Appearance
Growth model
🔵 Research. Assessment of the proposed acquisition engine.
The central claim
The product's distribution is supposed to be the product itself: a profile is shared to 10–30 people, some of them open it, some of those create their own, and the loop repeats. Formally, the viral coefficient:
K = (viewers per profile) × (viewer → registration) × (registration → published profile)If K ≥ 1, growth compounds and paid acquisition becomes optional. If K is 0.1, this is an ordinary product that needs marketing money it does not have.
The source analysis is right that this is the single most important number, and right that it should be a formal product KPI rather than a hope. It is also right that the 90-day question is not revenue:
"Can one published marriage profile reliably bring another person into QR Setu?"
That is the correct first experiment, and everything else is downstream of it.
⚠ The weak link, and the source analysis does not isolate it
The chain has three multiplied terms and the analysis treats them as roughly equal in risk. They are not. Viewers per profile is nearly guaranteed; viewer → creator is where the model lives or dies, and the reason is specific to this use case:
Most viewers of a marriage profile are not candidates to create one. A shared biodata reaches:
| Who receives it | Roughly | Could they create their own? |
|---|---|---|
| Parents, siblings, close relatives | the majority | ❌ No — they are not seeking marriage |
| Family friends, community contacts | several | ❌ Mostly no |
| Matchmakers / bureaus | 0–2 | ⚠ Different mechanism entirely (they create for others) |
| Prospective families | a few | ✅ Yes — and this is the whole pool |
So "15 viewers" is real, but the creator-eligible subset is perhaps 2 to 4 of them. Any model that multiplies 15 by a generic conversion rate will overstate K by roughly 4 to 7 times.
The redeeming nuance, and it is genuinely important: the eligible subset is also the highest-intent subset. A prospective family opening a proposal has a marriageable son or daughter by definition. They are not a cold audience; they are the exact target, arriving at the exact moment. So a low base rate applied to a very high-intent group can still work — but it means the CTA must be aimed at that viewer, and generic "create yours" copy shown to an uncle is wasted surface.
What follows practically:
- Measure creator-eligible viewers, not raw viewers, or the funnel will look healthier than it is.
- Place the create-CTA where intent concentrates — the interest and proposal-closed moments, not every footer.
- Expect K < 1 initially. A sustained K above 1 is rare and should be treated as an upside case, never as the plan.
Where the loop is architecturally strong
Two of the three preconditions are already true on this platform, which is unusual:
- No install to view. ADR-0019 mandates one DOM, server-rendered public renderer. The recipient opens a real web page. The source analysis's §32 warning — "if QR Setu says 'download our app first', your viral loop dies" — is already prevented by platform law.
- Rich share previews. OG/meta tags are already produced and driver-inspectable on the public card route today, so the WhatsApp preview is a content problem, not an engineering one.
- No login to view. Anonymous-first public reads already exist for the vendor card.
The one genuinely new mechanism is the CTA and its placement.
"Don't authenticate curiosity. Authenticate intent."
This principle from the source material is the best single line in it, and it matches CLAUDE.md's existing rule that "a signup wall in front of a scanned card destroys the platform's entire growth mechanic". View freely; ask for OTP only when someone expresses interest.
Channels, ranked with my read
| # | Channel | Assessment |
|---|---|---|
| 1 | WhatsApp (the product itself) | Correctly ranked first. The behaviour already exists; the product substitutes the artifact inside it. Nothing else in the plan has this property. |
| 2 | Matchmakers / marriage bureaus | Most underrated. One bureau can produce dozens of profiles, and ⚠ the architecture already exists — ADR-0022/0023 organizations, workspace trees and seat licensing. See Architecture fit Decision 4. The trap is ownership: a bureau-created profile must be member-owned with delegated editing, never org-owned, or the client loses control of their own identity. |
| 3 | Free biodata generator + SEO | Long-horizon and, per Reference landscape, table stakes rather than a growth hack — the incumbents already occupy these search terms with 50+ free templates. Entering costs real content work. |
| 4 | Family referral ("share with family") | Cheap, natural, but mostly produces viewers, not creators — see the weak link above. |
| 5 | Marathi micro-creators | The best use of the ₹2,000/month, because relevance beats reach at this budget. |
| 6 | Instagram paid | ⚠ Not an acquisition channel at ₹24,000/year. Correctly described in the source as a creative-testing budget. Treat any user it produces as incidental. |
On the ₹2,000/month budget
₹24,000 for a year is roughly ₹67/day. Against the Indian CPC ranges the source analysis cites, that is a handful of clicks daily. The analysis's own conclusion is the right one and worth preserving as a rule:
Only increase advertising once the funnel is known to work. If ₹2,000 produces 300 registrations and no published profiles, the defect is in the product, and ₹20,000 would buy 3,000 non-publishers instead of 300.
That is a disciplined position and I would keep it.
What the plan does not model
- Seasonality. Indian marriage activity follows the muhurat calendar — concentrated roughly November–February and April–June, with a near-dead Pitru Paksha period. Both acquisition and revenue will be lumpy. A flat monthly target will mislead in both directions: panic in a dead month, false confidence in a peak.
- Profile lifespan vs measurement window. A profile is active for months and then retires. Cohort retention curves borrowed from ordinary consumer apps will read as catastrophic churn when they are in fact successful outcomes. The analytics need to distinguish matched from abandoned from day one, or the numbers will be uninterpretable.
- Supply/demand asymmetry. Bride-side and groom-side biodata circulate differently, and the sharing networks are not symmetric. Worth measuring separately rather than assuming one funnel.
My assessment
The loop is well-designed and the platform removes its two hardest technical preconditions. The strategy of substituting the artifact inside an existing behaviour, rather than relocating the behaviour, is the strongest idea in the source material.
Two corrections I would make before anyone builds to this model:
- Instrument creator-eligible viewers, not viewers. Otherwise K is measured against a denominator four to seven times too large, and the first 90-day experiment returns a number that cannot be acted on.
- Plan for K < 1. Treat compounding growth as upside. Matchmaker distribution — a channel that multiplies without needing K ≥ 1, and which the platform can already express — is the more reliable path to the first ten thousand profiles.