UPDATED 15 September 2026 · R2: Use the paid experiments and pivot gates in the multi-sector strategy for the expanded launch. This protocol remains the agriculture-specific reference.
Added: 15 September 2026 · Release 2026.09.15-r1
Day 0 starts when the team and pilot resources are actually committed. Dates below are relative, not claimed project progress. All numerical gates are proposed management targets. Spending must be priced from real quotes and approved through the cash model. No budget amount is fabricated.
| Window | Owner role | Work and record | Gate / decision |
|---|---|---|---|
| Days1–15 | Commercial founder | 20 structured interviews across budget owners, independent growers and operators. Ask for prior invoices, workflow timing, alternatives, failed jobs and purchase authority. Record refusals. | Identify a repeated paid problem and specific buyer. Verbal enthusiasm is insufficient. |
| Days1–30 | Field/legal lead | Verify one crop/service/region legal path and credentials; request insurance and supplier quotes; inventory seasonal availability. | No field execution without permission. If spray cannot be validated, treat surveys as a distinct test. |
| Days1–30 | Engineering lead | Collect consented original files from two OEM workflows. Parse units/timestamps; preserve provenance; compare to operator records. | Demonstrate usable data rights and reliable imports. Pivot the live-view claim if inaccessible. |
| Days16–45 | Commercial + field | Offer three paid anchor pilots with scope, price, cancellation, acceptance and data terms. Collect deposit/invoice evidence where appropriate. | Three paying anchors is a target. Failure to close triggers problem/pricing review, not a free-pilot vanity metric. |
| Days31–75 | Field + engineering | Execute10–30 external paid jobs using five qualified active operators. Instrument the current process and the Canopy process. | Positive job contribution after all measured variable costs; documented reasons for any losses. |
| Days45–90 | Product + finance | Test five small-parcel clusters where operationally sensible; compare matched trips, setup and admin. Track repeat demand and support load. | Target20% improvement in productive crew hours,95% evidence completeness and ≤10min administration/job; small sample is not causal proof. |
| Day 90 | Founders / investment committee | Review paid conversion, repeat cohorts, safety/quality, margin, evidence and market entry costs. | Continue only with paid repeat use, viable margin and reliable data. Otherwise narrow, change price/channel or stop. |
Experimental design
Use matched parcels and service conditions where possible: crop, application volume, field size/shape, crew, aircraft, travel and weather. Predefine productive hours as completed billable field work divided by total paid crew hours, including travel/setup/cleaning. Record a pre-Canopy baseline before selecting favorable jobs. Report every canceled, rejected and reworked job.
For small plots compare individual dispatch with a shared window, using actual quotes and time logs. The experiment must include mobilization, each parcel's setup and cleanup, customer communications and potential delays. A low nominal per-hectare price without these costs is not a successful result.
Ask the buyer to purchase again without a founder subsidy. Target at least two independent repeat buyers, adjusted for genuine crop seasonality. Separate completed paid repeat work, paid next-season commitments, nonbinding letters, free demonstrations and founder-owned land. Do not annualize a peak spraying month as ARR.
Decision metrics
| Metric | Exact definition | Instrument |
|---|---|---|
| Paid conversion | External buyers paying ÷ eligible quoted external buyers | Quote/invoice/payment reconciliation |
| Fill rate | Eligible jobs accepted inside promised window ÷ eligible requested jobs | Request/accept timestamps; preserve rejected demand |
| On-time completion | Accepted jobs completed inside contracted window ÷ accepted jobs due | Clock times and cancellation reasons |
| Contribution/job | Recognized Canopy job fees less payment/payout, support, data, rework, claims and other variable costs borne by Canopy | Model1 with invoices and measured staff time |
| Operator contribution | Operator service proceeds less actual labor, travel, equipment, consumables and job-specific costs | Operator cost log; exclude customer-supplied chemical from both sides |
| Evidence completeness | Required, validated fields present ÷ fields required by that jurisdiction/job | Versioned schema, missing-field audit |
| Repeat rate | Eligible first buyers making a second paid purchase ÷ eligible first buyers | Cohort and seasonal eligibility dates |
| Realized take | Earned marketplace fee ÷ matching completed service GMV | Same period, taxes/refunds consistently treated |
Quote and interview pack
Request signed service rates, minimum charges, travel/mobilization, materials, tax, validity dates, cancellation, insurance, aircraft data export rights, and example invoice/record bundles. Request loaded monthly payroll or contractor quotations, infrastructure estimates tied to usage, legal fees, settlement timing and loss allocation. Record document URL/file, supplier, currency, date, scope and expiry for every financial input.
Buyer interview: What did the last job cost? Who approved it? How often was capacity unavailable? Which software and records were required? What makes a new supplier unacceptable? What would you stop paying for if you bought Canopy? What specific paid pilot can you authorize this season?
Operator interview: What fraction of paid time is productive? Which fields are declined? What are minimum dispatch economics? Can original aircraft logs be exported lawfully? Who pays for weather aborts/rework? What evidence resolves a dispute? What would make you keep or leave the platform?
Seasonal and stop rules
September–December2026 in Bulgaria is not equivalent to a full spray season. Validate a suitable lawful in-season workflow or schedule a bounded additional test in an independently qualified region. Keep financing contingent on the unresolved spray hypothesis. Historical logs may validate imports; they cannot validate current demand. Stop a service test immediately when required permission, insurance or operator competence is missing. Stop expanding a cohort when losses lack a credible measured remedy.