B2B SME app marketing strategy: launching a business app for small businesses in India

India's SME software market is real and structurally unprofitable at current acquisition costs. Vyapar, the category's mobile-first leader, posted FY25 revenue of ₹69 crore (about $7.8 million), up 53 percent, against a net loss of ₹63 crore (about $7.2 million). It spent ₹2.04 to earn ₹1 of operating revenue. That is the fact that should shape your plan.

Dig one level into that filing, reported by Entrackr, and the shape becomes clear. Total expenses were ₹141 crore (about $16 million), of which employee costs were ₹102 crore — 72 percent. Cash reserves fell 93 percent to ₹6 crore (about $682,000).

Read that cost structure carefully, because it contradicts how the category is usually described. This is not a low-touch, self-serve software business in India. It is a sales-and-support business with an app attached. A launch plan that models India SME B2B as app-store-led self-serve acquisition is modelling something the category leader could not make work — and the budget section of this post therefore buys people before it buys media.

The competitive set:

The ledger apps are the cautionary tale. Khatabook and OkCredit acquired enormous SMB user bases on a free digital-ledger hook, then struggled to convert them; the intended lending monetisation never scaled to justify the valuations. Khatabook's public pricing is now "price on request" — what a self-serve product looks like after it becomes sales-led. Free-utility-to-fintech is a failed pattern in India SME, not an aspirational one. The winners charge for the software.

For contrast only: the US market is mature, expensive and web-primary — QuickBooks, Xero, FreshBooks, Wave, plus vertical players like Jobber and Gusto — with mobile as a companion rather than the product, and ticket sizes an order of magnitude above India's. That gap is why US B2B advice about "high LTV" does not transfer, and why this post is India-only.

You are acquiring one person: the owner-operator, usually on a mid-range Android phone that is the only computer the business has. The activation event is the first invoice created and sent — not previewed. Someone who taps through creation has not activated; someone who sends a real invoice to a real customer has entered their revenue workflow.

That distinction is the whole funnel:

Install → Open
  → Business profile setup (name, GSTIN, logo, address)
        ⚠ BIGGEST DROP-OFF: GSTIN entry is a hard gate. The owner
          must fetch a number they may not have to hand.
  → First item or customer added
  → ★ ACTIVATION: first invoice created AND sent
        (WhatsApp is the dominant delivery channel in India)
  → Second invoice within 7 days        ← the habit signal
  → Payment recorded against an invoice ← the loop closes
  → ≥5 invoices/week                    ← real adoption
  → Feature-limit collision             ← the paywall trigger
  → GST filing period arrives           ← seasonal conversion spike
  → Subscribe → renew, year after year  ← where the value actually is

Three things about this ladder have no consumer equivalent.

Monetisation is limit-driven, not feature-driven. Every price ladder here is gated on invoice counts, GSTIN counts and user seats, so your paywall fires when a customer's business grows. Your conversion rate is partly a function of your users' growth rather than your persuasion, and your revenue per customer grows the same way — through seats and GSTINs added over years, not through a bigger first purchase. Choose limits that bind at the point the customer can afford you.

There is a calendar. GST return deadlines and the 31 March financial year-end create predictable demand spikes. Build your launch and creative calendar around GSTR filing dates, not around generic consumer seasonality.

Switching cost compounds fast and then locks. Once a business has a quarter of invoices in your app, migration is painful, so the window to win a customer is narrow: business formation, or a GST-threshold crossing. Target the trigger event, not the category.

The retention question behind all of this is whether you become the system of record. A tool that is not where the books live is a tool the owner has to remember to open, and it competes with WhatsApp, a paper ledger and the accountant's spreadsheet — losing quietly over about a quarter. Consumer apps churn when the user loses interest; B2B SME apps churn when a parallel record survives alongside yours. If the owner still writes the day's sales in a notebook and enters some of them in your app, you are a duplicate, and duplicates get dropped at the first busy week.

The practical test: would the business suffer a real consequence if your app vanished tomorrow? If they would lose their invoice history, GST filing trail and outstanding-payments list, you are the system of record. If they would lose "a nicer way to make invoices," you are not. So instrument record depth, not sessions — invoices stored, customers stored, months of continuous history.

On retention benchmarks, take the honest answer rather than a borrowed one. The figures usually quoted here — Admiral Media's Business and Productivity panel at D1 24 to 32 percent, D7 10 to 16 percent, D30 5 to 10 percent — are Tier-1 iOS. Wrong platform, wrong geography, and an installed-base curve rather than an activated-user one. They are not a target and this post will not pretend otherwise. Instead: report retention on activated users only, because a blended figure describes your GSTIN gate rather than your product, and target D90 far above any consumer curve — a business that has put a quarter of its invoices in your app and has not left is a customer, and if your activated D90 looks like a consumer app's D30, the product has not become the system of record. Measure it against your own first cohort. Do not benchmark it against anyone.

Four things: an onboarding path that survives GSTIN entry, a Play production-access clock that has already started, an event taxonomy built around business records rather than screens, and a decision about whether you will ever touch money. That last one determines which regulatory regime you are launching into, and it cannot be reversed cheaply.

GSTIN entry is your single biggest product risk — a hard gate at the exact moment your user has the least commitment. Build the mitigations before launch: allow full invoice creation without a GSTIN and prompt at send time, offer GSTIN lookup by business name, and let the user save a draft profile and return. Each moves the drop-off later in the funnel, where you have more evidence to work with.

Start the Play production-access clock early. Personal Play Console accounts created after 13 November 2023 need 12 testers running a closed test for 14 continuous days before production access is granted, with review typically inside seven days after that. Organisation accounts are exempt. Testers who opt in for fewer than 14 days and then opt out do not count, and if a tester opts out and returns, the clock restarts. Twelve real small-business testers are harder to recruit than twelve consumers, and they are also your first onboarding research — start recruiting them a month before you want to launch. Older guides saying 20 testers are out of date.

Event taxonomy. Instrument these, and keep them distinct:

Mark invoice_sent and payment_recorded as key events in Firebase, and note the ceiling: Google's guidance is not to select more than one action for target CPA, so you get one bid target, and here it should be invoice_sent.

Decide about money now. The natural monetisation path for a billing app is "we can see your receivables, let us finance them" — invoice discounting, working-capital credit, inventory BNPL. That single feature converts your business-tools app into a regulated lending app under Google Play policy, and it is precisely the turn Khatabook took. The policy detail is below; the point here is that this is an architecture decision, not a roadmap decision.

Treat GST e-invoicing as your strongest, most honest marketing hook, not a compliance chore. E-invoicing is mandatory above ₹5 crore (about $570,000) of annual aggregate turnover, and the threshold is cumulative and permanent: crossed in any financial year since 2017-18, even once, and the business is in scope forever. Most affected owners do not know this.

Verify the thresholds against cbic-gst.gov.in and einvoice.gst.gov.in before publishing any of them — the summary here comes from consultancy reporting corroborated in outline, not from the primary notification. The rules that matter commercially: businesses at or above ₹10 crore (about $1.14 million) aggregate turnover must report invoices to the invoice registration portal within 30 days of issue, and the portal hard-rejects after that with no grace period. Scope covers B2B supplies, exports and supplies to government, and turnover aggregates across all GSTINs under one PAN. Without a valid IRN and QR code the invoice is legally invalid, and the buyer loses input tax credit.

B2C retail is exempt from IRN — but not from QR codes, and that distinction is a product requirement. Businesses above the relevant turnover threshold must display a dynamic QR code on B2C invoices, enabling digital payment against that specific invoice. So a retail-facing customer who is out of scope for e-invoicing may still be in scope for dynamic QR, and a billing app that ships IRN but not dynamic QR has a gap exactly where the retail vertical lives. Confirm the current threshold and the exemption list against the primary notifications before you build a marketing claim on it.

That input-tax-credit consequence is the marketing lever, and it is much stronger than the penalty. Your customer's non-compliance hurts their customer. A missing input tax credit is a relationship problem with a buyer who may not place the next order — far more motivating than an abstract fine. Build creative and content around that, not around "stay compliant."

The "crossed it once, in scope forever" rule creates a large population of businesses who do not know it applies to them. That is a content wedge with genuine search demand and very little honest competition.

One warning to take seriously. If you market GST compliance and your IRN generation fails, you have caused a customer's customer to lose input tax credit. Ship compliance features conservatively, disclaim clearly, and monitor IRN success rates before you build marketing on them. The fintech and payment app playbook covers the adjacent regulated-product discipline.

Two gates sit between your keyword list and your product, and both are procurement items with lead times rather than engineering tasks.

You cannot file returns or generate IRNs by yourself. Access to the GST Network for filing and e-invoicing runs through the GST Suvidha Provider and Application Service Provider framework: a GSP holds the licensed connection to GSTN, and an ASP builds on top of it. Practically you either become a GSP, which means an application, an agreement, fees and sandbox certification, or you buy access from one, which means a commercial contract, per-transaction pricing and a dependency on their uptime. Either path takes weeks to months and has a cost per invoice attached.

That has a direct marketing consequence that most keyword plans miss. The two highest-intent terms in this category — "gst return filing app" and "e invoice generator" — describe functionality you cannot legally deliver without that access. Do not build ASO, ad copy or landing pages around them until the GSP or ASP contract is signed and the sandbox has passed. Ranking for a promise you cannot fulfil generates installs that churn at the first attempted filing, which is the most expensive kind of install in this category.

DPDP applies to you with more force than it does to most consumer apps. India's Digital Personal Data Protection Rules were notified in November 2025, with substantive obligations enforceable on 14 May 2027. An invoicing app holds PAN, GSTIN and bank details for the business, and — the part people forget — the full customer contact list of that business, including individuals who are not your users and never dealt with you. Notice, purpose limitation, retention periods and erasure rights attach to all of it.

What that means before launch, not after:

• A retention period you can state and enforce, plus a working deletion path that survives a business closing its account and asking for its customers' data to be removed.
• Notice in English and the user's own language, matching your Play Data Safety declaration exactly. SME apps routinely collect GSTIN, PAN and bank details, and every one of those is a Data Safety line item; a mismatch between the declaration and the code is an enforcement trigger on its own.
• A clear internal answer to who the data principal is when the data is a business's customer list. Take Indian counsel on that before launch rather than after — this is a summary of an obligation, not legal advice, and the answer depends on your data flows.

Your listing has two jobs that consumer listings do not: it must survive Google Play's Financial Services policy if you touch money at all, and it must rank for sub-vertical intent, because "restaurant billing app" converts many times better than "billing app."

Start with the policy question, because it can stop the launch. Google Play's Financial Services policy triggers the moment your app moves money, offers credit or facilitates lending, at which point you must complete the Financial features declaration. For personal loans in India, Play requires a valid RBI financial services licence and presence on RBI's public list of digital lending apps deployed by regulated entities; new apps must be RBI-listed before first publication, in force since 30 October 2025, and the deadline for existing apps was 28 January 2026. Facilitators routing to partner NBFCs must still be on the RBI list and name every partner in the app description. Enforcement is not theoretical — press reporting puts Google's action at 3,500-plus loan apps in India.

A pure GST-invoicing app is out of scope. A GST-invoicing app with an embedded credit offer is not, and the second needs a completely different launch timeline.

The same logic reaches your ad accounts. Google Ads requires financial services advertisers in India to verify through a third-party vendor, demonstrating they are licensed or exempt, with business details matching the registry records exactly. Two things are widely missed. The net catches non-financial advertisers who target financial-services audiences, so a billing app advertising "manage your business finances" against finance-intent audiences may be pulled in. And an entity-name mismatch between your Play listing, GST registration and ads account will fail verification — align naming before you start, and budget four to eight weeks if there is any chance you are in scope.

No verified keyword volume data exists publicly here. The lists below are informed inference — validate in Play Console search-terms reporting first.

India, Play, head terms: gst billing app, billing software, invoice app, gst invoice, bill banane wala app, billing app for small business, e invoice, khata book, invoice maker, tally alternative.

India, sub-vertical long tail — where the conversion actually is: retail billing software, restaurant billing app, medical store billing software, kirana billing app.

India, gated on GSP or ASP access: gst return filing app, e invoice generator. High intent, and off-limits until you can actually file.

Sub-vertical terms carry far higher intent than generic ones, because this buyer self-identifies by trade rather than by "small business." Custom store listings per vertical are genuinely valuable — but routing paid traffic to them is not free, and the next section prices it.

What does it actually cost to run vertical store listings?

More than most plans assume, because of how App Campaigns are built. App campaigns for Installs have no keyword-level ad groups, so "point vertical ad groups at vertical listings" is not a thing you can configure. Custom store listing selection happens at campaign level. One vertical listing means one campaign, and every campaign has to clear the target-CPI floor on its own.

Run the arithmetic at an ₹80 planning CPI:

That is the floor, not a working budget — a campaign sitting exactly on its floor gathers conversions too slowly to clear Google's 100-conversion learning threshold in a useful timeframe. Give each campaign roughly twice its floor and a five-way vertical split needs about ₹12 lakh a month of Google spend alone.

This directly contradicts the consolidation advice later in this post, and both are right at different budgets. So state the threshold plainly:

• Below about ₹5 lakh a month: one consolidated campaign, one custom store listing aimed at your largest vertical, and vertical differentiation carried entirely by creative assets inside that campaign.
• ₹5 to ₹12 lakh a month: two campaigns at most — your strongest vertical, and everything else.
• Above about ₹12 lakh a month: a three-to-five-way vertical split becomes affordable, and the intent lift is worth the fragmentation.

Custom product pages on iOS behave differently and are cheaper to exploit: up to 70 per app, search-results ad groups only, one active custom ad per ad group, and no new app version needed. On Apple Ads the vertical routing genuinely does work at ad-group level. It is just that on an India-primary plan, iOS is a few percent of the audience.

Put people in the budget from the first month. The category leader spends 72 percent of its costs on employees, and this post has already argued that means humans close and retain these customers. A validation budget that is 100 percent media is testing the wrong hypothesis.

The single highest-return activity in this category does not appear in any media plan: call every business that entered a GSTIN and did not send an invoice within 48 hours. That user has told you they are serious, has hit the gate you already know is your biggest drop-off, and is reachable by phone because they gave you a number to verify. Nothing you can buy converts as well as that call.

Three capabilities, all of which belong in the Validation tier:

Inside sales and telecalling. Start with one or two callers working a queue built from gstin_entered without invoice_sent, plus limit-reached users and renewal-window users. As a planning figure, a fully loaded junior telecaller in a metro runs roughly ₹25,000 to ₹35,000 a month — replace that with local quotes, but do not replace it with zero. Two callers is roughly ₹60,000 a month, which is a tenth of a modest media budget and will move activation further than the tenth spent on ads would.

WhatsApp-led onboarding support. Invoice delivery, support, onboarding and referral all run through WhatsApp for this buyer, so support has to live there too. That means a WhatsApp Business API account through a business solution provider, template approval for anything you initiate, and per-conversation pricing that varies by template category — check current rates when you scope it. The important design point is that a template-based onboarding sequence pegged to GSTR filing dates and renewal dates is the re-engagement channel this category has instead of the one Google will not sell you.

The CA and accountant channel, on commission. Chartered accountants advise, and frequently veto, the software decision for Indian SMEs. Recruiting 20 to 50 CAs with a referral commission on first-year licence value costs almost nothing fixed, because it is variable by construction, and it reaches the buyer through the person they already trust. Give them a co-branded onboarding link, a client dashboard and a filing-calendar asset they can forward. Measure the CA-sourced share of paid conversions from month one — if it is not the largest single source by month six, either the channel or your commission is wrong.

Metrics for the motion, alongside the media metrics: connect rate, calls per activation, cost per activated business including telecalling cost, CA-sourced share of paid conversions, and renewal-desk contact rate in the 30 days before expiry. A cost per activated business that counts only media is not a cost per activated business.

Usually not as a primary channel, and the reason is arithmetic rather than preference. India iOS reach sits at roughly 4 to 6 percent of devices, so for an India-primary SME app this is a small, high-value slice. Run it deliberately or not at all.

Treat the cost anchors with care. AppTweak's India Apple Ads CPI of $0.89 is an all-category median across its India panel, not a B2B figure — and it is low because the India iOS auction is thin at that 4 to 6 percent reach, not because the channel is efficient. Cost per tap is panel-dependent too: AppTweak's 2025 dataset (about 3,500 apps, 50,000 campaigns, $1 billion in spend across 38 countries) puts the global median CPT at $0.92, the US at $1.91 and the US Business category at $2.70, while Adapty's separate panel of a million-plus ad groups puts the US at $1.58. Publish and plan on the range, never on a single number. There is no published Apple Ads benchmark for B2B SME apps at all.

Where you do run it, use Apple's own documented four-campaign structure rather than agency folklore:

Add every keyword from the first three into Discovery as exact-match negatives, and build one custom product page per trade, routing the matching keyword theme to each.

Now the honest part, and keep the two Apple products separate. The manual four-campaign structure runs on Manage Bids with a maximum cost per tap and has no conversion-volume gate. Maximize Conversions, launched 25 February 2026, is a different product: automated bidding to a target CPA, with documented guidance to budget for at least five conversions a day, run at least two weeks before judging, and Search Match mandatory on its automatic ad group. A B2B SME app optimising to invoice_sent will frequently not produce five conversions a day at any sane budget. That is a volume floor you either clear or you do not. If you cannot clear it, run Manage Bids manually and accept that you are operating the channel by hand. Lifetime budgets were paused in June 2026, so daily budget is the only model.

Android is the whole business in India. Start on App campaigns for Installs with target CPI, respect the budget floors, change nothing before 100 conversions, and then move to target CPA on invoice_sent. What you cannot do is comfort yourself with "high LTV" — the ticket is ₹699 to ₹1,599, and the CPI has to answer to that.

Google's documented minimums are the numbers most people get wrong:

Google's exact wording is that making changes before the first 100 conversions have registered may disrupt learning. That is a conversion count, not a time window, and here 100 activation events can take a month.

What is a good CPI here? Nobody publishes one. The closest India anchors come from a vendor's self-reported portfolio: utility ₹10 to ₹25, ecommerce ₹30 to ₹80, fintech ₹80 to ₹200. My estimate for India SME billing apps is ₹40 to ₹120 (about $0.45 to $1.36) — above utility because the audience is narrow, below fintech because you are not in a regulated-inventory auction. A reasoned estimate, not a benchmark.

What does an activated business actually cost, net of everything?

The version of this calculation that circulates is wrong because it stops three steps early. Here it is in full, at an ₹80 CPI.

Step one, cost per activated business. ₹80 CPI, a 45 percent GSTIN-completion rate and a 60 percent invoice-sent rate gives ₹80 ÷ 0.27 = about ₹300 (about $3.40) per activated business.

Step two, activation is not purchase. An activated business has sent one invoice. It has not paid you. If 15 to 25 percent of activated businesses convert to a paid licence — a range you must measure, because nothing is published — then cost per paying customer is ₹1,200 to ₹2,000, not ₹300. That step is missing from almost every version of this maths, and it is a factor of four to seven.

Step three, the licence is not what you receive. Take Vyapar's ₹799 Gold plan as the anchor. If that price includes 18 percent GST, the ex-GST amount is ₹677. Play's service fee then takes 15 percent of that on the first tranche of annual earnings and 30 percent above it, leaving ₹575 at the 15 percent rate and ₹474 at 30 percent. Confirm the exact base Play applies its fee to in your own payout reports; the direction is not in doubt.

Step four, support is the largest cost in the category and it is nowhere in this calculation. Vyapar's employee costs were ₹102 crore against ₹69 crore of revenue. Dividing ₹69 crore by a roughly ₹900 average realised licence implies on the order of 7.5 lakh paying licences, which puts people cost per paying customer somewhere near ₹1,300 a year — more than the licence. That division rests on an assumed average price and is illustrative rather than reported, but it is consistent with the disclosed ₹2.04 spent per ₹1 earned, and it is the honest scale of the number.

So does ₹300 hold up? No, not as stated. Year one at these figures loses money on every customer: ₹1,200 to ₹2,000 to acquire a payer, ₹474 to ₹575 of net first-year revenue, and a support cost that may exceed the licence. The business exists only in the renewal years, which is exactly what "the LTV comes from renewal duration and seat expansion" means in practice. At ₹575 net a year and a 60 percent annual renewal rate, lifetime net revenue is about ₹1,440; at 40 percent renewal it is about ₹960. Set those against a ₹1,200 to ₹2,000 acquisition cost and the conclusion is unavoidable: at a ₹799 ticket you cannot buy your way to a business through paid installs alone. The levers that make it work are pushing renewal above 60 percent, expanding seats and GSTINs within accounts, and moving a large share of acquisition into the CA channel and inside sales where cost per payer is lower. That is why the previous section exists and why the budget below spends on people.

Do the floor arithmetic before committing budget. At an ₹80 target CPI the floor is ₹4,000 a day (about $45); at a ₹300 target CPA on invoice_sent it is ₹3,000 a day (about $34). Both mean this channel needs a committed monthly budget, not a trial.

On burst installs. Two facts stay in every version of this advice. Apple added a named clause, 5.6.3 "Discovery Fraud," in the February 2026 Guidelines revision, under the Developer Code of Conduct where the stated remedy is termination of your Developer Program account rather than app rejection. Apple's sentence, in full: "Manipulating any element of the App Store customer experience, such as charts, search, reviews or referrals to your app, erodes customer trust and is not permitted." The Section 3 preamble covers third parties engaged to act on your behalf — "or engage with third-party services to do so on your behalf." Google separately states it filters incentivised installs out of its ranking systems, escalating to removal from top charts and then from the store, so the burst can be silently zeroed while you still pay for it. And Google's enforcement reaches past the app that ran it: "any related Google Play developer accounts will also be permanently suspended." In a category where a single customer is worth years of licence revenue, buying installs that will never enter a GSTIN is the worst possible trade.

Because the Indian SME owner is a consumer on Meta. There is no professional network layer between you and them: they are on Facebook and Instagram in the evening, and they respond to a video of an invoice being made faster than to B2B positioning.

The mechanics. Run App Installs optimisation to start. Move to App Event Optimisation on invoice_sent only when you clear the volume threshold, which is roughly 50 optimisation events per ad set per rolling seven days. Note the sourcing honestly: Meta's Business Help Center is blocked to crawlers, so that threshold comes from trade press rather than from Meta, and every number circulating is second-hand.

Targeting that works in India, in rough order of reliability: interest and behaviour proxies for business ownership, including engagement with competitor pages; lookalikes built from your activated users rather than your installers, which is a fundamentally different audience; and broad targeting with strong vertical creative, because in a market this large letting the creative qualify often beats narrow targeting that starves the ad set below the learning threshold.

Consolidate aggressively. A B2B SME app running six narrow ad sets will have six ad sets in permanent learning. Run two, run them wide, and let the creative segment the audience. Note that this is the same trade-off the vertical store-listing section priced on Google, and the answer moves with budget in the same way.

For contrast rather than as a plan: the equivalent US budget would go to search, review sites and accountant partnerships, because that buyer researches software rather than encountering it. Running an India-shaped Meta plan against a US audience produces cheap clicks from people who do not own businesses.

You mostly cannot, and a plan that does not say so will misread its own data for a year. This is the first question a practitioner who has run B2B mobile will ask, so here is the answer in full.

The windows are shorter than the decision. On iOS, AdAttributionKit gives you a 30-day click-through window and a 24-hour view-through window, with the first conversion value update available up to 60 days after install and three conversion windows from first launch at days 0 to 2, 3 to 7 and 8 to 35. On Google, conversion windows are configurable per conversion action — commonly 30 days for click-through, with view-through shorter — so check what your account is actually set to rather than assuming.

Now put a real buyer against those numbers. A shop owner sees your GSTR-deadline video in November, does nothing because November is busy, remembers you at financial year-end, searches "billing app" in March and installs. That is roughly 120 days. Every window has closed. The install arrives as organic or as branded search, and the November video that caused it is recorded as having produced nothing.

Five practical responses, in the order they pay off:

1. Treat branded search and organic as a paid outcome, not as free supply. Track branded search volume and your Play Console acquisition report — Google Play search versus explore versus third-party referral — as a channel, and read it against paid spend from previous months rather than the current one. If paid spend goes up and branded organic follows six weeks later, that is your answer.
2. Instrument the web-to-app path properly. Many of these decisions start on a web landing page and finish in the store weeks later. Capture the GCLID on the web touch and carry it through a deferred deep link so the eventual install can be joined to it. Two silent failure modes make this not work: deferred deep linking has to be enabled in your measurement SDK, and your robots.txt must allow AdsBot-Google and AdsBot-Google-Mobile to reach assetlinks.json and apple-app-site-association. The second is the most common misconfiguration in mobile marketing and it fails without an error message.
3. Do not build one multi-touch model across both platforms. Android attribution is deterministic and user-level — the Privacy Sandbox for Android was cancelled in October 2025, so GAID persists — and a genuine multi-touch view is possible there. iOS is aggregate and windowed, and no amount of modelling makes it comparable. Report them separately or you will report a blend that describes neither.
4. Run incrementality holdouts, because they are the only method the windows cannot break. Hold out a set of matched geographies or an audience segment for three to four weeks and compare total activated businesses, not attributed installs. The reference point for what to expect: the MIT advertising-shutoff experiments (Ju, Zhao and Aral, 2025 — three shutoff experiments across six apps over 500 days) found roughly 37 paid installs and about 3 organic installs per $100 of ad spend, an organic halo near 8 percent. That is the only causal estimate of its kind in public, and it is the shape of experiment to copy.
5. Ask. A single "how did you hear about us?" field on the business-profile form the user is already filling is the cheapest multi-touch instrument available, and the only one that spans 120 days. It is self-reported and biased, and it is still better than nothing, which is what the windows give you.

And one rule that follows from all of this: never judge a long-consideration channel on a seven-day report. Set your evaluation window to a quarter, agree it with whoever is funding the spend before you start, and hold to it when month one looks bad.

Here is the constraint this category cannot dodge: at realistic budgets, a B2B SME app may never generate enough weekly conversion events to optimise to a deep rung on any platform. That is structural, not a skill problem.

Work the numbers. Meta reportedly needs about 50 optimisation events per ad set per seven days. If your install-to-activated-business rate is 25 percent — optimistic, given the GSTIN gate — you need 200 installs per week per ad set to optimise to invoice_sent. At an ₹80 CPI that is ₹16,000 a week (about $180) per ad set: achievable. One rung deeper, to payment_recorded at perhaps 10 percent of installs, needs 500 installs per week per ad set. Deeper still, to subscription at 2 to 3 percent, needs 1,700 to 2,500 installs a week to feed a single ad set. Very few SME launches in India will fund that.

So: optimise to invoice_sent and stop. It is the deepest rung most launches can feed, and chasing subscription optimisation on sparse data produces campaigns permanently stuck in learning while the model fits noise — exactly why Google advises against selecting more than one action for target CPA. Consolidate ad sets and campaigns below the budget thresholds set out earlier. And accept manual operation where automation cannot work — Manage Bids on Apple Ads, longer evaluation windows on Google.

What replaces Google's re-engagement product when you cannot use it?

That 50,000-install gate on App campaigns for Engagement deserves more than a row in a table, because of what it costs you specifically. This is a renewal-driven business with a published calendar: GSTR deadlines every month or quarter, financial year-end on 31 March, and a licence expiry date for every customer. Re-engagement is therefore the highest-value motion available to you — and Google's re-engagement product is closed until 50,000 installs, which most apps in this category never reach. The channel gap sits exactly where your best opportunity is.

Three things fill it, and all three are already in this post:

• WhatsApp Business API sequences pegged to GSTR filing dates and to each customer's renewal window. Templates need approval and conversations are priced, so design a small number of high-value sequences rather than a newsletter.
• Email against the filing calendar, which is free, unrestricted by install counts, and works because the deadline is real.
• Inside sales at the renewal window, which is the same telecalling capacity you staffed for activation, pointed at expiry dates instead. A renewal call 30 days before expiry is the highest-value call in the business, because renewal duration is where the lifetime value in this category comes from.

Treat those three as your re-engagement channel and budget them as such, rather than waiting for an install threshold that a low-volume category is not designed to cross.

Screen-record creating a GST-compliant invoice end to end in under 30 seconds and sharing it on WhatsApp. In utility B2B, showing the job done beats any claim about doing it. Then peg your rotation to the GST filing calendar, because a dated deadline outperforms evergreen feature creative in a category where the anxiety is scheduled.

Six things that work, offered as a synthesis to test:

1. The demo is the ad. One job, done fast, filmed on a phone — not a feature list.
2. Lead with the deadline. "GSTR-1 due in three days" beats any benefit statement. Build a filing-calendar rotation and refresh it monthly.
3. Vertical creative beats horizontal. A kirana owner, a pharmacy, a garment wholesaler and a contractor have visibly different invoices. Show theirs — "small business" as a frame under-converts where the buyer identifies by trade. This is also how you get vertical differentiation below the budget at which vertical campaigns become affordable.
4. Show the WhatsApp share step. Invoice delivery, support and referral all run through WhatsApp in Indian SMEs, and showing it signals "this fits how I already work."
5. Trust proxies matter more than polish. Businesses served, accountant endorsement, and explicit data-backup messaging — losing the books is the SME owner's biggest fear.
6. Treat the accountant as a separate campaign. Chartered accountants advise, and often veto, the software decision. Creative aimed at CAs is a distinct audience with distinct copy, and it feeds the commission channel described earlier.

Two things to avoid. Enterprise-SaaS visual language — abstract dashboards, gradient hero shots — signals "expensive and complicated" to this buyer. And be careful with discount anchoring: every major competitor advertises 28 to 50 percent off list, so the category has trained buyers to expect a deal. A clean, low, honest price is differentiating, but you are fighting a convention.

Budget for people first, media second, and a payback measured in years. There are no published India CPI or LTV benchmarks for SME B2B software from any tier-one measurement company — none exist, and any page quoting them is fabricating. Build bottom-up from your own price ladder and measured activation rate.

Your price anchors are public: Vyapar at ₹699 to ₹1,599 (about $8 to $18) per plan, myBillBook at ₹2,599 to ₹4,999 (about $30 to $57), and Zoho Books India at ₹749 to ₹8,332 per month by tier, excluding 18 percent GST, with a genuine free tier below ₹25 lakh (about $28,000) of revenue. Zoho's free tier sets a hard floor on what anyone can charge for basic invoicing, pitched exactly at the boundary below which businesses are not compelled to be sophisticated. And Vyapar earns 90 percent of income from software licences rather than monthly subscription — Indian SMEs resist recurring charges, and an annual licence framing generally converts better than monthly SaaS here.

Creative production and people come out of the monthly figure, not on top of it. Note the shape: at Validation, roughly a quarter to a third of the budget is people rather than media. That is deliberate, and it follows directly from the finding at the top of this post.

Why the top tier is capped, when other posts in this series scale past it. ₹35 lakh a month at an ₹80 CPI buys about 44,000 installs a month, or half a million a year, in a category whose leader does ₹69 crore of total revenue. Uncapped budgets in this category do not buy more customers, because the binding constraint is the number of Indian small businesses actively shopping for billing software in a given month, not your budget. The way you find the ceiling is the same CPI-drift check that works anywhere: if your CPI rises more than about 20 percent while install volume stays flat, the auction is exhausted and the next rupee belongs in the CA channel, the renewal desk or the product — not in media.

The KPI ladder, in priority order:

1. GSTIN completion rate. The onboarding gate, and the biggest lever in the funnel.
2. Install to invoice-sent rate, reported by channel and by vertical creative.
3. Cost per activated business including telecalling cost, not cost per install.
4. Activated-to-paid rate, the step most plans omit and the one that turns ₹300 into ₹1,200.
5. Second invoice within seven days — the best early predictor of retention.
6. Record depth at D90 — invoices and customers stored, not sessions, on activated users only.
7. Annual renewal rate. What the whole model rests on, and one you will not measure for a year — build a leading proxy now.

The honest closing note: the category leader spent ₹2.04 to earn ₹1 in FY25. Plan for payback measured in years and a first year that does not pay for itself.

Six failure modes account for most of the damage in this category, and the expensive ones are structural rather than tactical.

You believed the "high LTV" story. Cost: a media budget calibrated to a ticket that does not exist. At ₹699 to ₹1,599 a year, net of Play's fee and GST, a customer returns under ₹600 in year one. If your plan quotes lifetime value without stating a renewal rate, it is not a lifetime value.

You budgeted media and not people. Cost: an activation rate that never improves. The businesses that entered a GSTIN and stopped are your highest-yield audience and they are reachable by phone. A plan with no telecalling line is a plan that leaves them there.

You added a credit feature. Cost: a regulated launch you did not plan for. The moment your app facilitates lending, Play's Financial Services policy applies, RBI listing becomes a precondition of publication, and your ads account needs financial services verification. Weeks to months, at best.

You marketed filing before you had GSP or ASP access. Cost: installs on your highest-intent keywords that churn at the first attempted filing, plus a store listing making a claim the build cannot meet.

You judged a four-month decision on a thirty-day window. Cost: killing the channel that was working. Branded search and organic installs weeks after a campaign are the campaign's output, and no attribution window will tell you so.

You never became the system of record. Cost: everything, on a delay. A duplicate record beside a paper ledger churns at the first busy week, and it churns quietly — your dashboard shows ordinary decay rather than the structural failure it is.