Skip to Content
Blog (drafts)Style guide

Blog Style Guide — read this before drafting any post

The shared contract every Obsrv blog post follows. Owner: Salil. Post briefs (titles, hooks, keywords, outlines): ../go-to-market/content-seo-plan.md. Product facts below are the source of truth for any claim — do not exceed them.

Who we write for

Indian lending professionals: small/mid NBFC underwriters and credit heads, DSAs / loan connectors, CA / audit firms, early fintech lenders. They underwrite bank statements, often manually in Excel, and are cost- and time-pressured. Write to a sharp practitioner, not a layperson.

Voice

  • Precise, ledger-like, credible. Show the math. Confident, never hypey.
  • Plain punctuation. No em dashes in any post (these are customer-facing). Use commas, periods, parentheses, or rewrite. (Internal docs may use em dashes; published copy may not.)
  • Indian context and terms: ₹, NBFC, DSA, NACH, ECS, FOIR, UPI/NEFT/RTGS/IMPS, Account Aggregator, CIBIL, RBI, LOS. Indian number formats where natural (lakh/crore) but keep digits clear.
  • Second person (“you”), active voice. Short paragraphs.

Product facts (accurate claims only — DO NOT overclaim)

  • Obsrv: self-serve bank statement analysis for Indian lenders. Upload a statement (PDF or CSV), get a decision-ready underwriting report in about a minute.
  • Pricing: 1 page = 1 credit = ₹5; CSV = 40 rows per page. Prepaid credit packs, no subscriptions, no seat fees, no sales call. Free trial page on signup.
  • What it extracts: income (salary/business/rental/interest/govt), obligations + FOIR + disposable income, NACH/ECS bounces, cheque returns, penal charges, top counterparties, channel mix, balance analytics (AMB/ADB, negative days), spend categories.
  • Trust architecture: the AI only transcribes; all money math is deterministic, computed by auditable code. Every row is reconciled against the running balance AND column totals (dual reconciliation gate). The risk score is deterministic and reproducible (same statement, same score), never the AI’s guess. Uncertain items are flagged for human review, never silently passed.
  • Borrower Case: consolidate a borrower’s multiple statements/accounts into one report, with inter-account self-transfers detected and netted out (auditable, with confidence scores).
  • Decision support: the engine sizes eligibility and recommends approve / approve-with-conditions / counter-offer / refer / decline against the lender’s OWN policy. A human owns the decision; we are the engine, not the credit decision. Override/approvals queue with delegated authority and an append-only audit trail; CSV export for the audit file / LOS.
  • Partial processing: if credits cover only part of a statement, we process that part, label it clearly, and let you resume after topping up.
  • Data: processed in India (AWS ap-south-1); raw files deleted after analysis by default; optional retention vault. DPDP-aware posture.

CRITICAL — do NOT claim these (not built yet, roadmap):

  • Tamper / forgery detection (PDF metadata/font forensics, digital-signature validation) is NOT shipped. For the fraud post: position Obsrv on what IS true today, namely the dual reconciliation gate + deterministic math (doctored totals/balances cannot pass silently), circular-transfer / high-cash / large-one-off flags, and needs-review gating. You may say tamper forensics is “on our roadmap.” NEVER say we currently detect forged/tampered documents.
  • Account Aggregator ingestion is roadmap, not shipped. We are PDF/CSV-first today.
  • Do not invent statistics. If you cite an industry stat (e.g. AA ~120M accounts, ~38% coverage; AI document fraud up ~5x; bank statements ~59% of flagged fraud docs), keep it directional and attributable; prefer framing over hard numbers you cannot source.

Structure every post must follow

  1. Frontmatter (YAML): 
    ---
title: "<SEO title, <=60 chars where possible>"
description: "<meta description, 140-160 chars, includes primary keyword>"
date: 2026-06-25
author: "Obsrv"
slug: "<kebab-slug>"
primaryKeyword: "<one>"
keywords: ["<long/short tail list from the brief>"]
---
  2. H1 = the article title (compelling, includes primary keyword).
  3. Opening answer (AEO): 2-3 sentences that directly answer the title’s question up top, before any preamble. Answer engines lift this.
  4. Question-shaped H2 sections (e.g. “How is FOIR calculated?”). Each section opens with a 1-2 sentence direct answer, then detail. Use tables and numbered/bulleted lists liberally (AEO favors them). Include worked examples with real ₹ numbers where relevant.
  5. A comparison or formula or checklist box where the topic supports it.
  6. FAQ section (”## Frequently asked questions”) with 4-6 Q&A pairs in ### Question? + answer format (this maps to FAQPage schema later). Use the secondary/AEO keywords here.
  7. Internal links: link to the pillar post (what-is-bank-statement-analysis) and 1-2 sibling posts where relevant, using relative markdown links like [FOIR](./how-to-calculate-foir.md).
  8. Soft CTA at the end: one short paragraph on how Obsrv does this automatically (₹5/page, self-serve, every number reconciled), with a link to https://obsrv.in . Never pushy.

Keyword integration

Weave the post’s primaryKeyword into the H1, the opening answer, at least one H2, and the meta description. Distribute the long/short-tail keywords from the brief naturally across H2s, lists, and the FAQ. Do not keyword-stuff; readability first.

Length & quality

~1,500-2,200 words. Genuinely useful to a practitioner (it should teach, not just sell). Accurate, specific, and citable. Indian examples throughout.

Overview1. What is BSA (pillar)