Astrology app marketing strategy: no benchmarks, a real playbook

India is the larger and more profitable market: Astrotalk booked ₹1,214 crore (about $138M) total revenue in FY25, up 85% year on year, with adjusted profit before tax of ₹285 crore ($32M), on audited figures reported by Outlook Business. The structural fact: India runs a per-minute human consultation marketplace. The activation event consumes a paid human being's time, which is why a media plan alone cannot budget this category.

Revenue from operations was ₹1,176 crore ($134M), up from ₹656 crore in FY24, and the Astrotalk Store did ₹140 crore ($15.9M) in CY25 in its first year. Engagement rose 27%, led by Tier-I cities rather than the small-town growth story most people assume. The company became a unicorn at $1bn through an ESOP buyback, per Entrackr. The FY24 comparative field, from Storyboard18:

Astrotalk carries 10 crore+ Play installs (100M+) at 4.6 stars across 16.5 lakh reviews.

The number that should set your expectations: Astrotalk spent ₹162.7 crore ($18.5M) on advertising and promotion in FY24 — about 25% of that year's revenue. That is the CPM floor you are bidding against, and it is the reason this post refuses to plan against a cheap CPI. A leader spending a quarter of revenue on media, profitably, is not a category you enter with a ₹5 lakh test.

And the install pool is not growing. India ran about 6.3 billion app downloads in Q2 2026 and has been flat at that level since 2023, on Sensor Tower figures reported via TechCrunch. The growth number people quote alongside it — up 35% year on year — is consumer spend, $345M in the quarter, and it belongs to money rather than to volume. India's astrology story is monetisation depth per user, not install volume, which is exactly what a wallet-recharge model is built to harvest.

For contrast rather than as a plan: per Statista and AppMagic for Q1 2026, CHANI was the highest-grossing US astrology app, followed by Co-Star, with dollar figures paywalled. Nebula is the largest download-side player there, selling subscriptions from $2.99 to $49.99 plus chat credits at $9.99 and $24.99 — the same human-consultation primitive wrapped as credits rather than a wallet. Both stores classify astrology under Lifestyle in both markets.

The activation event is first_wallet_recharge. It is the money event, it is a real bid target with a real threshold, and the rest of this post configures it. The earlier rungs are volume and diagnostics, and naming one of them "the activation event" — as an earlier version of this post did with paywall_hit — is how a taxonomy ends up with an activation event that nothing ever bids on.

The India ladder, reconstructed from Astrotalk's live product and reported model:

install → birth details entered (DOB, time, place) → free Kundli generated →
first free chat with an astrologer → paywall at free-minute expiry →
FIRST WALLET RECHARGE → first paid minute → repeat consultation → membership pass

The birth details are the data moat, not a form. Date, time and place of birth make every subsequent screen personal, and under India's DPDP Act 2023 they are sensitive personal data alongside religious affiliation — explicit consent, purpose limitation, deletion rights.

The free chat is the whole conversion mechanism. Astrotalk pairs it with explicit risk reversal — switch astrologers at any time, low-rated minutes refunded within 24 hours. Your free-minute allowance is the most important number in your funnel model, and it is a marketing variable, not a product default: it sets your conversion rate and your supply cost at the same time, in opposite directions.

Where the recharge lands, and what iOS will actually tell you about it. For converting users the recharge happens in the same session as the free chat, which puts it inside AdAttributionKit's days 0–2 window. That is genuinely unusual and it is worth engineering for. But timing alone does not buy you a signal. A fine conversion value only arrives in the first postback and only at Tier 2 or Tier 3 crowd anonymity, and a launch-volume app sits in the low tiers — where the first postback is coarse-only, and at Tier 0 there is no postback at all. So the correct statement is: good timing puts you in the only window that can ever carry a fine value, and your own install volume decides whether you receive one. That advantage arrives in month three, not launch week. Until then, validate on Android, where attribution is deterministic and Privacy Sandbox's cancellation on 17 October 2025 means GAID persists.

The economics, from reported figures on Astrotalk's model: ₹20–75 per minute on the live listing page, most astrologers between ₹23 and ₹50; a membership pass at ₹99 a day across ten days; a take rate of about 20%; roughly 90% of revenue from calls and chats; ARPU of ₹200–500 a month for new users, up to three times that for repeat users; wallet recharge IAPs up to ₹9,900 ($113). Two operating numbers to plan around: a reported CAC payback of roughly 8 months on contribution margin, and about 50% of monthly transacting customers coming from paid — this category never goes organic-led, and equally, half of the leader's transacting base carries no acquisition cost at all. Your launch cohort will not have that half.

Install-to-payer conversion is not published for this category. Instrument your own, and log one second-order event from day one: repeat consultation within 30 days. Usage here is event-based and sporadic — people consult when something happens in their life, not on a daily loop — so D7 will look bad while the business is healthy.

This is the most consequential decision in the post and the one most likely to be answered wrongly by confident blog advice, including an earlier version of this one. Apple guideline 3.1.3(d), Person-to-Person Services, verbatim:

"If your app enables the purchase of real-time person-to-person services between two individuals (for example tutoring students, medical consultations, real estate tours, or fitness training), you may use purchase methods other than in-app purchase to collect those payments. One-to-few and one-to-many real-time services must use in-app purchase."

Here is what that text plausibly supports and where it stops.

A per-session charge for a live 1:1 consultation, billed at the point of service, is the product 3.1.3(d) describes. Two individuals, real time, the money attached to that specific session. If you charge ₹250 for a twenty-minute reading with a named astrologer at the moment it happens, you are inside the shape of the rule.

A prepaid wallet is a different product. Stored value bought in advance for unspecified future digital consumption is not a payment for a real-time service between two individuals — at the moment of purchase there is no session, no counterparty and no service. It looks much more like the digital currency Apple has consistently required to run through in-app purchase, and it becomes more so the moment the balance can be spent on anything other than 1:1 consultations: a live stream, a report, a Store product. That is the most likely reason the category leader lists a wallet-recharge in-app purchase despite operating a business built almost entirely on 1:1 chat. I am describing the constraint, not that company's configuration, which is not published.

Do not build an off-IAP rail on a blog's reading of a guideline, including this one. The remedy for getting 3.1.3(d) wrong is not a margin adjustment, it is removal. Put the exact question to App Review in writing and build against the answer you get back. Describe your actual flow — prepaid balance, what it can be spent on, whether unspent balance is refundable, whether any non-1:1 surface can consume it — and ask whether it may be collected outside in-app purchase. Use the App Review contact form or an App Review Appointment through "Meet with Apple". Budget four to eight weeks for this and start it before you build the billing layer, because the answer changes your take rate by 15 to 30 points and therefore changes every CAC threshold below.

What is settled: a live-stream astrologer session is one-to-many and must use in-app purchase. That makes your live-stream line structurally lower-margin than your chat line whatever happens with the wallet question, and it should change how you value traffic to each surface.

Apple's rules are not the binding constraint on recurring billing in India. RBI's e-mandate framework is, and it is the direct cause of India's high subscription billing failure rates. If you are modelling LTV off renewal counts, this is where the model breaks.

Three requirements to design around.

Two consequences for the plan. First, a pre-debit notice is a monthly cancellation prompt you are required to send. Price and package accordingly: a ₹99-a-day product generating a daily notification is a different retention problem from a monthly pass generating twelve. Second, billing failure is a bigger share of your churn here than in the US comparisons you will read. RevenueCat's 2026 dataset across 115,000+ apps puts billing failure at 31% of Google Play cancellations against 14% on the App Store — and that is a global blend, before India's mandate friction. Model involuntary churn as a separate line from voluntary churn, and hold a dunning budget.

This applies whether you bill through in-app purchase or your own rail. Apple and Google subscriptions charged to Indian cards sit inside the same mandate regime.

Apple guideline 3.1.2 requires subscriptions to deliver ongoing value, run a minimum 7-day period, be available across all of a user's devices, never require tasks such as social posts to get what was paid for, and disclose price, duration and content in full before purchase.

Read the 7-day minimum against a daily-pass product. An auto-renewable subscription cannot have a period shorter than seven days, so a genuinely daily auto-renewing product has to be built as a consumable instead. That is the constraint. What any specific competitor has configured behind their store listing is not published and I am not going to infer it from a price point.

On trial pricing, the honest number to design against: RevenueCat's 2026 data shows 55.4% of 3-day trial cancellations happen on day zero and 84% by day one. A trial that is not obviously worth keeping in the first hour is not a trial, it is a refund queue.

Your chat surface is user-generated content between two people, and it carries the same obligations a dating app carries. Post 01 covers both regimes at length and the requirements are identical, so this is the short version with the astrology-specific edges.

Apple 1.2 requires four mechanisms, all of them shippable features rather than policy text: filter objectionable material before it posts, a report mechanism with a timely response, the ability to block abusive users, and published contact information. The astrology edge case is that your "abusive user" reporting has to work in both directions — a consultation is one paid stranger and one paying stranger, and the paid side needs the block button as much as the paying side does.

Google Play's UGC policy asks for the same four capabilities in different words: consented terms of use, reporting and blocking that each work against content and against the person who posted it, and enforcement that actually happens on a clock you can evidence.

India's IT Rules 2021 make you an intermediary: publish a Grievance Officer with contact details, and remove unlawful content within 36 hours of actual knowledge. At 5 million registered users you become a Significant Social Media Intermediary, adding a Chief Compliance Officer, a Nodal Officer and monthly transparency reports. Astrotalk's 10 crore Play installs tell you that threshold is reachable in this category, not theoretical.

One astrology-specific exposure worth naming to counsel: the consultation transcript is the UGC, and it will contain health, relationship and financial disclosures made by the user. Your moderation and retention policy has to handle content that is sensitive personal data under DPDP at the same time as being the evidence you would need in a grievance.

Five things, and two of them have lead times measured in weeks: the App Review answer on your payment rail, your Play production access, your event taxonomy, your attribution stack, and a claims-safe creative library. Start the slow ones before the build, not after it.

Play production access needs 12 testers for 14 continuous days — personal Play Console accounts created after 13 November 2023 only; organisation accounts are exempt. A tester who stays fewer than 14 days contributes nothing, and one who leaves and comes back resets the clock to zero. Budget another week for review on top. If a guide tells you 20 testers, it predates the current rule.

The event taxonomy. Name these once and never rename them. Note where each one sits: exactly one is the activation event, exactly one is the volume rung, and the rest are diagnostics that must never become bid targets.

Why paywall_hit is a diagnostic and not a bid target. The advice in this post is to keep the free-minute allowance short so the paywall lands inside the first session. Follow that advice and nearly every user who starts a free chat hits the paywall — which makes paywall_hit almost a copy of first_chat_start, carrying the same volume and almost no extra information. Two near-identical events in an eight-slot priority list is a wasted slot, and bidding on the second one buys you the first one at a worse name. Keep it instrumented; it is how you tell "nobody starts chats" apart from "everybody quits at the meter". Do not optimise to it.

Firebase or an MMP. Mark bid targets as key events in Firebase — maximum 30 per standard GA4 property, and marking is not retroactive, so mark first_wallet_recharge on day one even though you will not bid on it for two quarters. Firebase alone is sufficient while Google is your only paid channel. It stops being sufficient the moment Meta enters, because Firebase does not feed Meta and you would be running two definitions of "recharge" against each other.

Both stores put astrology in Lifestyle. The rejection risk is Apple guideline 4.3(b), which names fortune telling explicitly, in the same sentence as dating. Apple 4.3(b), verbatim:

"Certain app categories — dating, flashlight, sound effects, wallpaper, simple timers, and fortune telling — are well established on the App Store and will not be accepted unless they offer a meaningfully different or improved experience."

Repeated submissions can lead to removal from the Apple Developer Program, and there is documented precedent of Apple rejecting horoscope apps because the store already has enough of them. Write your differentiation argument before you submit, and make it concrete: a verified-astrologer supply model, a regional-language consultation network, a practice no incumbent supports. "Better charts" is not one. Apple's remedy is procedural — an App Review call or an appointment through "Meet with Apple", which is also where your 3.1.3(d) question goes. Bundle the two conversations.

Live listings are your ASO template:

The India pattern is unmistakable: the service verb carries the title, not the noun. Every leader puts "talk to astrologer" in the title or short description — not "horoscope". India head terms, from leaders' metadata: talk to astrologer, astrologer online, chat with astrologer, kundli, kundali matching, rashifal, panchang, tarot, free kundli, janam kundali. Carry both the transliterated tokens and the service verb. The US listings invert it into chart nouns, which is a useful illustration of how differently the two markets search and not a keyword set you should be buying.

Age rating. Astrotalk is 13+ on the App Store but "Rated for 3+" on Play — an inconsistency between two listings for the same product, which tells you to pick yours consciously rather than by default. Apple's five-tier questionnaire was due 31 January 2026, and the social media declaration questions activated on 9 July 2026 with responses mandatory from September 2026 for new submissions, updates and notarization. Those dates come from third-party analysis of Apple's timeline rather than from an Apple page, so verify them in App Store Connect. If you have a chat or live-stream surface, assume the declaration applies to you.

Category review risks: one-star reviews cluster on wallet balance disputes, astrologer quality variance and unexpected charges as free minutes expire. Watch the share mentioning money — it is your leading indicator that the free-minute expiry is landing badly.

Use the structure Apple itself documents rather than an agency's variant of it: four campaigns, with the first three locked to exact match and Search Match switched off.

Add every keyword from the first three into Discovery as exact-match negatives, so Discovery only spends on new terms.

The India keyword split that matters more here than in any other category. Category India contains two intent groups that behave completely differently, and they should never share an ad group. The service group — talk to astrologer, chat with astrologer, astrologer online — is expensive, converts to a recharge, and is where your money should go. The utility group — free kundli, kundali matching, rashifal, panchang, janam kundali — is cheap, high-volume, and converts to a free chart rather than a consultation. Report them separately or your blended CPI tells you nothing at all: the utility group will drag your average CPI down while contributing nothing to the wallet, and you will read that as efficiency. Competitor India: astrotalk, astroyogi, astrosage, instaastro. Use Custom Product Pages to land the two groups on different screenshot sets — a user searching "free kundli" and one searching "talk to astrologer" should not see the same first screen.

On how much of the budget Apple Ads should carry: iOS is roughly 4–6% of Indian devices. Any share above that is a deliberate bet that your iOS users recharge at higher values, which is plausible in a wallet model and unproven for your app. Set it above device share on purpose or not at all. And carry the panel spread with the cost data: on US cost per tap AppTweak reads $1.91 while Adapty's panel of over a million ad groups reads $1.58. Same metric, same year, a fifth apart. India's $0.89 CPI has no second panel checking it at all.

Bidding, and why the automated option fights the structure above. Since 25 February 2026 you can hand a search-results campaign a target CPA and a daily budget and let Apple bid. Four documented conditions come with it: fund at least five conversions a day, judge nothing for two weeks, it is unavailable on pre-order campaigns, and Search Match is compulsory on the automatic ad group. That last condition is the problem — the four-campaign structure works precisely because the first three campaigns are exact-match with Search Match off, and Apple has issued nothing reconciling the two approaches. Pick one per campaign and keep the accounting separate. Two housekeeping facts: lifetime budgets were paused in June 2026, so daily budget is the only model, and your monthly ceiling is daily × 30.4.

Fund ACi to Google's documented floor and then leave it alone. The floor is a daily budget of at least 50 × your bid on tCPI. The leave-it-alone part is Google's, not folklore: "making changes to your in-flight campaign before the first 100 conversions have registered may disrupt learning." Read that carefully — the gate is counted in conversions, so a campaign starved of budget stays inside it indefinitely rather than ageing out of it.

The rates this post plans against, stated once and used everywhere. Two rates drive every threshold below, and both are derived, not cited — nobody publishes either for this category.

Three percent is the number, and it is used in every calculation in this post. An earlier version of this post carried 7.8% in one worked example and 3% twenty-five lines later, which understated the cost of the money event by a factor of 2.6 and mis-budgeted the entire Scale scenario. One rate, applied everywhere, or the plan is decorative.

The floors, run across the planning CPI band of ₹78–₹128 derived in the budget section below.

That last row is the real cost of optimising Google to the money event, and it is roughly three to five times what a plan built on a 7.8% recharge rate would tell you. It is why most astrology apps sit on the chat-start rung for two quarters, and why the Scale scenario in this post is budgeted at ₹45 lakh a month rather than ₹30 lakh.

App campaigns for Engagement, with both of its gates. Given how sporadic usage is here — a user who consulted three months ago and has a life event today is genuinely warm — ACe is unusually valuable in this category. It also has two gates, and most write-ups quote only the first. You need 50,000 installs, and a daily budget of at least 15 × your bid. At a recharge-level tCPA of ₹2,600–₹4,267 that is ₹39,000–₹64,000 a day, or ₹11.9–₹19.5 lakh a month, for that campaign alone. ACe is a Scale-tier instrument, not a re-engagement afterthought. Two mechanics break it silently even when funded: on Android, ACe needs session_start events carrying the GCLID or it measures Display and YouTube only, never Search; and Adjust and Branch both need the inactivity window set to 0 days.

Astrology creative gets rejected in ad review queues, not in courtrooms, and the operative documents are Meta's and Google's ad policies — not ASCI's code. Both matter and they are enforced by completely different mechanisms. Here is the platform half, which the Indian-regulator half of this post used to be missing.

Neither Meta nor Google publishes an astrology-specific advertising policy. The exposure runs through two policy families that astrology creative walks into by default.

Two operational consequences. Write in the first person, never the second. "I did not know whether to take the job" is a story; "Are you unsure about your career?" is an assertion about the viewer. The first survives review, the second is a coin flip that will eventually cost you an account-level strike. And build the compliant creative library before launch week, because Google and several other channels require pre-approved creative for restricted verticals and the approval time is not something you want to discover mid-flight.

Read both policy centres the week you brief creative. Clause text moves; the two families above have been stable.

Distinct from the ad platforms, and frequently conflated with them. Two bodies, two very different sets of teeth.

ASCI is a self-regulatory body operating a voluntary code. It does not prohibit anything and it levies no fines. What it does is take complaints, rule on them, publish the outcome and refer non-compliant advertisers onward to the statutory regulator. Its code disallows absolute claims — "100% accurate", "guaranteed" — and in practice an adverse ASCI ruling is a reputational and referral event, not a penalty notice.

The CCPA, under the Consumer Protection Act 2019, is statutory and does levy penalties: up to ₹10 lakh ($11,400) for a first misleading-advertisement offence and ₹50 lakh ($56,800) for a repeat, with liability extending to endorsers. The CCPA has actively penalised "100%" claims in other verticals, so the enforcement pattern is live rather than theoretical. Those penalties are the CCPA's, not ASCI's — a distinction the category's marketing content gets wrong constantly.

The Drugs and Magic Remedies (Objectionable Advertisements) Act 1954 is the classic India exposure for remedy and cure claims. I could not verify a specific application to astrology apps, so treat it as a question for counsel rather than an established finding — but it is where a "this gemstone will cure X" line in your Store copy would bite.

DPDP Act 2023 covers the birth data and religious inference at the heart of your product: explicit consent, purpose limitation, deletion rights.

Lead with the unresolved question, not the prediction. The reference asset is Astrotalk's September 2026 campaign: three films directed by Nitesh Tiwari, built on emotional restraint rather than mysticism, on the insight that "uncertainty today is not always about the absence of opportunity, but about the absence of clarity." Astrology is positioned as a companion for decision-making, not a solution.

None of the three scenarios is mystical: a marriage drifting into emotional distance; a millennial romance with a deferred commitment conversation; a small-business owner's confidence crisis despite a growing sector. That is deliberate category expansion beyond love into career and business anxiety, a much larger emotional territory that lets you buy inventory love-only creative cannot. It was amplified during IPL 2026.

Notice that the leader's campaign is also, structurally, a Meta personal-attributes workaround: it dramatises a character's doubt rather than diagnosing the viewer's. That is not a coincidence, and it is the single most useful thing to copy.

• Hook on the moment of doubt, phrased as a character's question rather than an accusation aimed at the viewer.
• The offer is the free first chat, because it maps directly to first_chat_start, your volume rung.
• No accuracy claims, no guaranteed outcomes, no testimonials implying certainty.
• Volume discipline: 50 to 80 or more variants a month, UGC-style talking heads, hook rate as the primary lever.

The audience insight that should shape your geo split: Astrotalk's growth is led by Tier-I cities, not the small-town assumption most India media plans start from. Cheap tier-2 and tier-3 CPMs buy installs with no wallet behind them, which shows up as a healthy CPI and a broken recharge rate.

No published creative-intelligence study exists for astrology apps in either market. The bullets above are extrapolated from Astrotalk's own campaign plus general mobile-UA findings, and I would rather say so than present them as measured.

Meta's numbers here are trade-press consensus, and that needs saying every time: Meta's Business Help Center is robots.txt-blocked, so the widely cited ~50 optimisation events per ad set per rolling 7 days cannot be sourced to Meta. It is consistent across every credible secondary source and it is the number practitioners plan against, but do not cite Meta for it. Three things that are documented on developers.facebook.com are worth carrying instead: the App Installs goal is Meta's own recommended default; Value Optimization is gated behind purchase data and, in Meta's phrasing, "available on a limited basis to partners and advertisers", so do not plan around it; and once your SDK is live, Meta advises against Link Click Optimization outright.

The rung arithmetic, at the 3% recharge rate and the ₹78–₹128 planning band.

So recharge optimisation on Meta needs a single ad set carrying ₹5.6–₹9.2 lakh a month before it is even eligible to learn. Below that, first_chat_start is the correct target and it clears comfortably inside a first-velocity budget.

The iOS priority list has eight slots and you should not fill all of them. Meta reports only the highest-priority event a session triggers, and re-ordering the list freezes the affected ad sets for 72 hours, so treat the ranking as a decision you make once. Rank first_wallet_recharge first, then repeat_consultation_30d, first_chat_start and birth_details_complete. Deliberately leave paywall_hit off it: because your paywall lands inside almost every free chat, it would win the priority contest against first_chat_start in nearly every session and report a strictly less informative event in its place. A note on scope, since this is where playbooks go stale — the standalone AEM tab disappeared for many web accounts, but the eight-event model still governs iOS app campaigns.

Descending the ladder without stalling. Every rung you skip costs you a month. Launch on tCPI; hold until Google has 100 conversions and Meta has cleared roughly 50 events a week; move to first_chat_start; hold again; only then attempt first_wallet_recharge, and only if the budget table above says you can fund it. The failure modes are cumulative rather than alternative: a starved campaign never exits learning so CPA inflates, a sparse signal makes the model fit noise (which is why Google states "it isn't recommended to select more than one action" for tCPA), and each corrective edit restarts the clock you were waiting on. On iOS, add signal decay to that list — windows two and three return coarse values only, and Tier 0 returns nothing.

There is no CPI benchmark for astrology apps, anywhere. No MMP, no store, no analyst report with an astrology cut. Nor is there a published cost per first chat, cost per recharge or D1/D7/D30 series. What genuinely exists, with sourcing:

The nearest retention proxy is Adjust's combined cross-category series — D1 26%, D7 13%, D30 7% — on a page carrying no stated period, so label it undated wherever you repeat it. It is a sanity check, not a target: expect D7 to underperform it while your 30-day repeat rate carries the business.

Deriving the planning CPI, out loud, because the number an earlier draft used was indefensible. That draft planned at ₹70 — below the only published India datapoint — in a category whose leader spends ₹162.7 crore a year and which the same draft called the CPM floor you are bidding against. Both cannot be true.

• The floor is ₹78 ($0.89), AppTweak's Apple Ads India CPI. It is search-intent inventory in a thin auction. It is the cheapest install available in India and it is not a blended planning number.
• The ceiling is ₹128 ($1.46), Adjust's global blended dating CPI, January 2023 to December 2024. Dating is the nearest category with a real MMP dataset: same store category (Lifestyle), named alongside astrology in the same Apple 4.3(b) sentence, similar broad-reach buying. It is geo-blended across markets far more expensive than India, which is why it sits at the top rather than the middle.
• Plan the band ₹78–₹128 and run every threshold twice, once at each end. That is arithmetic, not a published benchmark. Replace it with your measured CPI in week two — it is the deliverable of your first month.

Budget scenarios, with the astrologer-supply line that belongs in all of them. Every free chat your ads generate consumes a paid human's minutes. At a planning consultation rate of ₹30 a minute and a roughly 80% astrologer share, a five-minute free chat costs you about ₹120 in astrologer payout, and a roster can serve about eight free chats per astrologer-hour including queueing and handover. Those two numbers turn install volume into a headcount and a bill.

The supply line adds roughly 30% to your media CAC and it is missing from every published plan in this category. That is the headline finding of this section. Note also that it does not fall with scale — it is a variable cost per free chat, so it scales linearly with the volume your media buys.

Now put the true CAC against the take rate, which is where this gets uncomfortable. At a 20% take rate, recovering a fully loaded ₹4,200 per first recharge requires about ₹21,000 of gross consultation value from that user — roughly 700 paid minutes at ₹30. Whether that is a viable business depends on a figure nobody has published cleanly:

• **If the reported ARPU of ₹200–500 a month is platform revenue, recovery takes roughly 8 to 21 months** — which brackets the reported ~8-month CAC payback for the category leader, and that leader has brand pull and gets half its transacting customers without paying for them.
• **If it is gross consultation value, platform revenue is ₹40–₹100 a month and recovery takes three and a half to nine years**. The model does not work at launch prices.

Resolve which one it is before you commit budget. The reported figures do not specify, and the difference between the two readings is the difference between a business and a very expensive experiment. The levers if the answer is unfavourable: shorten the free-minute allowance (raises recharge rate, cuts supply cost, hurts conversion), raise the take rate above 20%, or push repeat rate — the repeat user is reportedly worth up to three times the new one, and repeat_consultation_30d is the only event in your taxonomy that moves that number.

The KPI ladder, in order: free-chat start rate as a share of installs, recharge rate as a share of free chats, average first recharge value, astrologer payout per recharge, 30-day repeat rate, true cost per recharge including supply, cost per recharge on media alone, blended CPI last. Blended CPI is last because the utility keyword group drags it down while adding nothing to the wallet.

Five things break: the payment-rail question is answered by a blog instead of App Review; 4.3(b) stops you shipping; astrologer supply cannot absorb the free chats your ads generate; you optimise to the recharge before you can fund the threshold; and you buy a burst that gets filtered or terminated.

The supply failure is specific to this category and consistently underestimated. Astrology acquisition converts into a request for a human being's time within minutes of install. If you cannot staff the free chats your Saturday-evening Meta campaign generates, you have paid full price for a queue, a bad review and no recharge. The capacity row in the budget table is the fix: it is a headcount, not a hope.

Before you buy a burst

The two facts that turn a burst from a tactic into a company-level risk, and neither can be engineered around.

Apple named the clause. 5.6.3 "Discovery Fraud" was added in February 2026. It sits in the Developer Code of Conduct, which matters more than the clause text: the remedy attached there is losing the Developer Program account, not losing the submission. "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 extends it to anyone you hire — "or engage with third-party services to do so on your behalf" — so an agency does not insulate you.

Google's version is quieter and wider. Incentivised installs are, by Google's own account, filtered back out of the ranking systems they were bought to influence — with escalation to removal from top charts and then from the store. The installs get zeroed; the invoice does not. And the consequence is not confined to the app or the account that bought them: "any related Google Play developer accounts will also be permanently suspended." If your company holds more than one Play account, one burst puts every one of them in scope.

If you buy CPI inventory anyway, warrant "no incentivised or rewarded inventory" in the insertion order and make blocked installs non-billable. And note that the compliant construction produces an identical velocity curve — stack your pre-registration cohort, waitlist email, embargoed coverage, creator posts and paid launch on a single date and the algorithm sees the same spike, from real people.