Iced-coffee attach in Vegas reached approximately 27% of guests, compared with ~17% in Miami and ~15% in Denver. Just as telling, the hot-versus-iced coffee preference split in Vegas was effectively even (hot 80 / iced 77 / either 68), while Miami skewed hot (102 / 59 / 67) and Denver skewed hot by default. The $1.99 offer is doing what an iced-beverage promotion should do: pulling forward cold demand and recruiting guests into an iced occasion.
Unaided recall of the $1.99 construct was strong
Unaided mentions of the test offer (Vegas)
On the surface, about a third of Vegas guests said they noticed a deal, sitting between Denver and Miami. The deeper signal is unaided recall of the specific construct: 21 Vegas guests volunteered $1.99, and another 18 referenced iced coffee in the deal-recall question without the price. The $8 Breakfast & Brew received only 2 unaided mentions in Vegas (it was removed from the menu in this cell), confirming the offer the cell is meant to test is the one breaking through.
Procedural improvements registered here
Vegas guests, more than Denver's, actively commented on the operational delivery of the iced coffee, both positively and as constructive feedback. Multiple guests called out that the chocolate flavor was now properly mixed (a known prior pain point), and others noted comparison-favorable taste. A small number still raised concerns about chocolate settling at the bottom. The pattern suggests the improvements register more when the offer is in market: guests who came in for the deal have a reason to evaluate the drink.
Flavor pricing: guests don't expect to pay more
Asked whether they expect to pay more for a flavored iced coffee versus plain, Vegas guests leaned toward "no, I shouldn't have to." Roughly 13 of the substantive responses said flavor shouldn't be an upcharge, versus roughly 7 who said they would expect to pay more. The Cell 2 construct, which treats flavors as a paid add-on inside the $6 combo, swims against this expectation. The Cell 3 construct, which includes all four flavors at the $1.99 price, is materially better aligned with how guests think about iced coffee.
Return intent in Vegas
Of 335 complete Vegas interviews, 124 said they were very likely to return in the next couple of weeks, 89 likely, 12 maybe, and zero unlikely. Survey-administered return intent is directional rather than predictive, but the absence of any "unlikely" responses across that volume is notable. Even guests with constructive product feedback said they would be back.
A consistent story across all three cells.
Side-by-side, the equity baseline is healthy, targeted value moves generate incremental behavior, and iced coffee is the more responsive lever of the two tested. Three patterns hold across every cell.
| Metric | Denver · Control | Miami · $6 / $8 | Las Vegas · $1.99 IC |
|---|---|---|---|
| Completed responses | 207 | 319 | 335 |
| Noticed a deal (% yes) | ~20% | ~51% | ~32% |
| Unaided mention of test offer | n/a | 27 ($6) · 16 ($8) | 21 ($1.99) |
| Expected combo price (median) | $8.00 | $7.00 | $7.99 |
| Value perception · "fair" mentions | 93 | 173 | 172 |
| Value perception · "expensive" mentions | 16 | 16 | 20 |
| Iced coffee attach (% of orders) | ~15% | ~17% | ~27% |
| Coffee preference (hot / iced / either) | hot-skewed | 102 / 59 / 67 | 80 / 77 / 68 |
| Top spontaneous improvement theme | Pricing (35) | Pricing (48) | Pricing (54) |
Iced-coffee attach: the headline lift
Iced-coffee attach rate by cell
Expected combo price: Miami pulled the anchor down
Median expected combo price by cell
Top spontaneous improvement themes by cell
Coded themes from the one-change open-end
Three patterns that hold across every cell
Pricing/affordability is the #1 spontaneous improvement ask in every cell.
This is a structural finding, not a market-specific one. Guests mention price as the lever to pull even when they call current value "fair."
Rewards / app and "more deals" appear in the top themes everywhere.
The digital channel and the value channel are complementary, not substitutes. The $6/$8 in-bakery signage in Miami left app orderers under-served.
Iced coffee underperforms hot coffee on attach in two of three cells.
Only the $1.99 promotion brought iced coffee to parity with hot. Consistent with the framework hypothesis that cold beverages are an opportunity area requiring a targeted intervention.
Five reads worth carrying forward.
Each Q3 question the framework set out to answer has a clearer-than-expected directional answer in the qualitative record. Behavioral measurement (transaction counts, ticket size, mix) is still required to size the moves, but the survey signals are directionally consistent.
Carry a value combo into Q3, with two tiers, not one
Miami showed that a $6 combo breaks through and pulls expected combo price down, but it also showed that nearly half of guests want the upgraded sandwich pairing at $8. The two tiers are doing work for different segments. Position the $6 as the entry point that drives incremental low-ticket trips; let the $8 carry attach value and protect average ticket.
Treat $1.99 iced coffee as a recruitment lever, not just a discount
The Vegas pattern (hot/iced near-parity, ~27% iced attach, strong unaided recall, zero "unlikely to return") reads as a recruitment moment more than a margin-eroding one. The Q3 question is whether $1.99 needs to be the carry-forward price point, or whether $2.49 / $2.99 with the same media treatment captures most of the lift.
Fix the digital cue: in-bakery signage missed app orderers
Multiple Miami guests said the offer didn't influence them because they ordered on the app. For Q3, deal cues should appear in-app at the same prominence as in-bakery signage, or the offer will under-deliver against the meaningful portion of guests who never see the window banner.
Pair iced-coffee procedure improvements with an active reason-to-try
Denver guests largely did not notice the procedural improvements on their own. Vegas guests, pulled into the iced occasion by the offer, did notice and reinforce them. Procedure improvements are an enabling condition, not a marketing message, and should be scaled alongside any iced-coffee promotion rather than ahead of it.
Don't include flavors as an upcharge inside the value combo
Vegas guests leaned against expecting to pay more for a flavored iced coffee versus plain. The Cell 2 construct, which treats flavors as a paid add-on inside the $6 combo, runs against guest expectations and should be reconsidered for the Q3 carry-forward. The Cell 3 construct (all four flavors included at the price) is materially better aligned with how guests think about iced coffee.
One read worth flagging: the equity baseline is genuinely healthy. Don't over-correct.
Across 207 Denver guests with no active promo, "fair" outweighed "expensive" by 6 to 1, and spontaneous return-intent language was overwhelmingly positive. The equity campaign and the operational base are doing their job. Q3 value moves should be designed as additive trial drivers and visit-frequency drivers, rather than as a response to a value problem the data does not show.
How the data was collected and coded.
In-bakery / post-purchase intercept survey administered across three test markets between May 7 and May 12, 2026. Approximately a dozen open-ended questions per guest covering trip driver, deal awareness, value perception, combo appeal, iced-coffee experience, and one open-ended improvement prompt.
Completed conversations by cell
| Market | Total responses | Completed | Pending | Fraud flagged |
|---|---|---|---|---|
| Denver · Cell 1 | 237 | 207 | 29 | 1 |
| Miami · Cell 2 | 391 | 319 | 55 | 17 |
| Las Vegas · Cell 3 | 402 | 335 | 64 | 3 |
| Total | 1,030 | 861 | 148 | 21 |
Definitions and conventions
- Inclusion
- Only rows where status = "complete" are included in qualitative coding. Fraud-detected and pending responses are excluded.
- Theme coding
- Open-ends were tagged with a lightweight keyword-and-stem approach. A single response can be tagged into multiple themes (e.g., a comment mentioning both "price" and "app" counts once in pricing/affordability and once in rewards/app). Theme counts are response counts, not mention counts.
- Price extraction
- Expected combo prices were parsed by extracting any dollar amount between $2 and $30 from the response to Q6 (Miami / Vegas) or Q10 (Denver). When a guest gave a range ("$5 to $7"), each endpoint was counted once.
- Comparability
- The Denver questionnaire has a longer structure (35 columns including a return-intent question) than Miami and Vegas (33 columns). Question wording for the directly compared metrics (value perception, deal recall, expected combo price, simpler-vs-upgraded combo preference, iced-coffee experience, and one-change ask) is functionally equivalent across all three cells.
- Read with caution
- Survey-administered behavioral self-report (e.g., return intent, "would you buy") is directional rather than predictive. Pair these findings with same-store sales, transaction counts, ticket size, and product mix from the actual test period before sizing any Q3 move.