LabelLint — preflight checker for India’s small food brands

1. One-liner

Upload your packaging PDF and LabelLint flags every FSSAI violation before the printer runs 10,000 non-compliant pouches.

2. Trend signal — why now?

Three things landed in the same 12 months.

First, the law moved. FSSAI notified the Food Safety and Standards (Labelling and Display) First Amendment Regulations, 2026, coming into force 1 July 2027 — new rules on allergen highlighting, added-sugar/sweetener declarations, small-pack exemptions, fortified-food marks and non-retail packaging. Every food brand that prints a label between now and then has to re-check it against a freshly changed ruleset, and most don’t have the staff to read a gazette amendment.

Second, the enforcement got teeth and got measured. The LabelBlind study analysed 5,058 labelling claims and found 33.6% non-compliant or unsubstantiated (21.3% outright non-compliant, 12.3% lacking substantiation) — 52.5% non-compliance on nutrient/health claims specifically. Compliance writers report 92% of early-stage FMCG brands fail 4–5 label checks. And Amazon’s own dietary-supplement claim crackdown went live 31 March 2026 — non-compliant listings get deactivated, no small-seller exception.

Third, the money is real and the pain is quantified. India’s D2C food market crossed $4B in 2024 inside a $121.3B packaged-food sector growing 10–12%. A reprint on a 10,000-unit run costs ₹1.8–3.2 lakh — and that’s before the delisted Flipkart SKU, the missed shipment, the ₹3 lakh statutory fine. Brands are already paying consultants ₹2,000–15,000 and using LabelBlind’s annual subscription to avoid exactly this.

Provenance:

• Signal 1 (demand): LabelBlind study — 33.6% of 5,058 food-label claims non-compliant/unsubstantiated; reprint costs ₹1.8–3.2L per 10K run — https://www.nutritioninsight.com/news/india-packaged-food-label-noncompliance.html / https://foodsure.co.in/blog/fssai-label-rules/ — May 2026
• Signal 2 (feasibility): FSSAI Labelling First Amendment Regulations 2026, in force 1 July 2027 — a fresh, machine-checkable ruleset; vision+OCR models now read artwork PDFs for font height, contrast and declared text cheaply — https://blogs.regacats.in/fssai-labelling-update-2026-compliance-guide — May 2026
• Signal 3 (economic): India D2C food market crossed $4B (2024) in a $121.3B packaged-food sector; consultants and LabelBlind/FoLSol already sell paid label-audit services — https://www.havstrategy.com/food-beverage-brand-marketing-india/ — May 2026 Category: Regulatory arbitrage

3. The opportunity

Label compliance in India is sold two ways today, and both leave the bottom of the market wide open.

Consultants and labs (TÜV SÜD label review, regional food consultants) do it manually — email a PDF, wait three to five days, get a marked-up report, pay ₹2,000–15,000 per round. Slow, opaque, and useless when you’re iterating artwork at 11pm before a printer deadline.

LabelBlind / FoLSol is the closest software incumbent — genuinely good, “India’s first AI-led digital food labelling service.” But it’s an annual subscription priced on cloud-storage slabs (a 2GB plan sized for a 400-SKU portfolio). It’s built for established multi-SKU manufacturers with a regulatory-affairs function. Esko’s Artwork Flow has a free FSSAI demo proofer but it’s a lead magnet bolted onto enterprise artwork-management software, and it talks “FDA guidelines.”

Nobody is serving the 1–5 SKU founder — the home-kitchen brand that just crossed the ₹12L line, the D2C snack startup, the spice repacker. That founder doesn’t have a regulatory team. They have a Canva file or a designer’s PDF, a printer quote, and no idea that their allergen line is the wrong weight or their “100% natural” claim is now an enforcement target. The 10× move: a self-serve, pay-as-you-go preflight check — upload artwork, get a violation-by-violation report in two minutes, fix, re-upload, free re-checks. Annotate the exact pixel, not “see Schedule II.”

4. Target market

• Primary customer: Founder or marketing lead of a small Indian packaged-food brand — D2C snacks, spices, condiments, beverages, nutraceuticals, ready-to-eat — typically 1–5 SKUs, ₹20L–10Cr turnover, no in-house regulatory staff. Skews to brands that just upgraded from FSSAI Basic Registration to a State License and are now printing “real” retail packaging.
• Why they buy: In their words — the fear is “a warehouse full of unsellable stock.” They’ve either eaten a ₹2L+ reprint, watched a peer get a SKU delisted on Flipkart, or got an Amazon claim flag. They want certainty before the print run, not a consultant’s three-day turnaround.
• Rough TAM reasoning: India has hundreds of thousands of FSSAI State-licensed food businesses; the D2C food segment alone is thousands of active brands and growing double digits. Even capturing a few thousand small brands at ₹6–15k/year is a ₹2–5Cr+ ARR business.
• Why now for them: The 2027 amendment forces a re-check of every label they own. Marketplace enforcement (Amazon, 31 March 2026) made non-compliance an immediate revenue event, not a someday-fine. And artwork now changes weekly because D2C brands iterate packaging like landing pages.

5. Product sketch (MVP)

• Upload artwork — drop a print-ready PDF or image of front-of-pack + back-of-pack; OCR pulls every text element and its size/position.
• Violation report in ~2 minutes — line-by-line: missing mandatory panels (14-digit FSSAI number, veg/non-veg mark, net quantity, FBO address, best-before), font-height failures, allergen-highlighting failures, ingredient descending-order checks, date-format errors.
• Claim risk flags — scans marketing claims (“immunity,” “100% natural,” “no added sugar,” “high protein”) and flags ones that need substantiation or are FSSAI/ASCI enforcement targets.
• Pixel-level annotation — the report points at the exact spot on the artwork, with the rule citation and a plain-English fix (“allergen names must be bold; yours are regular weight”).
• 2026 Amendment mode — toggle to check against the post-1-July-2027 ruleset so brands print labels with a 2+ year shelf life correctly.
• Free re-checks — fix and re-upload as many times as you want within a job; you pay per label, not per attempt.
• Shareable PDF report — hand it to your designer or printer as a punch-list.
• Category templates — pick “spices / snacks / beverage / nutraceutical” and get category-specific rule sets (e.g. proprietary-food declarations).

6. AI angle — what’s load-bearing

Remove the AI and this is a PDF and a checklist — i.e. nothing. The load-bearing work is two-fold. Vision + OCR extracts every text element from a messy real-world artwork file and measures font height, weight and contrast ratio against the actual print dimensions — the part humans get wrong because eyeballing 1.5mm type is hard. An LLM reasoning layer maps extracted ingredients, claims and declarations against the FSSAI ruleset: is this ingredient list in descending order, is “source of protein” substantiated at the declared value, does this claim trip ASCI guidance. That semantic check — “your label says X, the rule requires Y, here’s the gap” — is the product. A deterministic linter can’t read free-text claims; a human can’t read 40 SKUs in two minutes. The AI does both.

7. Localization angle (if any)

This is the localization play. The entire product is India-specific: FSSAI Schedule II font rules, the 14-digit license format, the veg/non-veg mark, Legal Metrology net-quantity rules, the 2026 Amendment, ASCI claim guidance. A generic global label checker (or Esko’s “FDA guidelines” tool) is structurally wrong here. Pricing is localized too — a ₹499/label or ₹6,000/year tier works where a US-priced SaaS seat cannot. Distribution is WhatsApp- and vernacular-creator-friendly: many small FBOs live on WhatsApp, and a “send your label, get a report” WhatsApp intake is a credible v2 channel. Hindi/regional-language report output is a fast follow.

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

• Pricing: Pay-as-you-go ₹499 per label check (free re-checks within the job) for occasional users; Brand plan ₹6,000/year for unlimited checks on up to 10 SKUs; Studio plan ₹24,000/year for designers/printers/consultants running checks across many client brands.
• ACV: Blended ₹7,500 ($90). Studio accounts pull the average up.
• Rough math to $1M ARR: ~₹8.5Cr. ~11,000 Brand-plan equivalents at ₹7,500 — or, more realistically, 6,000 brand subscriptions + 400 studio accounts + PAYG volume. Achievable within the D2C-food brand population.
• Rough math to $5M ARR: Needs the Studio/printer channel to become the primary engine — printers and design studios checking every client job — plus the 2027 amendment re-check wave driving a one-time spike, plus expansion into adjacent label-heavy verticals (cosmetics under CDSCO, ayurvedic/AYUSH labelling). $5M means becoming the default preflight step printers run, not a tool founders remember to use.
• Expansion path: SKU count → seats for designers → Studio multi-client → API for marketplaces and printers to check at upload/intake → re-check subscription tied to the 2027 deadline.

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

• Printer and design-studio partnerships. Packaging printers eat reprint disputes too. Sign 10–15 packaging printers and food-label design studios as Studio accounts; they run LabelLint on every client job and become a referral funnel. This is the highest-leverage channel — one printer touches dozens of small brands.
• Scrape the marketplaces. Amazon India and Flipkart food/nutraceutical listings expose the brand, the label image and the seller. Pull the back-of-pack image, run it through LabelLint, and send the founder a free one-page report of their actual violations with a “fix before the next print run / Amazon flag” CTA. Personalised, undeniable, and the work is automated.
• D2C founder communities. India D2C founders cluster in specific WhatsApp/Slack/Discord groups, Inc42/StartupTalky comment threads, and the comment sections of FSSAI-compliance blogs (foodsure, LabelVeda). Drop a free “check your label” link where the pain is already being discussed.
• The 2027 amendment hook. Publish a plain-English “what the 2026 Amendment changes for your label” explainer and a free amendment-mode check. Time it as the deadline approaches — every brand with printed stock needs to know if their next run is still legal.
• FSSAI consultants as resellers. The ₹2,000–15,000 consultants can white-label the Studio plan to speed their own audits — turn a substitute into a channel.

10. Build complexity — justification

Medium. The web stack, file upload, PDF/image handling and report rendering are all off-the-shelf, and the AI is API-based vision+OCR+LLM — no model training. The genuinely hard part is encoding the FSSAI ruleset accurately (Schedule II font tables, mandatory panels, allergen rules, the 2026 Amendment deltas) and tuning OCR to measure font height reliably on real-world print PDFs across varied artwork. That’s domain work, not research — it needs a regulatory advisor and disciplined test cases, not a breakthrough. A technical founder plus a part-time food-labelling consultant ships a credible v1 in roughly 10–14 weeks; the ruleset deepens continuously after launch.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Compliance-assist tool; no license needed. Must disclaim it’s not legal advice.
Ethical — no harm / dark patterns	✅	Helps brands comply with safety law — net-positive for consumers.
Market exists (evidence above)	✅	LabelBlind study, reprint-cost data, existing paid consultants/software.
1–5 person team can build this	✅	Technical founder + part-time labelling advisor.
Launchable with <$50K / ₹40L	✅	API costs + one advisor; well under budget.

All five gates pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Hair-on-fire at the print-run moment — a wrong label is ₹2–3L plus delisting. But it’s episodic (per print run), not daily, so not a 20.
Demand evidence	15	12/15	Hard numbers: 33.6% non-compliance, ₹1.8–3.2L reprint, paid consultants + LabelBlind already exist. Loses points for thin direct verbatim founder voice.
Build feasibility	15	11/15	Off-the-shelf AI, but accurate FSSAI ruleset encoding + OCR font-measurement tuning is real domain work. ~10–14 weeks.
Distribution clarity	15	11/15	Printer channel and marketplace-scrape outreach are concrete and named. Conversion on cold reports is unproven.
Revenue mechanics	15	11/15	Pricing benchmarked to consultants/LabelBlind; ₹6k/yr fits the wallet. $1M ARR needs ~6k+ subs — large but plausible.
Time to first revenue	10	8/10	PAYG check can sell within weeks of launch; pre-sell via printers during the build.
Defensibility	10	5/10	Moat is the ruleset accuracy + printer/consultant relationships + brand. LabelBlind is well-funded and could move down-market. Execution-and-focus moat, not structural.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required — needs a builder comfortable with vision/OCR/LLM APIs plus genuine FSSAI labelling knowledge (or a tightly-engaged advisor).

Key assumptions to validate (3–5)

1. Assumption: Small brands will pay ₹499–6,000 for a preflight check rather than trust their designer or eat the risk. How to test: Run free scrape-reports for 50 marketplace-listed brands; measure how many ask “how do I fix this” and how many would pay to re-check.
2. Assumption: OCR can measure font height and weight reliably enough on real print PDFs to flag violations without unacceptable false positives. How to test: Build the checker, run it on 100 real label files of known compliance status, measure precision/recall.
3. Assumption: Packaging printers and design studios will adopt LabelLint as a standard step and refer brands. How to test: Pitch 15 printers/studios; convert at least 3 to paid Studio pilots within 30 days.
4. Assumption: The 2027 amendment creates a re-check wave brands will act on early. How to test: Gauge response to the amendment explainer and free amendment-mode check — sign-ups and report runs.

Risk flags

1. Competitive risk: LabelBlind/FoLSol is well-funded, India-focused and AI-led; if it launches a cheap self-serve PAYG tier it attacks this wedge directly. Speed and the printer channel are the defence.
2. Accuracy / liability risk: A false “compliant” verdict that leads to a reprint or fine destroys trust and invites blame. Must ship with clear “assist, not legal sign-off” framing and conservative flagging.
3. Regulatory-timing risk: The 2026 Amendment could be deferred or revised (FSSAI has form here), softening the re-check hook. The core compliance pain survives regardless, so this dents urgency, not the thesis.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder fluent in vision/OCR/LLM APIs, paired with a part-time FSSAI labelling consultant
Time to revenue:        8–12 weeks (PAYG checks; pre-sell printers during build)
Capital to launch:      ₹4–7 lakh ($5–8K)
Top 3 assumptions to validate first:
  1. Small brands pay for a preflight check — test with 50 free marketplace scrape-reports, measure fix-intent and pay-intent
  2. OCR font-height/weight measurement is reliable on real print PDFs — run 100 known-status label files, measure precision/recall
  3. Printers/design studios adopt and refer — convert 3 of 15 pitched to paid Studio pilots in 30 days
Kill criteria:
  - Abandon if OCR precision on font-size violations stays below ~85% after tuning — false positives kill trust
  - Abandon if <10% of 50 free scrape-report recipients show fix-intent or pay-intent
  - Abandon if LabelBlind ships an equivalent sub-₹10k self-serve PAYG tier before your v1

15. Next step — 1-week validation sprint

• Day 1–2: Pull 50 real back-of-pack label images from Amazon/Flipkart small-brand food and nutraceutical listings. Manually audit each against FSSAI Schedule II + claim rules with the help of a labelling consultant. This both builds the test set and quantifies the real non-compliance rate in the target segment.
• Day 3–4: Send each of the 50 founders a free, accurate one-page violation report (“found 4 issues on your label — happy to share the fixes”). Track replies, fix-intent and “would you pay for this before your next print run.”
• Day 5: Pitch 8–10 packaging printers and food-label design studios on a Studio account; gauge whether they’d run it on every client job.
• Decision: Go if ≥10% of the 50 founders show clear fix-intent/pay-intent and at least 2 printers/studios want a paid pilot. Both numbers are falsifiable — no “people seemed interested.”