Headline data points driving the verdict
Revenue trajectory
Crumbl is the only brand in the set below its starting level: indexed proxy spend 86 vs peers 133–250 at Sep-2026
Monthly proxy spend = Transactions × Avg transaction amount derived
Source: Carbon Arc — Credit Card US Core Panel, Transactions (insight 627, T1) & Avg Txn Amount (insight 629, T2); entities Crumbl(13677), Dutch Bros(64186), Krispy Kreme(64283), NBC(342165), Insomnia(41080). Indexed series = D5; absolute = D1. inference Oct-2024 looks partially covered — all five brands sit 45–50% below Nov-2024 — so the index base and the CAGR start point are both soft.
New-unit vs same-store split for Dutch Bros: ~15–17pp of its systemwide growth is new units
Same-shop sales % are SEC-disclosed; new-unit contribution = Systemwide sales YoY − systemwide same-shop % derived.
Source: Carbon Arc — Company-Reported Actuals/KPI (insight 121840, SEC-sourced, T10/D8), entity Dutch Bros Inc (64186), 2024Q4–2026Q2.
Traffic vs price
Crumbl Cookies — decline is traffic-led: transactions −27.8% YoY vs avg ticket +7.1% at Sep-2026 (net −22.7%)
YoY contribution to proxy-spend growth, decomposed into transaction (traffic) and avg-ticket (price) effects derived.
Source: Carbon Arc — Credit Card US Core Panel (insights 627, 629; T1/T2 → D4). Decomposition and YoY are derived; levels are observed.
Customer behaviour — repeat base, churn, acquisition
Crumbl's repeat base shrank fastest in the set: 27,855 → 5,142 panel cardholders (−81.5%) by Aug-2026
Repeat shoppers indexed to each brand's own Oct-2024 = 100 derived; darker = further below its own starting base.
Source: Carbon Arc — Retention, Number of Shoppers (insight 124734, T4), all 5 entities, Oct-2024–Sep-2026. inference The simultaneous Sep-2026 drop across all five brands is flagged as a pull/panel anomaly, not a brand signal; the −88.7% headline in D9 uses that month, the −81.5% above stops at Aug-2026.
Loyal-shopper churn roughly doubled (0.47% → 0.96%) by Jul-2025
Monthly churn rate, as reported observed.
Source: Carbon Arc — Retention, Loyal-Shopper Churn Rate (insight 581, T5), Oct-2024–Jul-2025 (10 months). The Apr-2025 collapse across every brand coincides with the Apr-2025 spike in new-loyal customers and is flagged as a panel-reset anomaly.
Acquisition was already falling: 2,387 → 1,035 new loyal customers / month (Oct-24 → Jul-25)
New loyal customers per month observed; Apr-2025 panel-reset spike shown with a hollow marker.
Source: Carbon Arc — Retention, Number of New Loyal Customers (insight 577, T6), Oct-2024–Jul-2025.
Engagement diverges from spend: iOS app MAU troughed at 2.78M (Mar-2026) then recovered +64% to 4.58M
Monthly active users, iOS observed.
Source: Carbon Arc — Monthly App Users iOS (insight 192635, T9), entity Crumbl(13677). inference The U-shape in MAU against a still-falling card panel points to a product/marketing cycle (app relaunch, flavour-drop promotion) rather than a single structural demand trend.
Web demand signal peaked May-2025 at 20,401 sessions and is −32% since
Monthly website sessions, as reported observed, all five brands.
Source: Carbon Arc — Website Traffic (insight 379, T7), all 5 entities, Oct-2024–Sep-2026.
Unit-level signals
Foot traffic confirms peer strength: Dutch Bros +132.1% and Krispy Kreme +79.5% modelled visits since Oct-2024
Modelled monthly visits observed.
Source: Carbon Arc — Foot Traffic (PassBy, insight 45862, T8), entities Dutch Bros Coffee(8242) and Krispy Kreme(42329), Oct-2024–Sep-2026 (Oct-2026 partial excluded).
Competitive position
Crumbl is the only brand with negative growth at comparable scale: −7.5% CAGR vs +16% to +61% for peers
Growth (proxy-spend CAGR, Oct-24 → Sep-26) vs scale (avg monthly proxy spend, Nov-24 → Sep-26) derived.
Source: Carbon Arc — Credit Card US Core Panel (627, 629) → D1/D3. Scale is panel proxy spend, not system sales; it is comparable across brands only to the extent panel coverage is comparable.
Share of wallet within the set halved: Crumbl 33.5% → 17.0%, Dutch Bros 26.0% → 38.1%
Share of combined 5-brand proxy spend, Oct-2024 vs Sep-2026 derived.
Source: Carbon Arc — Credit Card US Core Panel (627, 629) → D6. Share is within this five-brand set only, not a true category share of wallet.
Digital mix is not offsetting: Crumbl's online share flat at ~5–7% while Dutch Bros scaled 3.5% → 13.8%
Online / (online + physical) card users derived ratio of observed counts.
Source: Carbon Arc — Credit Card Users by Channel (insight 628, T3 → D7), entities Crumbl(13677), Dutch Bros(64186), Krispy Kreme(64283).
Risks & upside
Risks
Every line below is tagged Observed (direct Carbon Arc output), Derived (mechanical transform of Carbon Arc output) or Inference (analyst read, not observed).
Upside / mitigants
Same tagging convention. None of these reverses the verdict on the observed evidence; each is a condition to test in management diligence.
Data notes — datasets used, coverage, limitations
Window. All series are monthly, Oct-2024 → Sep-2026 unless stated. Oct-2026 is a partial-coverage month (transactions ~25–40% of a normal month for every brand) and is excluded from all growth figures and from the charts by default. inference Oct-2024 also looks partially covered — all five brands are 45–50% below Nov-2024 — which flatters every peer index and makes the CAGR start point soft; it is retained because the source index (D5) and CAGR (D3) are both based on it.
Proxy spend. Every "spend" figure on this dashboard is Transactions (627) × Avg Transaction Amount (629) — a derived panel proxy, not observed dollars, and not system-wide revenue. Levels reflect Carbon Arc panel coverage, so cross-brand level comparisons carry coverage risk; within-brand trends are the more reliable read.
Anomaly flags. Sep-2026 shows a simultaneous ~50% drop in repeat shoppers across all five brands (T4) — treated as a pull anomaly. Apr-2025 shows a simultaneous churn collapse (T5) and new-loyal-customer spike (T6) across all brands — treated as a panel reset. Insomnia's Jun-2025 avg ticket ($33.02 vs ~$18 either side, T2) is a small-panel artefact. None of these are read as brand signals.
Prepared from the Carbon Arc evidence pack (tables T1–T10) and the chart-builder's derived-metrics workbook (D1–D9). No value on this page is interpolated, extrapolated, smoothed or estimated.