# Provisional niche ranking and INR economics

Research date: **6 October 2026**. The lead’s integrated decision selects **Bengaluru grooming quote and consultation preparation** for the first photo-focused POC. That is supported by stronger photo fit and concrete published prices. The lead’s final photo-oriented assessment scores grooming **3.60/5** and trainer preparation **3.45/5**, with equal new-tool willingness-to-pay scores of **1**. The earlier market pain-weighted ranking remains a documented sensitivity only. A still photo cannot explain behavioural cause. Keep clinical advice, individualized diet and photo diagnosis outside the POC. These are feasibility and analyst judgments, not measured demand. See the [lead decision](poc-decision.md).

The completed owner report records 54 inspected relevant India discussions/reviews, including 52 Reddit threads and two dog-service reviews. That is the owner specialist’s achieved count, not 54 new market observations. OWN-002, OWN-011, OWN-016 and OWN-019 describe difficulty judging trainers or expensive quotes; OWN-040 alleges a photo follow-up failed to reach the doctor. OWN-022 records grooming and food spending; OWN-048 describes home-bath effort versus conflicting professional quotes and a self-service preference. OWN-027/029 describe boarding compatibility/distance concerns. These support preparation, fit questions and an explicit handoff boundary. **No willingness to pay for this new AI/photo tool was found.** The final [owner report](owner-research.md) owns sample construction, bias, counts and final interpretation.

The final integrated weights are pain25%, new-tool willingness-to-pay5%, photo fit20%, data feasibility15%, operating cost10%, safety15% and competitive gap10%. These totals are supplied by the lead; the revised full component matrix was not supplied, so they are not recomputed from the earlier market matrix or extended to an invented seven-niche final ranking. The final recommendation is a provisional feasibility choice, not measured demand.

The following table preserves the **earlier pain-weighted sensitivity**, not the final integration ranking:

| Sensitivity rank | Candidate | Weighted score /5 | Why plausible; main constraint |
|---|---|---:|---|
| 1 | Trainer consultation preparation + local budget handoff | 3.55 | Trust/quote problems and costly packages; still photo offers little behavioural evidence, history essential. |
| 2 | Grooming preparation + inclusive quote checklist | 3.45 | Concrete Bengaluru prices, visible coat context; app alternatives and vet/grooming boundary. |
| 3 | Non-emergency vet consultation preparation | 3.30 | Factual timeline/questions useful; professional review, strong AI/tele-vet competition and emergency limits. |
| 4 | Boarding eligibility and full-cost comparison | 2.90 | Dates, handling, supervision and transport matter; live capacity unknown and photo weak. |
| 5 | Travel document and fee checklist | 2.85 | Expensive expert/flight alternatives; rare usage, changing acceptance and free official rules. |
| 6 | Generic trusted local discovery | 2.70 | Recurring search/trust difficulty; existing free directories, missing licensed price/slot feeds and weak payment rationale. |
| Excluded | Individualized diet/diagnosis from photo | 2.00 before gate | Unsafe/inadequate input; weighted average cannot override safety exclusion. |

Earlier sensitivity weights: pain30%, indirect existing-service payment evidence15%, photo fit10%, data feasibility15%, operating cost10%, safety of the **scoped** workflow15%, competitive gap5%. Higher is better. Those historical payment scores describe existing service spend and cannot be interpreted as new-tool WTP. Small score gaps are not statistically meaningful. An additional earlier sensitivity with photo weight30% also favored grooming; neither historical weighting overrides the lead’s final photo-oriented assessment above. Exact dimensions, sources, contradictions and validation gates are in [candidates.json](../evidence/market/candidates.json).

Data feasibility scores assume a small maintained set of source-linked records and unknown-field handling, not an existing price/availability integration. The completed [data report](data-api-feasibility.md) records a live Overpass probe with five veterinary features and zero grooming features, without prices or slots, plus six seed candidates with null prices/credentials/availability and production_usable:false. Provider pages are neither licensed databases nor APIs. Professional and provider review is pending. The friend’s screenshot suggests relevant training distribution, but follower counts are not current verified reach, owner tests or expected conversions.

Published amounts provide bounded alternatives, not a market-average budget. ThePetNest publishes Bengaluru grooming packages ₹999/₹1,499/₹1,999 and trainer rates ₹1,195 ×12 or ₹1,335 ×15. The arithmetic yields **₹14,340** basic and **₹20,025** behaviour packages; no final bundle checkout or discount was tested. [MKT-001](https://thepetnest.com/pet-grooming/bangalore), [MKT-020](https://thepetnest.com/dog-training/Venkatesh--M-in-bengaluru-with-tli-227517). Remote expert anchors range from ThePetNest’s advertised starting ₹199, Wiggles’ ₹399/30 minutes and Supertails’ ₹599 instant consult; these differ in scope and access. [MKT-037](https://thepetnest.com/online-veterinary-service), [MKT-005](https://www.wiggles.in/pages/faqs), [MKT-002](https://supertails.com/products/instant-consult).

HUFT’s ₹250 minimum service is not a full groom. PetBacker’s ₹500/night example is explicitly unavailable; use it only as price/availability contradiction. Its inspected daycare example starts ₹799/day with duration/add-on distinctions. Anvis lists dog boarding ₹750 in its catalogue; a per-night use below is a **planning assumption pending duration and inclusions confirmation**. [MKT-003](https://headsupfortails.com/pages/faqs), [MKT-008](https://www.petbacker.com/india/boarding/karnataka/bengaluru/aadi-n-ashs-pawris-home-based-pet-hotel), [MKT-009](https://www.petbacker.com/india/daycare/karnataka/bengaluru/jimmys-papa-~-pet-daycare-day-nursery), [MKT-034](https://www.anvisinc.com/collections/all).

Owner budget scenarios annualize occasional services and keep an emergency reserve separate from consumed spending. Food ₹3,000/₹6,000/₹10,000, frequency, supplies and reserve amounts are assumptions, not recommended nutrition, representative Indian spend or observed current medians. Historical OWN-022 reports ₹3,000 food/month; OWN-005 supplies a pack budget rather than a monthly total, and OWN-007 reports a roughly 25-day food purchase. They illustrate variation without establishing the chosen assumptions.

| Monthly-equivalent INR | Budget case | Routine case | High service use |
|---|---:|---:|---:|
| Food assumption | 3,000 | 6,000 | 10,000 |
| Grooming | 499.50: 6 × ₹999/year | 1,499: 12 × ₹1,499/year | 2,998.50: 18 × ₹1,999/year |
| Consults | 33.17: 2 × ₹199/year | 199.67: 4 × ₹599/year | 399.33: 8 × ₹599/year |
| Boarding | 0 nights/year | 437.50: 7 × assumed ₹750/year | 2,625: 21 × assumed ₹1,500/year |
| Supplies assumption | 250 | 500 | 1,000 |
| Spending excluding reserve | **3,782.67** | **8,636.17** | **17,022.83** |
| Emergency saving assumption | 500 | 1,000 | 2,500 |
| Cash budget including saving | **4,282.67** | **9,636.17** | **19,522.83** |

Training is a separate project cost: adding the ₹14,340 arithmetic package to the routine first year adds ₹1,195/month-equivalent, not a recurring monthly contract. Similarly, one grooming cancellation at ₹1,499 exposes ₹299.80 at 20% or ₹449.70 at 30%, with timing-policy ambiguity to confirm. No cancellation probability was invented. Taxes, vaccinations, diagnostics, medications, transport, initial equipment and exceptional emergencies are absent unless separately quoted; these scenarios do not budget complete lifetime ownership.

Travel shows why a photo cannot set a price: official Air India domestic charges are ₹7,500 cabin or ₹16,000 checked **per sector**, with tax and acceptance conditions. Actual carrier-plus-dog weight, route, dimensions and documents are necessary. Anvis publishes ₹7,500 international DIY consultation and a ₹100,000 customized relocation amount; neither is the airline fee. [MKT-027](https://www.airindia.com/content/air-india/in/en/travel-information/travelling-with-pets.html), [MKT-026](https://www.anvisinc.com/collections/international-pet-relocation/products/international-pet-relocation-consultation-from-anand-vishwanath), [MKT-011](https://www.anvisinc.com/products/anvis-pet-relocation).

Operating calculations use official **GPT-6.1 Sol Standard short-context** rates: $2 input and $10 output per million tokens, verified by opening the model page; output estimates include billed reasoning tokens. The multiple tables on the pricing page represent different tiers, so $1/$5 Batch/Flex is not the Standard rate. A separate GPT-4.1 Mini comparator has $0.40/$1.60, without assuming equivalent quality or changing the mandated research-agent model. [MKT-039](https://developers.openai.com/api/docs/models/gpt-6.1-sol), [MKT-016](https://developers.openai.com/api/docs/models/gpt-4.1-mini).

INR conversion uses **₹85/₹90/₹95 per USD as explicit planning sensitivities, not a verified current exchange rate**. Token loads, retry multipliers, human rates, incident rates and data costs are assumptions. In the base case: 5,000 total billed input tokens, 4,000 billed output including reasoning, ₹90/USD and 1.25 attempts cost ₹5.625 API/case. Add ₹1 retrieval budget, 5% × five minutes at ₹600/hour = ₹2.50 human review, and 10% × five minutes at ₹300/hour = ₹2.50 support. Variable case cost becomes ₹11.625 before payment. Neither image-token usage nor latency was measured.

| ₹99/month **hypothesis**, no observed purchases | Optimistic | Base | Stress |
|---|---:|---:|---:|
| Cases per payer/month assumption | 2 | 4 | 8 |
| API/case | 1.496 | 5.625 | 21.888 |
| All variable/case before payment | 3.071 | 11.625 | 77.388 |
| Variable/month including assumed payment leakage | 8.4784 | 48.8364 | 621.4404 |
| Contribution/month before fixed | **90.5216** | **50.1636** | **−522.4404** |
| Fixed/month assumption | 6,000 | 12,000 | 30,000 |
| Payers covering those fixed costs | **67** | **240** | **No finite break-even** |

Base fixed ₹12,000/month assumes hosting/security/logs ₹1,000, record refresh/editorial 10 hours × ₹500 = ₹5,000, and professional template review 10 hours × ₹600 = ₹6,000. Payment leakage 2.36% is an uncontracted modeling sensitivity, not a tariff or tax opinion. Founder engineering, acquisition, provider minimums, insurance/liability, legal, app-store/platform shares, output taxes and incident compensation are excluded; this is contribution-cost coverage, not profit. Mandatory five-minute professional review on every base case raises variable cost to ₹59.125/case and makes the ₹99/four-case plan uneconomic. Sampling alone is not professional safety validation.

Backend API cost applies only if this developer operates inference. A ChatGPT-hosted conversation may carry inference under the owner’s platform plan; do not assume an extra API charge unless architecture calls one. A deterministic local replay POC incurs none of these live API costs. Data retrieval ₹0.50/₹1/₹3 is a placeholder, not a verified Places/Mappls/API license quote. If standard OpenAI web search is added, the published fee is $10/1,000 calls plus content tokens; for Mini the fixed 8,000 input block makes one call ₹1.188 at assumed ₹90/USD before retries/tax. [MKT-017](https://developers.openai.com/api/docs/pricing). Refresh-time search, per-owner search and licensed integration have different cost/permission implications.

Provider compensation must remain visible. PetBacker’s inspected general fee is 15–25%, so a hypothetical ₹750 platform booking leaves **₹637.50–₹562.50 before the provider’s own costs**. This does not claim Anvis pays PetBacker commission; ₹750 is a generic arithmetic illustration. ₹500 yields ₹425–₹375; ₹799 yields ₹679.15–₹599.25. Labor, meals, rent, travel, supplies, taxes and idle capacity are not net profit. A 35% training-fee statement was discovered in a blocked FAQ snippet and is excluded from the verified base model. [MKT-029](https://www.petbacker.com/help-center/policies), [MKT-036](https://www.petbacker.com/help-center/pet-service-providers/how-much-does-it-cost-to-be-a-pet-sitter), MKT-028 access gap.

An illustrative ₹599 expert bundle, with base tool cost ₹11.625, extra support ₹50 and assumed payment leakage, leaves ₹173.2386 when provider pay is ₹350, ₹23.2386 at ₹500, and **−₹176.7614 at ₹700**. Those compensation amounts are sensitivity assumptions; no expert accepted them. Existing ₹399/₹599 consultations constrain markup plausibility. A ₹199 guided-preparation case has ₹182.6786 contribution before fixed/acquisition under the base assumptions, but contains no expert consultation and has **unvalidated willingness to pay**. Do not label it cheaper clinical care.

For provider leads, `fee × completion probability` is a sensitivity formula, not conversion evidence: ₹100 × 15% would yield ₹15 gross per referred owner, only ₹3.375 after base case cost before further expenses. ₹50 × 5% loses ₹9.125 against that case cost. Nine combinations are in the results. No referral contract, consent, completed lead or partner revenue exists. Marketplace anti-circumvention terms prevent treating its listings as permission to take contacts off-platform; prefer transparent owner links and no paid ranking until authorized agreements exist.

The credible initial business decision is **build a free/read-only preparation prototype, then test usefulness and price separately**. Existing low-cost AI subscriptions (Petraah base ₹1,199/year, Soriz ₹399/month) and expert alternatives are competitive price anchors, not proof anyone will pay this team. Neither follower count, advertised price nor owner complaints establish conversion or market size. Candidate safety and core professional handoff should not depend on revenue assumptions.

Reproduce with `python evidence/market/calculate_economics.py` from the workspace. It reads [economics-inputs.json](../evidence/market/economics-inputs.json) and writes [economics-results.json](../evidence/market/economics-results.json), without network, credentials or inference. The source/rank/arithmetic validation script and results are in evidence/market/. Required next evidence: professional-approved preparation questions, available/provider-authorized fresh records, inclusive quotes, actual owner usability, a payment experiment and a negotiated provider compensation model. None is fabricated as completed here.