Qaf · Monetization Strategy · Confidential Working Draft
How to brand, price, and place Qaf's paid tier — and how to launch it around Ramadan 2027 — grounded in Qaf's own beta analytics and verified market evidence.
Direct answers to the five questions asked, with the evidence behind each in the sections that follow.
Yes as the wrapper, no as the substance. Sell a real product (Qaf Pro) whose paid value is capability — a deep-research answer mode, not permission to ask — and wrap it in honest mission language: "Your membership keeps Qaf free for a million seekers of knowledge." Never use the words donate, donation, sadaqah, or zakat for payments to a for-profit: Apple prohibits for-profit donation flows in-app, a company is not zakat-eligible, and implied-charity framing invites a Muslim Pro-scale trust collapse when users remember Qaf is a company. Publish a Hallow-style "Why does Qaf charge?" page and real cost transparency instead.[7][10]
One paid plan at launch — Qaf Pro at $9.99/month or $59.99/year (annual pre-selected) — plus two zero-build-cost wrappers: a one-time Founding Member offer for the beta cohort (~$199 lifetime, fair-use capped) and a Tarteel-style pay-what-you-can hardship program that answers "never forced to pay" better than any price cut. Add a Family plan (~$99/yr, 5–6 seats) as the first fast-follow. Paywall data consistently shows fewer visible options convert better; a second feature tier is only worth revisiting once Deep Study usage data justifies a ladder.[3][12]
Yes — ship it, with three guardrails. It sits between Duolingo (1 week) and Perplexity (1 month); at ~$0.01/question it costs $0.02–$0.35 per grant against a $30–80 industry cost to acquire a subscriber. Make it two-sided and gift-framed ("give your friend two weeks of Pro"), credit it only after an activation event (invitee asks 3+ questions within 7 days), and cap it (~24 weeks/user/year). One prerequisite from your own data: referral attribution is currently broken — 51 users shared links last month and only 1 attribution was recorded.[13][14]
Never in front of an answer. Free tier gets a monthly meter (60–90 questions/month — untouched by 91–95% of active users) that degrades gracefully to a lighter model rather than blocking. Paywalls appear at three moments: a skippable multi-page intro of Pro at the end of onboarding (~50% of all conversion volume industry-wide), the moment the meter approaches, and the moment a question deserves Deep Study. Your guest wall already proves the mechanic: 228 guests hit the 3-question wall last month — that moment, moved up-market, is the paywall.[4][5]
Both, in a specific shape: a reverse trial plus a native intro trial on annual only. Give every new user 14 days of full Pro automatically, then downgrade to free — no card, no dark pattern, and it neutralizes trial abuse (Toggl doubled premium revenue this way). Keep a 7-day card-on-file trial on the annual plan via App Store / Play intro offers, and require a card on web trials. Promise — and send — a reminder before any charge: Blinkist's transparent trial lifted trial starts 23% and cut complaints 55%, and billing honesty is amanah for this audience.[6]
The unifying principle: Qaf sells scholarship-grade research tools and capacity — it never sells access to Islamic knowledge. Every pricing, paywall, and branding decision below is that one sentence applied.
Mixpanel, 30-day window ending July 7, 2026. 2,500-person open beta; volume became meaningful June 8.
Usage is a textbook power law, and that is the entire free-tier problem. The median asker sends 4 questions a month; the 95th percentile sends ~88; the single heaviest user sent 1,571. The top 1% of askers generate roughly a third of all serving cost, and the bottom half generates 4%. A pricing design that treats "the average user" as the unit of analysis will be wrong on both ends.
Exhibit 1
Half of users ask fewer than 5 questions a month; 9% of users create two-thirds of the cost
Distribution of questions per asker, trailing 30 days (n = 1,662). Highlighted: the 60+ bucket — 148 users, ~26,100 questions.
The 60+ bucket averages ~171 questions/user/month (p95 ≈ 88, p99 ≈ 255, max 1,571). Source: Mixpanel frequency-per-user histogram, chat_message_sent.
The habit is weekly, not daily. Askers are active ~2.2 days per week (DAU/WAU ≈ 0.31) and usage inside an active day is bursty — median 3, p90 15. People sit down and research a question. This favors a monthly allowance over a daily cap (a daily cap punishes exactly the natural burst behavior while missing the every-day whale), and it favors trials of 14+ days, which span 4+ usage occasions.
Exhibit 2
The beta grew ~10x in eight weeks, and cohorts keep coming back
Weekly unique askers and questions asked (left); average cohort retention by weeks since first question (right).
Weekly askers (final partial week omitted).
% of cohort asking again, by week. Avg across weekly cohorts.
Payers and cost live in different places. US/UK/Canada supply 36% of askers (591 of 1,662) but only ~25% of question volume — Western users actually ask ~30% fewer questions than average. The heavy tail lives in Pakistan (51.6 q/user), India, and European diaspora pockets. Two consequences: the free tier is largely a da'wah subsidy to the rest of the world (which makes the support story true), and upgrade prompts fired at cap-hitters will mostly reach low-willingness-to-pay geographies — so forecast conversion from heavy users in payer geos only.
Exhibit 3
The UK, not the US, is Qaf's largest market today — and payer geos are lighter users
Monthly askers by country, top 12. Highlighted: US / UK / Canada — the expected payer core (36% of askers, ~25% of volume).
Platform mix: iOS 48%, Android 38%, web ~18% of askers (multi-device overlap). Caveat: ~50% of message volume reports platform=undefined — fix before per-store pricing analytics.
Three found assets, one found bug. The guest experience already has a working conversion wall (3 questions/day; 228 guests hit it last month and the next step is sign-in). A referral screen already exists (270 viewers → 51 sharers last month) — but only one referral was ever attributed, so attribution is effectively broken and must be fixed before the referral program launches. And citations are the product's felt differentiator: 37% of askers opened a citation last month — the thing free AI chatbots don't give this audience is exactly the thing to build Pro around.
Modeled at 1M registered users using the beta's observed usage distribution (Pareto tail α ≈ 1.5), across monthly-asker scenarios of 150k / 250k / 400k and the beta's own 66% activation as the stress case.
At today's ~$0.10 per question, no free tier survives contact with 1M users. Uncapped, realistic scenarios burn $300–960k per month. Even a stingy 15-questions-per-month cap costs over $1M a year at modest activation. The entire monetization design is therefore sequenced on the cost curve: generosity is a function of unit cost.
Exhibit 4
A 60–90 question monthly meter keeps free serving inside the paid-revenue envelope — at $0.01 per question
Annual free-tier serving cost at 1M registered users, by free-tier monthly cap and monthly-asker count, at $0.01/question. Recommended zone highlighted.
| Free cap / month | % of askers ever capped | 150k askers | 250k askers | 400k askers | 660k (beta-rate) |
|---|---|---|---|---|---|
| Uncapped | 0% | $426k | $709k | $1.14M | $1.87M |
| 300 (≈10/day) | 0.8% | $343k | $572k | $915k | $1.51M |
| 90 (≈3/day) | 4.8% | $273k | $455k | $728k | $1.20M |
| 60 (≈2/day) | 8.9% | $238k | $396k | $634k | $1.05M |
| 30 (≈1/day) | 16.4% | $175k | $292k | $467k | $771k |
At today's $0.10/question, multiply every cell by 10. Paid-revenue envelope at 10k subscribers ≈ $735k/yr net (Section 09). Model: beta usage distribution with capped Pareto tail; caps rarely bind below the 90th percentile.
The honest cross-subsidy arithmetic — which is also the marketing copy. At the recommended pricing, one Pro member nets Qaf ~$6.10/month after store fees. At $0.01/question that funds ~600 free answers every month — a month of Qaf for somewhere between 25 heavy users and 150 median users. "One membership keeps dozens of students of knowledge asking freely" is not framing; it is the unit economics, and publishing it is the strongest honest version of the support story.
Tie the public generosity of the free tier to realized unit cost, not to the calendar: launch the meter at 30–45 questions/month while cost per question is above ~$0.03, move to 60–90 as it approaches $0.01, and advertise increases loudly (generosity ratchets up; it must never visibly ratchet down). Route free-tier questions to efficient models first — the flagship model is a Pro feature, graceful degradation is the free tier's safety valve.
The question isn't whether support-framing works — it's what it must be wrapped around, and which words are radioactive.
Position: "Qaf is free for the ummah, and it always will be. Pro members get scholarship-grade research tools — and they're the reason it stays free for everyone else." Both halves must be literally true, and Section 03 shows they are.
Users will benchmark Qaf against Hallow, Tarteel, and Muslim Pro — not against ChatGPT. Price in the faith-app band, differentiate with AI capability.
Exhibit 5
$59.99/year sits in the proven faith-app band, at half the price of the $20/mo AI assistants
Annual subscription price, US, mid-2026 (verified against official pricing pages). Highlighted: proposed Qaf Pro.
Tarteel also sells a $250 lifetime; Hallow adds Family $119.99 (6 seats); Muslim Pro's low annual reflects an ad-supported free tier. RevenueCat's cross-category "structural anchor" is $9.99/month.
Exhibit 6
What each rail nets on a $9.99 subscription — and why not to over-steer
| Rail | Fee | Net to Qaf | Note |
|---|---|---|---|
| Stripe web (with Billing) | ≈6.6% | $9.33 | Fixed $0.30 fee penalizes cheap monthlies — default web checkout to annual |
| Apple IAP — Small Business Program | 15% | $8.49 | Applies under $1M/yr proceeds; at the recommended mix Qaf stays eligible until ~15–20k subscribers |
| Google Play (from Jun 30, 2026) | 15% | $8.49 | External-offer link-outs save only the 5% billing fee — Android steering isn't worth the friction |
| Apple IAP — standard | 30% | $6.99 | The post-$1M world; plan for it in year 2 |
US iOS apps may link to external checkout at 0% today (post-Epic contempt ruling) — but the Ninth Circuit has authorized a future cost-based fee, so treat it as temporary. Web checkout converts at only ~63–67% of IAP, so while on the 15% rate, hard steering loses net revenue: ship IAP as the primary CTA with a same-price "subscribe on the web" secondary link (US only; Canada must stay IAP-only in-app; UK likely opens late 2026–2027).[15]
Two rail-specific notes. First, the choose-your-amount supporter checkout and any "support Qaf" page must live web-only — for-profit fundraising flows are prohibited in iOS apps. Second, implement referral Pro-time and all entitlements server-side, independent of either billing rail, so earned weeks work identically for IAP and Stripe subscribers.
Freemium costs Qaf the hard-paywall conversion premium (10.7% vs 2.1% median download-to-paid) by design. The compensation is placement discipline: capture the Day-0 window without ever blocking an answer.
Exhibit 7
The paywall copy formula that fits both the data and the mission
Assembled from tested patterns: benefit-led bullets, goal-first framing (+27% trial starts), specific social proof above price, annual pre-selected with per-month framing, cancel-anytime above the fold, and the Blinkist reminder promise (+23% trial starts, −55% complaints).
| Slot | Recommended copy direction |
|---|---|
| Headline | "Go deeper into the sources with Qaf Pro" — capability first, not "unlimited questions" |
| Mission line | "Your membership keeps Qaf free for a million seekers of knowledge." |
| Benefits (4–6) | Deep Study answers with full evidence chains · every madhhab, compared · voice & saved research · priority in Ramadan nights |
| Social proof | "Join 12,000 members keeping Qaf free for everyone" (real number, above the price) |
| Price block | Annual pre-selected: "$4.99/mo billed annually — save 50%" · monthly beside it |
| Trust footer | "Cancel anytime. We'll remind you before any charge. We never sell your data." |
New users start in Pro for 14 days, then land on the free tier. This is the documented best fit for a generous-freemium mandate: loss aversion does the selling (10–40% conversion lift in reverse-trial case work; Toggl doubled premium revenue and killed trial-cycling abuse, since there's nothing left to farm). Qaf's weekly usage cadence argues for 14 days over 7 — that spans ~4 usage occasions. Cost is bounded: a p90 user's 14 Pro days cost ~$0.35 at $0.01/question.[6]
Subscription-time rewards are structurally fraud-resistant and nearly free for Qaf. The risk isn't cost — it's a reward nobody wants because Pro doesn't feel different yet.
(1) Fix attribution first: 270 users viewed the referral screen and 51 shared last month, but exactly one referral_attributed event exists — the loop can't be measured, let alone rewarded, until install-time attribution works. (2) Ship the Pro delta first: free weeks of a tier whose only difference is "more questions you weren't running out of" is worthless currency (the Duolingo lesson). Deep Study must exist before referral rewards mean anything.
Hallow earns roughly a quarter of its annual net revenue in the single post-Lent month. Ramadan (Feb 8 – Mar 9, 2027) is structurally stronger — 30 nights of engagement on a fixed calendar — and Islamic apps reliably see 1.6–3x download spikes.[9][16]
Deep Study mode, monthly meter (remotely configurable), server-side entitlements across IAP + Stripe, referral attribution fix, platform-tracking fix, Van Westendorp + MaxDiff survey on the beta cohort (n ≥ 200, oversample US/UK/CA).
Quiet rollout: debug billing on both rails, publish "Why does Qaf charge?", open the Founding Member offer to the 2,500 beta users, start the reverse trial for new signups, enroll in Apple's Small Business Program. Honest baseline prices — no discounts yet.
Pro free for all during Sha'ban (no card), enrollment opens for the Qaf Ramadan challenge (khatam / daily-learning format with named scholars). Habit forms in the prep month; the paid decision lands during Ramadan at peak intent — not after Eid in the lull.
Challenge live (free for everyone — the mandate holds), referral gifting push, annual plan pre-selected: an annual bought in Sha'ban or early-mid Ramadan renews inside the next pre-Ramadan window — the best possible renewal timing (Hijri drift makes this a 1–2 cycle bonus, so bank it early). Pre-scale serving for a Hallow-grade spike.
Dhul-Hijjah 10-day challenge (~May), weekly Jumu'ah / al-Kahf habit loop, win-back automation for monthly churners (~20% reactivate within a year; annual churners almost never return, so pre-renewal engagement matters more). Do not read Shawwal churn as steady-state.
Launching paid after Ramadan 2027 would push the category's one compounding month into a window with 1.6–7x less traffic. The date is the strategy.
At 1M users, 1% paid, recommended pricing ($9.99 / $59.99, ~60% annual mix, ~70% of revenue via IAP at the 15% small-business rate).
Exhibit 8
Conversion sensitivity: 1% covers the AI bill; 1.5–2% funds a business
Net annual subscription revenue at 1M registered users vs total AI serving cost (free + paid, base case), at $0.01/question.
Serving band shown as reference rows in the table view. Above ~$0.03/question unit cost, every scenario is underwater — the launch gate is the cost curve, not the calendar. Upside levers not modeled: Supporter choose-your-amount uplift (Bandcamp's name-your-price runs ~+50% over the floor), Family plan, price at $69.99, reactivation.
Read it straight: at the base case, subscriptions (~$735k net) cover total AI serving (~$380–745k) with thin headroom — monetization v1 makes Qaf's free mission self-sustaining, which is exactly the support story, but it does not yet fund payroll. The realistic paths to margin: convert the Western cohort at 2–3% (they benchmark at 2.8% median; Duolingo shows values-driven freemium can double penetration over time), land the cost curve below $0.01, and let annual mix + Family + Supporter pricing lift ARPU. A 1.5% blended conversion at $69.99 annual is a ~$1.3M-net business on the same cost base.
Exhibit 9
Top risks and their mitigations
| Risk | Exposure | Mitigation |
|---|---|---|
| Unit cost stays > $0.03/question | Free tier structurally unaffordable at scale | Gate generosity on realized cost; launch meter at 30–45/mo; route free tier to efficient models; remote-config caps |
| Ramadan spike hits costs before revenue | 25x traffic on day 1 of campaign | Pre-scale; free-tier depth (not access) tightens during spike; annual-default paywall converts the wave |
| "Paywalling the deen" backlash | Category-defining trust risk | Never gate sources/answers; hardship program; no ads; no donation vocabulary; "Why we charge" page; never retro-gate beta features |
| AI-app annual retention (21.1% vs 30.7% non-AI) | Year-2 revenue cliff | Habit surfaces (saved research, study plans, Ramadan cycles); pre-renewal engagement; win-back for monthly churners |
| Apple fee cliff (15% → 30% past $1M proceeds) | ~$0.9k/yr per 1k subs | Annual + web mix keeps Qaf under threshold until ~15–20k subs; grow web-native acquisition before the cliff |
| Meter converts the wrong geography | Cap-hitters skew non-payer geos | Forecast from payer-geo p95 users only; treat ROW over-cap volume as mission subsidy, not lost revenue |
Internal data: Mixpanel project 4009978, trailing 30 days ending Jul 7, 2026 (events: chat_message_sent, chat_rate_limit_hit, referral_*, app_opened; frequency-per-user histograms and retention curves). Beta caveats: opt-in enthusiast cohort (66% monthly activation will not hold at scale); signed-in users currently uncapped, so usage reflects true unconstrained demand; iOS store attribution sparse. External research: 13 parallel research workstreams (Islamic apps, faith apps, AI assistants, subscription benchmarks, paywall placement, referrals, support-framing, trials, price architecture, billing rails, usage economics, Pro contents, Ramadan timing) executed with live web research in July 2026; 59 load-bearing claims were then independently re-verified against primary sources — 49 confirmed, 8 corrected or updated (corrections applied throughout, e.g., Hallow's #1-overall App Store moment was Feb 2024 not 2025; Perplexity's give-get referral program ended Oct 2025; Tarteel's current App Store monthly is $12.99 with ~$99 annual), 2 unverifiable (excluded).