SB StartupBasket
All ideas
74 /100 GO Medium complexity

GasRunway — retrofit timing planner for EU food retailers

Tells an independent grocer which refrigeration units to replace in which year, before their refrigerant prices itself out.

CleanTech SaaS EU SMBCompliance-drivenAI-agentCapital-planningMulti-site
— views
Evaluation Scores
74/100

GO

Overall Score

15
Problem
12
Demand
12
Build
11
Distrib.
11
Revenue
7
Time
6
Defense

GasRunway

1. One-liner

Tells an independent grocer which refrigeration units to replace in which year, before their refrigerant prices itself out.

2. Trend signal — why now?

The EU is not banning your refrigeration equipment. It is starving the gas that fills it, on a published schedule, and letting the price do the work. That distinction is the whole business.

The quota is being cut on a clock. The 2025–2026 HFC quota stands at 42.9 Mt CO₂-eq — a 48% cut from the 2023 level of 82.3 Mt. It tightens to 24% of the 2015 baseline in 2027. And effective 1 August 2026, an amendment cut Q4 2026 import quotas for high-GWP refrigerants — R134a, R404A, R410A — by 35% versus the same period in 2025. Supply is being removed from the market faster than most operators’ price lists have adjusted for.

A new fee landed on 1 January 2026. Article 17 of Regulation (EU) 2024/573 imposes €3.00 per tonne of CO₂-equivalent on every quantity of F-gas placed on the market. Because the levy scales with GWP, it lands hardest on exactly the old gases: roughly €5.77/kg extra on R-410A. Manufacturers and importers pay it, and they have passed it straight through. Critically — recycled and reclaimed refrigerants are exempt. That exemption is why there is no single right answer per operator, and why a generic “replace everything” article is useless.

The prices already moved, hard. Since EU quotas were reduced, legacy HFC refrigerant prices rose by over 1,000%. R404A now runs €446–€499 per 10 kg cylinder, and Q4 2025 alone saw a 9% quarterly rise in reclaimed/regenerated R404A. Virgin R-410A rose 15–25% across EU markets between 2024 and 2026.

The service cliff is dated, and it is not 2026. Since 1 January 2025 there is a service ban on virgin gas with GWP ≥ 2,500 (R404A at GWP 3,922, R507A). Existing systems may be serviced with reclaimed R404A until 1 January 2030. From 2032 the threshold drops to GWP 750, catching R134a and R407C. So an operator’s real question is never “is it banned” — it is “will there be affordable reclaimed gas in the year my compressor next leaks, and is that year before or after I was planning to spend the money?”

Nobody answers that question for a four-store grocer. The compliance software market answers a different one: Field Ascend, Collabit, and MSA Parasense all sell F-gas logbooks, leak-check scheduling, and CO₂e audit trails — built for contractors and enterprise facilities teams, priced and shaped for them. They record the past. Nobody prices the future for the person who owns the asset.

Provenance:

3. The opportunity

The entire F-gas software category points the wrong way. Field Ascend, Collabit, MSA Parasense — all of them build digital logbooks: per-equipment leak-check scheduling, CO₂e tracking, five-year audit-ready records for the Environment Agency. That product exists because the contractor is legally certified and the enterprise facilities team gets audited. Both are recording what already happened.

The independent grocer has a different problem and no product for it. He owns six refrigeration assets of varying age charged with three different gases. Each one has its own private deadline that is a function of gas type, charge size, leak history, and the reclaimed-gas market in his country. He will discover those deadlines the way operators always do: a compressor fails, the contractor quotes him a regas at a price that has quadrupled, and he makes a five-figure capital decision standing in his own stockroom with a technician waiting.

That is the gap. The incumbents sell proof of compliance to the party who owes it. Nobody sells forward capital timing to the party who pays for it. GasRunway is not a compliance product — it never files anything. It answers one question per asset per year: replace now, retrofit to a drop-in, or ride it and budget for reclaimed gas? — with the euro figure attached to each branch.

The 10× is not intelligence, it’s altitude. A refrigeration consultant would charge €2–5K to survey a site and answer this once. It goes stale the moment the quota schedule moves — which it did on 1 August 2026. A model that re-runs every asset against the current published schedule and the current reclaimed-gas price makes a stale one-off survey into a live plan.

4. Target market

  • Primary customer: Owner or operations manager of an independent EU food business running its own refrigeration — independent supermarkets and convenience stores, butchers, bakeries, fishmongers, small cold-store and distribution operators. 1–8 sites, 4–30 refrigeration assets, €1M–€25M revenue. Big enough to own real refrigeration capital, too small to employ an energy manager. Germany, Netherlands, Poland, Spain, Italy, Ireland first — markets with dense independent food retail and active F-gas enforcement.
  • Why they buy: Because the number moved and nobody told them. They have a mental model that says “my system is legal, so it’s fine” — which is true and irrelevant. What is actually happening is that the consumable their asset depends on is being deliberately withdrawn from the market on a published schedule, and the first time most of them price that in is at the moment of failure, when they have zero leverage and a technician on site. They will pay for the version of that conversation that happens twelve months early, at a desk, with a budget.
  • Rough TAM reasoning: IBISWorld counts 729,000 supermarket and grocery businesses in Europe, plus the Food, Drink & Tobacco Stores category (bakeries, butchers, fishmongers) at €237.7B in 2026. The independent, self-owned-refrigeration slice across the target countries is conservatively in the low hundreds of thousands of businesses. Capturing 1,200 of them at €95/mo is €1.4M ARR. This does not need to be a big share of anything.
  • Why now for them: Three dated things all bite inside their planning horizon — the Article 17 levy (live since 1 Jan 2026), the Q4 2026 35% quota cut (1 Aug 2026), and the 1 Jan 2030 reclaimed-R404A service cutoff. A grocer writing a 2027 capex budget this autumn is making the decision right now, with no data.

5. Product sketch (MVP)

  • Asset capture from a photo. Snap the nameplate on each unit; the product reads model, refrigerant type, and charge size, and builds the site’s asset register. No engineering survey, no spreadsheet.
  • Per-asset runway. Every unit gets a plain-language verdict and a date: “Unit 3 — R404A, 12 kg. Serviceable with reclaimed gas only. Hard stop 1 Jan 2030. Your next regas at current reclaimed prices: ~€540.”
  • Three costed branches per asset. Replace now / retrofit to a drop-in alternative / ride it and budget for reclaimed gas — each with a euro figure and the assumption behind it, so the owner can argue with the number instead of trusting it.
  • A year-by-year capex plan for the whole site. Which units to do in 2027, which in 2028, sequenced so the operator never faces two five-figure replacements in one quarter.
  • Price-and-quota watch. When the quota schedule changes or reclaimed prices move in their country, the affected assets are re-scored and the owner gets told which decisions changed — the thing a one-off consultant survey structurally cannot do.
  • Leak-history ingest. Drop in service invoices; a unit that leaks repeatedly has its runway shortened automatically, because its real exposure is charge size × leak rate × a rising gas price.
  • A contractor-ready brief. One page per decision the owner can hand to their refrigeration contractor to quote against — turning a vague worry into three competitive quotes.

6. AI angle — what’s load-bearing

Two places, both doing real work.

Nameplate and invoice extraction. The blocker on every asset-register product is that nobody has an asset register. Multimodal extraction turns a phone photo of a grimy nameplate and a stack of PDF service invoices into structured asset records — model, gas, charge, leak history. Without this the customer must do 3–6 hours of clipboard work before seeing any value, and they won’t. This is the difference between a 20-minute onboarding and no onboarding.

Reasoning over a messy, dated rule set against a specific asset. The answer for one unit depends on GWP threshold by year (2,500 now, 750 from 2032), the reclaimed-only service window closing 1 Jan 2030, charge size in tCO₂e, whether the Article 17 levy applies (virgin: yes; reclaimed: exempt), and national implementation quirks. Encoding that as static if-then rules is possible but brittle and re-breaks every amendment — the 1 August 2026 change would have required a rewrite. A model reasoning over the current published schedule, with the arithmetic done deterministically and shown, degrades gracefully when the rules shift.

Remove the AI and this is a consultant with a spreadsheet, at €2–5K a survey, going stale on the next amendment. That is the incumbent, and it is exactly what we are undercutting.

7. Localization angle

EU-first by construction, and the country layer is the moat. The regulation is EU-wide but the economics are national: reclaimed-gas availability and price vary by market, enforcement intensity varies by member state, and the contractor base is intensely local. A generic English-language “F-gas planner” is worth little to a Bavarian butcher.

The wedge is per-country depth: local reclaimed-gas pricing, the national certification body’s rules, native-language output, and — most importantly — a list of certified local contractors who can actually quote the recommended work. Ship Germany and the Netherlands first (dense independent food retail, strong compliance culture, high refrigerant prices), then Poland, Spain, Italy, Ireland. The UK is adjacent but a separate post-Brexit regime; treat it as a second product, not a translation.

Pricing localizes too: €95/mo works in Germany and the Netherlands, but a Polish or Spanish tier nearer €45–60/mo will be needed.

8. Business model — path to $1M–$5M ARR

  • Pricing: €95/month per site for the standard tier (up to 12 assets, full runway plan, price-and-quota watch, contractor briefs). €45/mo single-site starter for butchers and bakeries with 3–5 units. €249/mo for multi-site operators (up to 8 sites, consolidated capex plan).
  • ACV: ~€1,400 blended. Multi-site operators pull it up; the starter tier pulls it down.
  • Rough math to $1M ARR: ~700 customers at a €1,400 blended ACV ≈ €980K. Realistically ~450 standard single-site, ~150 starters, ~100 multi-site.
  • Rough math to $5M ARR: ~3,300 customers, which means being the default in three or four countries rather than one, plus the contractor channel described below carrying the majority of acquisition. Realistically requires the adjacent revenue line: paid placement / referral fees from refrigeration contractors who receive qualified, pre-costed replacement leads. That is the natural second business — the operator has already decided to spend €15K and needs three quotes — but it must not distort the advice, which means flat referral fees, never per-outcome commissions.
  • Expansion path: Assets per site grow the base tier; sites grow the multi-site tier; the second product is the post-decision layer — tracking that the replacement was actually done, its warranty, and the leak-check obligations on the new equipment, which pulls into the logbook territory the incumbents hold.

9. Go-to-market wedge — first 100 customers

  1. The refrigeration contractor channel is the whole game. Every independent grocer already has a certified F-gas contractor who visits 2–6 times a year and is the only person they trust on this. Those contractors are small (3–20 techs), they are being asked “should I replace it?” constantly, and they currently answer with a shrug or a self-serving quote. Recruit 30 contractors across Germany and the Netherlands — findable via national certification registers and trade directories — and give them the tool free to run for their own customer base, branded as their advice. Their existing book is the customer list. This solves both trust and CAC.
  2. Wholesaler and buying-group newsletters. Independent grocers in the EU cluster into buying groups and voluntary chains (Edeka and Rewe independents in Germany, Spar and Superunie affiliates in the Netherlands). These groups publish member newsletters and run supplier days. One placed piece — “the 35% quota cut that hit on 1 August and what it does to your 2027 budget” — reaches thousands of exactly-right operators with an urgent, dated hook.
  3. Refrigerant distributor co-marketing. Distributors like Schiessl already publish customer notices about the Article 17 levy — they are actively explaining the price rises to the same operators and have no tool to offer them. A distributor’s mailing list is the highest-intent list in this market: everyone on it just got a price increase letter.
  4. Trade-press hook with a free scan. A single-asset free check (“photograph one nameplate, get its runway”) is a low-friction top of funnel that converts on the moment of alarm, and gives the trade press (Cooling Post, national HVAC-R titles) something concrete to write about.
  5. Direct outreach on a dated pretext. Scrape butchers, bakeries, and independent supermarkets from national business registries in two countries; the outreach writes itself because it is factual and dated — the levy started 1 January, the quota cut hit 1 August, here is what it means for your walk-in.

10. Build complexity — justification

Medium. The application is a standard web stack over an asset register with a deterministic cost model — off-the-shelf. The AI work is multimodal extraction from nameplate photos and service invoices, which is a solved capability, not research. There is no hardware, no sensors, no integrations with refrigeration controllers in v1.

The genuine work is domain encoding and data curation: getting the GWP tables, dated thresholds, charge-size-to-tCO₂e arithmetic, and national reclaimed-gas pricing right, and keeping them right through amendments. That is not hard engineering, it is careful, ongoing, unglamorous work that needs a refrigeration-literate person. 8–12 weeks to a credible v1 for one country with a pair, plus a domain advisor. Each additional country is 2–4 weeks, mostly data and language.

The real risk to the calendar is not code; it is that the founder without refrigeration domain access will encode the rules wrong and get caught by a contractor in the first demo.

11. Gating checklist

GatePass?Note
Legal in target market✅Advisory and planning tool. Files nothing, certifies nothing, does not touch regulated handling of gas. No licence required to tell someone what a published schedule means for their asset.
Ethical — no harm / dark patterns✅Advice pushes toward earlier, cheaper, planned transitions off high-GWP gas — the regulation’s actual intent. One live hazard, flagged in section 13: contractor referral revenue must never become per-outcome commission, or the advice becomes a sales funnel.
Market exists (evidence above)✅1,000%+ price rises, dated quota cuts, a live levy, and an existing paid F-gas software category — just pointed at contractors and enterprises.
1–5 person team can build this✅Pair plus a refrigeration-literate advisor. 8–12 weeks to one-country v1.
Launchable with <$50K / ₹40L✅No hardware, no inventory. Cost is people, the domain advisor, and country data curation.

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2015/20Real money and a dated clock — but honestly, it is not hair-on-fire today. The pain is anticipatory: felt sharply at the moment of failure, dully the rest of the year. That is the single biggest weakness and the thing the GTM must overcome. Docked 5 for it.
Demand evidence1512/15Strong indirect evidence — 1,000%+ price moves, a paid F-gas software category, contractors fielding the question daily. Weaker on direct evidence that owners will pay for planning rather than absorb the shock. No verbatim owner quotes found; not fabricating any.
Build feasibility1512/15Standard stack, solved AI capability, no hardware. Docked for the domain-encoding burden and per-country data curation, which is real and ongoing.
Distribution clarity1511/15The contractor channel is specific, named, and findable via certification registers, and the distributor lists are high-intent. Docked because it is a two-step sale — recruit the contractor, then reach their customer — and contractor adoption is unproven.
Revenue mechanics1511/15€95/mo is defensible against a €2–5K consultant survey and against a single €500 regas. Docked because the $5M path leans on the contractor referral line, which is a different business with a conflict to manage.
Time to first revenue107/108–12 week build, then a contractor-mediated sale. First euro realistically 4–5 months out. Pre-selling to contractors during the build is the obvious accelerant.
Defensibility106/10Soft moat: accumulated per-country rule encoding, reclaimed-price data, and the contractor relationships. A competitor can copy the concept in a quarter but not the country depth or the installed contractor base. Not patent-grade, adequate for sub-$5M.
Total10074/100

13. Qualitative modifiers

Founder-fit tags

domain-expertise-required · sales-heavy

This is not a technical founder’s solo project. The encoding must be right or contractors will dismiss it in one demo, and the distribution is relationship work with small refrigeration firms. The ideal team is a builder plus someone with genuine HVAC-R sector credibility.

Key assumptions to validate (3–5)

  1. Assumption: Independent grocers will pay a recurring fee for anticipatory capital-timing advice, rather than absorbing the cost shock reactively at failure. How to test: Take 25 owners across two countries a hand-built runway plan for their actual site, then ask for €95/mo. Reactions to a free plan are worthless; only the payment attempt tests this.
  2. Assumption: Refrigeration contractors will hand this to their customer base rather than treat it as a threat to their own advisory position or their replacement-quote margin. How to test: Pitch 15 contractors directly. The signal is whether any will introduce it to five of their own customers, not whether they say it is a good idea.
  3. Assumption: Nameplate photos plus service invoices are sufficient to build an accurate-enough asset register without a site survey. How to test: Run 40 real nameplate photos from 6 sites through extraction; have a certified engineer verify gas type and charge. Below ~90% and onboarding breaks.
  4. Assumption: The per-asset recommendation is materially different from the generic advice a contractor gives for free. How to test: Model 30 real assets; count how many get a recommendation that differs from “replace at end of life.” If most collapse to the generic answer, there is no product.

Risk flags

  1. Anticipatory-pain risk (the main one): The customer is not bleeding today. Products sold against a future cost are structurally harder to sell than products sold against a present one, and churn is a live threat once the plan is delivered — the owner may feel he has already got the answer. Mitigation is the price-and-quota watch: the plan must visibly change often enough to be worth a subscription. If it doesn’t, this is a one-off report business, not SaaS, and should be repriced as such.
  2. Channel-conflict risk: Contractor referral revenue and honest advice pull in opposite directions. A contractor’s incentive is to replace; the honest answer is often “ride it and budget for reclaimed gas.” Flat referral fees only. If this becomes commission-driven, the product’s only asset — being the disinterested voice — is gone.
  3. Regulatory-drift risk (double-edged): Amendments land regularly; the 1 August 2026 quota cut is proof. Every amendment obsoletes stale advice — which is the reason to subscribe, but also means the curation burden never ends and a lapse produces confidently wrong five-figure advice.
  4. Incumbent-adjacency risk: Field Ascend, Collabit, and MSA Parasense already hold the asset registers of the enterprise segment. If any of them add a forward-timing view for their existing base, the differentiation narrows fast. The defence is the segment they don’t serve — the 1–8 site independent — and the contractor channel that reaches it.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Builder + refrigeration-sector domain advisor; sales-led into
                        the contractor channel, not product-led self-serve
Time to revenue:        4–5 months (8–12 week build + contractor-mediated first sale)
Capital to launch:      €25–40K ($28–45K) — two people part-time, domain advisor,
                        per-country data curation
Top 3 assumptions to validate first:
  1. Owners pay for anticipatory timing advice — take 25 of them a real plan and ask
     for €95/mo; only the payment attempt counts
  2. Contractors distribute rather than block — pitch 15, measure how many introduce
     it to five of their own customers
  3. Recommendations differ from the free generic answer — model 30 real assets and
     count the ones where the verdict is not "replace at end of life"
Kill criteria:
  - Abandon if fewer than 3 of 25 owners shown a real, site-specific runway plan will
    pay €95/mo
  - Abandon if fewer than 2 of 15 contractors will introduce it to their own customers
  - Abandon if under 40% of modelled assets get a recommendation that differs from
    generic end-of-life replacement — that means the model adds no decision value
  - Abandon if an incumbent logbook vendor ships forward-timing for sub-10-site
    operators before v1 launches

15. Next step — 1-week validation sprint

  • Day 1–2: Build the rule model by hand for one country (Germany). Encode GWP thresholds by year, the 2030 reclaimed cutoff, tCO₂e arithmetic, the Article 17 levy, and current reclaimed R404A/R410A pricing. Then model 30 real assets sourced from contractor friends and public equipment listings. Count how many produce a recommendation that differs from “replace at end of life.” If it is under 40%, stop here — there is no decision value and the rest of the week is wasted.
  • Day 3–4: Hand-build genuine site runway plans for 8 real independent operators (butchers, bakeries, small supermarkets) using their actual nameplate photos. Walk each one through it in person or on video. End every conversation with a direct price ask: €95/mo, card details today, product ships in 10 weeks.
  • Day 5: Pitch 15 refrigeration contractors from the German certification register. Ask one question only: “will you introduce this to five of your customers?” — not “is this useful.”

Go / no-go: Proceed only if ≥40% of modelled assets get a non-generic recommendation, ≥3 of 8 owners commit money, and ≥2 of 15 contractors agree to introduce it to named customers. Any one of those missing means the mechanism is wrong somewhere specific — decision value, willingness to pay, or distribution — and the failure tells you which.

Interested in a detailed proposal?

Get a deep-dive with market research, competitive analysis, and implementation roadmap.

Contact us

info@startupbasket.ai