Europe Partnership Workstream · August 2026
How this was built
Forty-seven moves for a German forecourt operator. The judgement behind them comes from nearly a decade of walking c-stores, QSRs and charging estates in every country in Europe, plus Turkey and Israel — cities and countryside, on long road trips. The measured core of that practice is 538 geotagged photographs and 22 video clips taken across ten European markets between 2017 and 2026. Every lever is scored on a published model, and every lever cites the frame it came from.
Start here
Two documents
Lever Board
All 47 levers in sortable rows with their transportability scores. Click any row to open its photo evidence, rationale and read design. Filter to either board, or to the eighteen marked start-now.
Open the board → The argumentFull Findings
Eight findings with the evidence attached: the method, the measured base rates, the German gap map, the two-model split across Europe, and both boards written out card by card.
Read the findings →Method
How 538 photographs became a set of base rates
- 01Read metadata before pixels. 537 of 538 frames carry both a timestamp and GPS, which makes the archive self-describing before a single image is opened.
- 02Cluster and segment. A greedy ~40 km clustering collapses those points into 51 locations; a 25-minute gap rule then splits each location into discrete store visits, producing the 174-visit denominator every prevalence figure is computed against.
- 03Montage into readable scenes. Nine frames to a sheet, stamped with index, city and date and sorted chronologically — so a whole store visit lands on one sheet and reads as a scene rather than nine disconnected pictures.
- 04Re-read at full resolution wherever the wording is the insight. Roughly 390 frames re-read four-up at 780–900 px, plus all 22 clips watched. Several findings exist only in video and were invisible to three still-photograph passes.
- 05Tag blind against a fixed codebook. 35 mechanism codes and 6 observability flags, applied by six independent taggers with no access to the emerging thesis, then rolled to visit level with a per-mechanism denominator — a mechanism is only scored against visits where that part of the site was actually in frame.
Reliability. A 54-photograph overlap between the main pass and an independent check gives Cohen's κ = 0.70 on mechanism codes and 0.75 on observability — substantial agreement. The disagreement has a known direction: the checker found more observability flags than the main pass, so denominators are likely undercounted and prevalences likely overstated by roughly a fifth.
The model
What "transportable" means, made computable
Every candidate is rated 0–5 on five friction dimensions — higher always means less friction — and those roll into a Transportability Index on 0–100. The weights are the argument.
Capex and ops carry half the model between them, because with an operator under pressure the thing that kills an idea is almost never that it is a bad idea — it is that it needs a remodel or a headcount. Friction alone, though, produces a ranking that cannot respond to evidence: it judges every candidate from the nature of the mechanism rather than from what was photographed. Four further terms fix that.
EVIDENCE how the claim is known — Measured 1.00 · Transported .85 · Simulated .55–.80 · Citation .45
PROOF how well the measured base rates support it
READABILITY could it ever become Measured? public data 1.00 · client POS .65 · none .40
GERMAN ADV. 1 − German prevalence. For a German board, the gap is the edge.
Structural rows — advantage from a market condition — are ranked separately from and above execution rows. They are not sequential: execution levers are the delivery mechanism for structural ones. A decal cannot outrank a legally protected trading window, and a protected window delivers nothing without the decal.
The short version
Germany is craft-competent and provocation-poor
The most useful finding is counterintuitive, and it reframes the brief before any lever is discussed. Measured across 174 store visits on matched denominators, Germany matches or beats the rest of Europe on open kitchens, per-item flavour cards, product photography at shelf, dietary navigation and kids provision.
Where it trails is almost entirely street-level provocation — A-frame signs at 7% of German sites against 24% elsewhere, window graphics at 0% against 7%, price roundels on the case glass at 0% against 8% — plus charger-to-basket conversion at 10% against 25%.
Three things that follow
- The top structural lever and the top execution lever are the same move. The biggest asset in the market is a legally protected trading window; the biggest measured gap is the A-frame sign. A protected window converts nothing if the person walking past does not know the site is open.
- Most of this already exists in Germany — at Christmas markets, behind bakery counters, on motorway table tents, and in a supermarket produce hall filmed in 2017. It is quarantined into seasonal and specialist formats and never reaches the everyday forecourt. The ask is not to import Polish practice; it is to do year-round what German retail already does at the Weihnachtsmarkt.
- The low-base board costs €890 in capital. Eleven moves, no capex above about €500 each, no new skill, no partner, live inside 30 days. The two highest-scoring moves cost nothing and €200. The real cost is a rota change, not a fit-out.
Caveats
On the numbers
Prevalence figures come from a convenience sample of what was worth photographing, not a random sample of European retail — so the cross-market contrasts are the load-bearing output, and absolute rates should be read as ordinal rather than cardinal. Photographs were tagged blind against a fixed 35-code codebook, and inter-rater agreement was κ = 0.70. Rows marked Citation rest on operator recall and are never dressed as measured.
Six absences are worth carrying into any recommendation: no labour evidence, no back of house, thin coverage after dark, no small-town or rural sites, a thin and dated Nordic sample, and no shelf-set depth. Every claim here is about demand generation, never about the economics underneath it.