AyByay is the Bangla-first, mobile-only operating system for Bangladesh's paper-run shops — billing, stock and the baki ledger on the phone in the owner's pocket — and AI does the data entry that has killed every prior attempt to put these shops on software. We are pre-revenue by design, raising a milestone-gated seed to answer the only two questions worth money: will a shop pay, and will it stay?
A torn khata for credit, stock in the owner's head, profit a feeling, an hour of midnight reconciliation after fourteen hours on his feet. The software to fix this has existed for years. The reason it never stuck is not price or features — it is data entry: typing a shop's catalog in by hand, or paying someone ৳10–20k to do it. AI is the first tool that can delete that wall — and 2024's collapse in inference cost is the moment it became affordable.
We delete data entry from two directions — one proven live, one being proven on camera this quarter: a computerized shop's whole catalog migrates in one AI-mapped session; a paper shop builds his by photographing purchase memos and scanning barcodes against a shared catalog. We headline what is proven and flag what is promised.
One definition throughout — active = at least one sale in the trailing 7 days. The last rung is the whole reason to fund the round.
Two internal tenants (demo, test) excluded. The only rows in the payments table are two SSLCommerz sandbox ৳299 test charges on an empty shop — integration testing, not revenue. We show you that so you don't wonder what else was hidden.
Built by hand at the counter, day after day.
Active 17 of 30 days in July — front-loaded, honest gaps. "17 of 30" out-credibles "daily."
Baki, precisely: 3 customers carrying ৳585 in active credit (of 21 customers on file). One stored SQL definition, reproducible live.
A PrimeERP export, AI-mapped in one session — the migration proof. Not a daily-use shop.
banbro is the one-session migration proof (DokanePOS's turf, not our wedge): 15 sales over 7 active days, then quiet. We describe it as onboarding proof and nothing more.
Reconciling the macro number with a 400–600-shop seed target.
Feature breadth is the monetization ladder. Economics assume the list schedule; Basic runs at a ৳99 launch price today — a step we disclose, not hide.
14-day free trial, no card · pay via bKash / Nagad. Switching cost accrues as the shop's whole history lives here.
The category is fragmented, not empty. TallyKhata won the ledger — but it deliberately stops at the khata. The full shop OS — POS + stock + baki + AI onboarding + offline, in Bangla, on the shopkeeper's phone — is still up for grabs, split between 1998-era desktop POS and sub-scale mobile challengers. AyByay is built to win that layer.
| Job to be done | AyByay | TallyKhata | Desktop POS | banijjo | Hishabee |
|---|---|---|---|---|---|
| Whole shop from the phone | ✓ | ✓ | — | ✓ | ✓ |
| Baki ledger in Bangla | ✓ | ✓ | ~ | ✓ | ✓ |
| Barcode POS + live stock | ✓ | — | ✓ | ✓ | ✓ |
| AI does the data entry | ✓ | — | — | ~ | — |
| Works offline through load-shedding | ✓ | — | ~ | — | — |
| 5M-listing BD catalog head-start | ✓ | — | — | — | — |
TallyKhata reports ~9.7M downloads / 5.5M businesses — proof of BD adoption, and proof it won't build the counter itself. banijjo is a real, contemporaneous same-category rival (an old 2012 domain repurposed into a 2025–26 SaaS), pre-traction (~100+ app installs) — not an earlier winner. Our edge over it is concrete: offline-first (banijjo advertises none), AI product-onboarding + the 5M-listing catalog head-start, and a fuller shipped stack. The 5M catalog is a public+licensed head start (~84% public scrape) — not itself a moat; the moat is the correction loop live counters write on top.
No khata or shop-OS company in South Asia has yet returned venture capital. India proved shopkeepers adopt this explosively (Khatabook, OkCredit) and proved free-love doesn't convert. Vyapar is the closest survivor — private, roughly profitable on about a third of Khatabook's capital, by charging for the whole shop OS from day one. That is the model we run — in Bangla, on mobile, with an AI onboarding weapon the Indian cohort never had.
From activation 20–35% × free→paid 10–20% = 2–7% net conversion at ৳300–600 a visit — a ৳4,286 floor. We publish the real band, not a number our own funnel refutes.
Defensible stretch target: ≤ ৳5,000, contingent on pushing net conversion toward the 7% ceiling.
Net LTV ৳2,360–4,720 (net of ~21% commission — never the gross band under a "net" label) against that CAC.
Ratio ~0.1–1.1× — below break-even in all but the best corner. Moving it above 1.5× (annual prepay lifting ARPU, hardware margin, conversion ≥5%) is exactly what the seed buys. This is the honest reason to fund the round, not a red flag we hide.
Agent-network line separates a costed hardware attach and 1–2 field-lead salaries from the ~৳500-per-prepay commission pool — the channel can't manufacture fake CAC. Retention is measured at 4/8/12 weeks, not assumed at month six; the structure is the standard offshore holdco + BD operating subsidiary.
Miss them and we spent very little finding out. That is the whole bet — a founder who did the math against himself, in Bangla, on mobile, with the one weapon the graveyard never had.
DRAFT — pre-evidence-sprint — not for external distribution. Every figure here is reproducible on demand from the live tenant database (evidence.sql) or a cited public source.