Marketing & Business Brief — Answers to the Team’s 16 Questions

Living doc answering the marketing/business team’s 2026-06-25 questions. Owner: Salil. Sources of truth this synthesizes (do not duplicate, link): product ../product/features.md, ../product/roadmap.md, ../product/pricing.md; market market-analysis.md, gtm-plan.md; content content-seo-plan.md, customer-problems.md. INTERNAL — contains margins, COGS, competitor teardowns, and the “avoid” list. Not for public hosting without access control (see §0).

0. Hosting these on handbook.obsrv.in — built, with one condition

Yes, and it’s done. The back-and-forth of mailing docs is the problem; a git-backed handbook site means the team always reads handbook.obsrv.in and sees whatever we last pushed. No versions to chase. (Built as the handbook/ Nextra app; renders this docs/ tree.)

The condition: it must be private. This folder holds unit economics, competitor teardowns, the org-types-to-avoid list, and pricing strategy. A public site leaks our playbook to the exact competitors named below. So:

The domain split: handbook.obsrv.in is this private GTM/strategy/product space behind auth. That keeps docs.obsrv.in free for a public docs surface later (API reference, help guides) and the blog ships through the public web app. To go live: add the handbook.obsrv.in DNS record and turn on Vercel password protection (5-min jobs on your side).

1. Feature → business-value map (per feature)

How each shipped capability converts to money/risk outcomes for the buyer. (Status from features.md; ✅ built · 🟡 partial · 🔲 roadmap.)

One-line pitch this map supports: “Verify the statement, consolidate the borrower, decide against your policy, file the audit trail — every number reconciled, the decision stays yours.”

2. What’s actionable after analysis — and is the roadmap aligned?

What the user can do the moment a report exists (all shipped ✅):

1. Run a decision — enter amount/tenure/product → eligibility sizing + approve / approve-with-conditions / counter-offer / refer / decline, with reasons tied to the computed signals.
2. Send to the approvals queue — refers/declines route to a queue; an authorized user (L1/L2/L3) claims, resolves or escalates, with reason codes.
3. Override — within authority; hard-blocks (KYC/fraud/tamper) are never overridable.
4. Export — CSV of the queue (ask + financials + engine call + resolution) for the audit file/LOS; PDF report for the credit committee.
5. Consolidate — group into a Borrower Case for one combined income/FOIR view.
6. Resume — finish a partially-processed statement after topping up credits.

Roadmap alignment: yes, strongly. The roadmap’s thesis is explicitly “we process a statement; the customer’s job is to underwrite a borrower,” and the owned arc is Collect → Verify → Consolidate → Decide → File/Audit. Post-analysis actionability is the “Decide” and “File” half, and it’s built. The gaps that would make post-analysis even more actionable (all on the roadmap, see roadmap.md):

• Case-level decision as the primary unit (decide on the consolidated case, not per statement) — the next build.
• Decision-packet PDF (consolidated report + decision record + policy version as one filed artifact) — 🟡.
• Borrower intake link — fills the case at the front of the workflow — 🔲.

Verdict for marketing: we can confidently market “from statement to decision to audit trail” today; we should not yet over-claim a one-click consolidated-case decision packet.

3. ICP and India-specific subtypes to target directly

Primary ICP: small-to-mid Indian lenders who underwrite bank statements manually and can’t justify an enterprise contract. Ranked subtypes:

Sharpest beachhead: small NBFCs + DSAs. They self-serve, feel the pain weekly, and don’t need AA or an enterprise suite to get value on day one.

4. Org types to strictly stay away from

5. International feasibility & the easiest first country

Feasibility: the engine travels; the domain layer is local. The reconciliation gate, deterministic math, extraction, scoring mechanism, and the credit-pack billing are locale-agnostic. What’s India-specific and would need localizing:

• Domain policy — FOIR norms, income bands, segment thresholds (policy.py).
• Bounce/channel taxonomy — NACH/ECS/UPI/IMPS are Indian rails; other countries have ACH/SEPA/Faster Payments/etc.
• Bank-statement formats — coverage is per-bank.
• Currency, date formats, number formats.
• Compliance — data residency + local lending/data law (the real gating cost).

Easiest country to launch with the least compliance handling: UAE / GCC.

Runners-up: Singapore (English, but PDPA + small market), Philippines/SE Asia (big thin-file lending, but localization + data law work). Avoid first: US/UK (huge but crowded and compliance-heavy), EU (GDPR). Recommendation: treat international as post-PMF; if we test one, UAE via the existing Dodo USD rail is the lowest-compliance probe — localize the domain policy + bounce taxonomy first.

6. B2B / B2C classification & category

• It is B2B SaaS — sold to lending organizations, not consumers. More precisely product-led-growth (PLG), self-serve B2B SaaS, with a B2B2C touch once the borrower intake link ships (the borrower uploads, the lender consumes).
• Category: Bank Statement Analysis (BSA) / cash-flow underwriting. Adjacent labels buyers search: financial document intelligence, transaction enrichment, bank statement analyzer/analyser, automated underwriting. See market-analysis.md.
• Our sub-position within it: the trust-first, self-serve, decision-owning BSA — “the verifiable underwriting workspace for a borrower.”

7. Migration from a competitor’s portal — challenges & whether we’re configured

Reframe first: our wedge is net-new self-serve users, not ripping out enterprise contracts. Most realistic “migration” is parallel-run — a lender runs Obsrv alongside their current tool for a few weeks and compares — not a data lift-and-shift. With that lens:

What “configured for migration” would mean for us: (a) re-analyze from the borrower’s source statements (we do this — no proprietary import needed), (b) a policy editor to mirror their FOIR/ score rules (we have this — ScoringPolicy + decision policy), (c) bulk upload + API (we have this), (d) a historical-report importer (we do not have, and it’s competitor-format-specific).

Build it or not? A historical-report importer is off-roadmap and low-ROI — it’s bespoke per competitor and our re-analyze-from-source path already gives a clean, verifiable baseline (which is on-brand: we’d rather show our reconciled numbers than trust theirs). Recommendation: invest in a smooth parallel-run / bulk-upload onboarding + policy-mirroring, not in importing competitor data. That stays on the roadmap (self-serve, verifiable) instead of pulling us toward bespoke enterprise migration work.

8. Competitor cost breakdown & the value we provide against each

Pricing is opaque for most players (enterprise “contact us”); figures flagged as estimates in market-analysis.md. Full table there.

The value sentence: “Everyone either hides the price behind a sales call, or sells OCR without the decision. We’re the only one that’s self-serve, transparently priced, and goes from a verified statement to a decision you can file.”

9. India market analysis (summary — full doc linked)

Full living analysis: market-analysis.md. Headlines:

• The category is core to every Indian digital-lending stack and increasingly bought by NBFCs, DSAs, fintech lenders, and CA firms.
• The market is enterprise-sales-gated and opaque — Perfios, CRIF, Signzy, ScoreMe, Ocrolus all hide pricing behind “book a demo.” Transparent self-serve is rare (Precisa, Pro Analyser).
• The long tail is underserved — individual CA firms, DSAs, small NBFCs can’t justify an enterprise contract. That’s our lane.
• Account Aggregator (~120M accounts linked by Dec 2024) changes ingestion, not the need for analysis, and covers only ~38% of borrowers — PDF-first analysis stays relevant for years.
• Fraud is the fastest-growing battleground — AI document fraud up ~5× (2025); bank statements are 59% of flagged fraud docs. This is why tamper detection is our #2 roadmap item (§10).

10. Fraud-detection models we’ve adopted (be precise — don’t over-claim)

Shipped, fraud-relevant signals (✅):

• Dual reconciliation gate — every row vs running balance and credit/debit column totals; a doctored balance or inflated total cannot pass silently. (This is our strongest anti-tamper today.)
• Deterministic money math — the AI only transcribes; all totals/FOIR/score are computed by auditable code, so inflated figures don’t survive.
• Document-integrity checks — math consistency, duplicate rows, date-sequence breaks.
• Risk flags — NACH/ECS bounces, cheque returns, penal charges, circular transfers, high cash deposits, large one-off credits, negative-balance days.
• needs_review gating — anything uncertain is held for human review, never silently passed.
• Adversarial / prompt-injection test harness — guards the extraction layer against statements crafted to manipulate the model.
• Confidence scoring per extraction.

NOT yet shipped (🔲 roadmap #2 — marketing must not claim it):

• Tamper / forgery forensics: PDF metadata/producer-chain analysis, font & object inconsistency detection, recomputed-balance tampering, OCR-layer-vs-text mismatch, digital- signature validation. This is the #1 gap on our own scorecard vs Precisa/Perfios/Ocrolus/ Inscribe and the next trust feature.

Marketing line that is true today: “Every number is reconciled and computed deterministically, so doctored totals and balances don’t pass — and anything uncertain is flagged for a human, never silently approved.” Do not yet say “forgery/tamper detection.”

11. Expected credit usage by ICP (predictions + worked examples)

1 credit = 1 page = ₹5. CSV: 40 rows = 1 page. Rough page counts: a single statement ≈ 3–6 pp; a 6-month single-account pull ≈ 40–70 pp; a multi-account borrower case ≈ 70–150 pp.

Takeaway for marketing & pricing: DSAs and CA firms sit comfortably inside the current packs; NBFCs and fintechs will exhaust ₹2,000 packs fast — which is exactly why auto-recharge and larger ₹5k/₹10k packs are flagged in pricing.md as the next billing build. Lead heavy users toward CSV uploads (near-total margin for us, same ₹5 to them).

12. Partial processing — already specced; confirm the “limited output” messaging

This is built and specced (see pricing.md §Partial processing; features.md “partial+resume ✅”). The rule the team is asking for is already the design:

1. If credits cover only N of M pages, process the first N, debit N — never reject (except at 0 balance).
2. Total transparency — the job, report, UI, and PDF all state “N of M pages processed; M−N remaining,” and analytics/verdict/decision carry an explicit partial-data warning (income/FOIR from partial months can mislead a decision — the report says so loudly).
3. Resume after recharge — only partially-processed docs offer “process the remaining pages”; raw file retained 7 days (disclosed) for this.

Action items for marketing/product to verify in the live UI (so the team’s requirement is demonstrably met):

• The “limited analysis output” banner is prominent on a partial report (not just a quiet flag).
• Low-credit nudges are live (currently 🟡 in features.md — balance shows, nudges pending).
• The buy-credits CTA appears at the moment of truncation (402 → “buy credits”).

Messaging principle to enforce everywhere (matches the no-em-dash, plain-copy rule): “We scanned the pages your credits covered. Add credits to analyse the rest; the report below is based on a partial statement.”

13. SEO keywords already shipped + new targets after the ICP work

Shipped today (we use no keywords meta tag — deprecated/ignored by Google; the real surface is titles, descriptions, llms.txt, FAQ, and JSON-LD):

• Core: bank statement analysis, bank statement intelligence for lenders, underwrite a bank statement (in seconds/60 seconds), bank statement analyzer.
• Feature/signal: income, obligations, FOIR, NACH/ECS bounces, risk flags, top counterparties, integrity checks, every number reconciled against the running balance.
• ICP: for NBFCs, DSAs, and lending teams, small and mid-size Indian lenders.
• Pricing/model: ₹5 a page, prepaid credits, no subscriptions/seat fees, free trial page, partial processing.
• Differentiator: borrower case / consolidation, inter-account transfer dedup, consolidated FOIR, deterministic / verifiable / reconciled, indicative aid, not a credit decision.
• Structured data shipped: Organization, SoftwareApplication (FinanceApplication, ₹5 offer), FAQPage JSON-LD; robots.txt, sitemap.xml (only /, /terms, /privacy), llms.txt.

Gaps in shipped SEO: static sitemap (no content URLs), no per-page keyword/topic pages, no blog, no OG images, no BlogPosting/BreadcrumbList schema.

New targets to add after the ICP work (full long/short-tail bank in customer-problems.md; blog mapping in content-seo-plan.md):

• Short-tail (high volume): bank statement analysis, FOIR calculation, fake bank statement, bank statement analyser, automated underwriting, cash flow underwriting.
• Long-tail (high intent, by ICP): how to calculate FOIR for home loan, how to detect tampered bank statement, bank statement analysis software for NBFC, bank statement check for loan DSA, self-employed income assessment from bank statement, account aggregator vs PDF statement, Precisa alternative, bank statement analysis API India, NACH bounce meaning, best bank statement analysis software India.

14. 500 customer problems & most-searched queries

Delivered as its own bank: customer-problems.md — organized by ICP × journey stage × theme, each entry phrased the way the customer actually searches/complains, tagged to the feature or blog that answers it. (See that doc for the count and how it’s structured to extend to the full target.)

15. 10 blog titles, subtitles & bodies with long/short-tail keywords

The titles, hooks, subtitles, primary + AEO keywords, and outlines for all 10 launch posts live in content-seo-plan.md (pillar + cluster, mapped to ICP and roadmap). Full body copy is the writing phase — each post drafted with its keyword set woven in. I can draft the pillar post (#1) end-to-end as the template the other nine follow; say the word and I’ll add full bodies into the content doc.

16. LinkedIn public figures from competitor companies

Delivered as its own sourced doc: competitor-people.md — founders, CXOs, and active “thought-leader” voices across the India and global competitors, with roles, LinkedIn handles (tagged confirmed vs inferred), posting themes, and a “who to follow first” watchlist, dated 2026-06-25.

Two corrections that came out of the research, worth knowing:

• Precisa is not a standalone company — it’s the BSA product of Finezza, so its “people” are Finezza’s founders. (Update mental models accordingly; market-analysis.md still lists Precisa as the closest competitor, which is correct at the product level.)
• Pro Analyser has a brand-only public presence — no founder/exec is named anywhere public, so there’s no person to track there (flagged, not fabricated).
• Some named execs have moved on (e.g. departures at FinBox and CRIF) — flagged in the doc; confirm before any outreach.

What’s done here vs. what needs your go-ahead

• Answered now (Q1–Q13): in this doc, grounded in our product/market/pricing docs.
• Q14: see customer-problems.md.
• Q15 (full blog bodies): plan is in content-seo-plan.md; bodies on request.
• Q16 (competitor people): needs a verified web-research pass — on request.
• Hosting (§0): built as the private Nextra handbook.obsrv.in (add DNS + turn on Vercel auth to go live).