# 13 — Bulgaria own-field experiment

**Added:** 2026-09-15  
**Date:** 2026-09-15  
**Index:** [`CHANGELOG.md`](CHANGELOG.md).  
**Owner:** AT / Velocity EU  
**Audience:** founders. Idea-validation experiment — not a product launch.  
**Companions:** [`04-bulgaria-pilot.md`](04-bulgaria-pilot.md), [`06-regulatory.md`](06-regulatory.md), [`03-competitors.md`](03-competitors.md), [`11-platform-vs-robotaxi.md`](11-platform-vs-robotaxi.md), [`14-hybrid-vs-oem-dock-autonomy.md`](14-hybrid-vs-oem-dock-autonomy.md), [`docs/Canopy_Startup_Plan.md`](../docs/Canopy_Startup_Plan.md).  
**Competitor context (this pass):** [`briefs/competitors-2026-09-15.md`](briefs/competitors-2026-09-15.md).

**Purpose:** stress-test what a test on *founder-owned land* can prove — and what it cannot — so “we can fly our fields” is not mistaken for “the network works.”

**Evidence grades (this note):** **A** behavioural (what a grower or crew actually does) · **B** primary documents · **C** corroborating · **D** proxy · **E** assertion. Design rows below are the experiment plan, not field results.

Facts about named contractors and ferry fees come from the 2026-09-15 competitor pass. Pass / fail thresholds are **HYPOTHESIS** pre-commit gates, not measured outcomes.

---

## Decision this experiment serves

Decide whether Canopy’s **thin slice** (request → operator accepts → live map → regulatory-shaped receipt) is worth more capital and outside pilots in Bulgaria — not whether drones can spray land in general (that is already true; Agro Drone BG, SkyAgro, Adronom and peers already fly).

The 90-day sketch in [`04-bulgaria-pilot.md`](04-bulgaria-pilot.md) still holds: thin booking, live map, Bulgarian Food Safety Agency (BFSA)-shaped PDF receipt, Sentinel-2 + OpenWeather, ten jobs, survey-first if the spray path is incomplete. This note is the *falsification design* for that sketch when the first plots are the founding team’s.

---

## Bottleneck beliefs to falsify

| ID | Belief | If false… |
|----|--------|-----------|
| **B1** | A grower (even the founder, then a neighbour) will complete a **small-plot job** (target ~0.6–2 ha) on a phone path without a phone call to the crew | Product is ops theatre; WhatsApp still wins |
| **B2** | At least one **independent licensed crew** will accept a Canopy-shaped job (not only a founder-owned drone) | This is a self-flying farm tool, not a network |
| **B3** | **Batching / ferry** of nearby small plots can be priced so the crew still shows up | The sub-2 ha mandate dies economically |
| **B4** | A **receipt artefact** can be produced that matches the fields BFSA / Regional Food Safety Directorate (ОДБХ) / Plant Protection practice needs | “Compliance operating system (OS)” is marketing |
| **B5** | **Live customer view** is valued enough that someone watches and would miss it if removed | Live map is a demo feature, not retention |
| **B6** | Satellite / weather **cue** changes *when* or *whether* a job is booked (not just a pretty tile) | Cue is analytics cosplay |

Own-land alone can strongly test **B1, B4, B5, B6** and weakly test **B3**.  
**B2** requires at least one outside crew. Liquidity, take rate, and civil-protection surge are **out of scope**.

SkyAgro already signals a ferry fee on small jobs (**A**, competitor pass). B3 is therefore a priced note, not a slogan.

---

## What own-land can prove vs cannot

| Can prove (high information) | Cannot prove (do not overclaim) |
|------------------------------|----------------------------------|
| End-to-end UX friction of request → live → receipt | Two-sided marketplace liquidity |
| Whether BFSA-shaped PDF fields are complete enough to file or hand to an advisor | National regulatory acceptance at scale |
| Ferry time / cost for small plots on *these* parcels | Bulgaria-wide operator density |
| Crew willingness *if* one licensed Agras / XAG crew is hired for a paid scripted job | Organic discovery of operators |
| Survey-first legal path vs spray-first risk | EU product-approval portfolio for every crop / chemistry |
| Whether on-demand dispatch still feels timely when the crew is twenty minutes away | Tesla-style dock economics |

The competitor brief’s “what a Bulgaria own-land test can / cannot falsify” list is the same split. Own-land does **not** falsify marketplace density, sustainable take rate, cross-border (Romania / Greece / Serbia) friction, original equipment manufacturer (OEM) dock unit economics, product-by-product aerial plant-protection product (PPP) approvals at EU scale, whether satellite support maps *create* paid jobs, or civil-protection procurement cycles.

---

## Experiment design (4–8 weeks)

### Phase 0 — Instrument (week 0–1)

- Pick **3–8 own parcels** spanning sizes (include at least one **under 2 ha** and one larger block).
- Confirm job type for version 1: **survey first** (repo thin slice) unless the spray legal path for that week is already clear with a licensed applicator. European Union Aviation Safety Agency (EASA) Specific / predefined risk assessment PDRA-S01 and Bulgarian Art. 109a still apply if anything is dropped — [`06-regulatory.md`](06-regulatory.md).
- Draft the **receipt field list** against Bulgarian treatment / survey documentation practice (cross-check [`06-regulatory.md`](06-regulatory.md) + BFSA / ОДБХ sources) — even if filing is manual at first.
- Matching mode for the test: **scripted hire** (invite 1–2 known Agras / XAG crews: Agro Drone BG, SkyAgro, Adronom, or an unnamed licensed crew) vs a pure marketplace (too early).

### Phase 1 — Concierge OS (week 1–4)

Run **10–20 scripted jobs** on founder land (and ideally **2–3 neighbour plots** if permission exists):

| Step | Concierge allowed? | Must still feel like the product |
|------|--------------------|----------------------------------|
| Request on phone / web | Ops may prefill | Grower taps confirm |
| Match | Human can assign crew | Grower sees ETA / crew identity |
| Live map | Crew or device beacon from the pilot app — not the aircraft path | Grower watches without calling |
| Receipt | PDF may be semi-manual | Fields stable; grower can download |

**Pass / fail (pre-commit — HYPOTHESIS gates)**

| Metric | Pass | Fail |
|--------|------|------|
| Jobs completed with live view used | ≥70% of jobs | <40% watch, or call instead |
| Time request → wheels-up (or takeoff) | Median <24 h for survey windows you control | Multi-day chaos with no learning |
| Receipt completeness | ≥90% of required fields populated | Crew refuses the format / fields missing |
| Outside-crew jobs (B2) | ≥3 jobs by a non-Velocity pilot | 0 — only self-fly |
| Small-plot economics note | Written ferry + setup cost per <2 ha job | Hand-waved “batching will fix it” |
| Grower NPS-style (founder + one neighbour) | Would book again without a phone call | Prefer WhatsApp forever |

### Phase 2 — Neighbour / second grower (week 4–8)

- At least **one paying or cash-cost job** where the landowner is **not** the founding team.
- Same rails. If they will not book without someone standing next to them, B1 fails for strangers.

### Explicit non-goals

- Building a consumer discovery marketplace.
- Buying an XAG XA1 / DJI dock.
- A civil-protection ministry sale.
- Claiming Bulgaria product-market fit from own-land-only flights.

---

## Kill / reshape rules

| Outcome | Interpretation | Next move |
|---------|----------------|-----------|
| Live + receipt delight; outside crew accepts | Thin-slice OS thesis alive | Recruit 3–5 Bulgarian crews; expand parcels beyond founder land |
| Self-fly works; no outside crew | Farm tool / integrator, not a network | Become premium operator software for crews, or pause “Uber” language |
| Receipt impossible / rejected | Compliance wedge fails | Narrow to a survey-only ops tool, or partner with existing compliance vendors |
| Live ignored | Cut live scope; compete on dispatch + receipt only | Re-price the build |
| Small plots never pencil | Drop the <2 ha mandate, or require dense clusters only | Honest wedge = mid / large estates |

---

## Capital allocation stance

**Fund:** concierge software + telemetry path + receipt template + paid crew days on founder land.  
**Do not fund yet:** owned autonomous docks, satellite purchase, national marketing, multi-country launch.

That is the same capital spine as [`11-platform-vs-robotaxi.md`](11-platform-vs-robotaxi.md): build the platform first; do not buy docks in seed.

---

## Link to autonomy (robotaxi) research

Own-land tests **demand and receipt**, not dock utilisation. Only after **B2** (outside crews) fails *in a region with demand* does Option C “field `canopy_fleet`” become an information-rich experiment.

See the companion note: [`14-hybrid-vs-oem-dock-autonomy.md`](14-hybrid-vs-oem-dock-autonomy.md). That note **reinforces** 11’s “build A first / don’t buy docks in seed” conclusion; it does not replace it.

---

*Experiment design for research validation — 2026-09-15.*