ReplyMeter

1. One-liner

Prices every WhatsApp reply your agents send, so the 1 October bill stops being a surprise.

2. Trend signal — why now?

On 1 October 2026 — 33 days from today — Meta starts charging for service messages on the WhatsApp Business Platform. Free-form replies typed by a human agent or a third-party AI bot inside the open 24-hour customer service window have been free since November 2024. From October they bill per message, at the same per-message rate as utility and authentication templates in the recipient’s country. Utility messages sent inside that same window lose their free status too.

Three details turn this from a price rise into a business.

One: service messages get no volume discount. Utility and authentication templates accrue toward Meta’s volume tiers and get cheaper as you scale. Service messages don’t. Every reply bills at the standard country rate forever, no matter how many you send. The one line item a growing support team can’t outgrow is the one that scales linearly with how chatty its agents are.

Two: the unit of cost changed from the conversation to the keystroke, and nobody has ever measured keystrokes. Under the old model, a customer conversation was a conversation — one window, one charge, and whether your agent resolved it in one message or eight was a UX question, not a finance question. Meta’s own illustration of the change shows a single customer interaction going from 3 charges to 5 charges for the identical conversation. Support teams have spent two years being trained to do exactly the wrong thing: split a reply into three short bubbles because it feels more human and reads faster on a phone. Every one of those bubbles is now a line item. No support platform on earth reports “average billable replies per resolved conversation,” because until 1 October that number cost nothing.

Three: Meta isn’t publishing the country rate card until 1 September 2026. Asksuite, writing to hotels, put it plainly: Meta “will definitively confirm the rate above by country by September 1, 2026, just 1 month before the change.” So every brand on WhatsApp is going into October with a cost they cannot yet model, on a behaviour they have never measured. That is a very specific, very short, very loud window of demand.

The rates that are public are wildly uneven, and that decides who feels this. Utility/auth per-message rates as of July 2026: India $0.0014, US $0.0034, Brazil $0.0068, Mexico $0.0085, Saudi Arabia $0.0107, UAE $0.0157, UK $0.0220, Indonesia $0.0250, Germany $0.0550. A German brand handling 20,000 support conversations a month at six agent replies each is looking at roughly $6,600/month of cost that did not exist in September. The same brand in India pays about $168. This is not a global panic — it’s an acute, expensive problem in a specific band of countries, which is exactly the kind of segment a solo operator can own.

And the money is real enough that the ecosystem is already selling against it. Asksuite told its hotel customers that with its first round of optimizations alone, “the average savings for customers is already 40% of the new cost.” Forty percent. That’s a vendor with skin in the game publicly conceding the waste is enormous.

The tell that this is unbuilt: every BSP blog post says the same thing — “pull your last 30–60 days of message volume broken out by category so you know your actual exposure rather than guessing at it” — and then none of them hands you the tool that does it. YCloud tells businesses to “update cost forecasts” and supplies no benchmark to forecast against. Wati frames itself as the answer while quietly recommending you “resolve in fewer turns.” Everyone has identified the metric. Nobody ships it.

Provenance:

• Signal 1 (demand): Meta charges for service messages and in-window utility messages from 1 Oct 2026; service messages get no volume discount; Meta’s own example shows the same conversation going from 3 to 5 charges; exact country rates not published until 1 Sept 2026 — https://chakrahq.com/article/whatsapp-api-pricing-update-service-messages-october-2026/ — observed 2026-08-29
• Signal 2 (feasibility): Per-country utility/auth per-message rates now published on a fixed rate card (India $0.0014 → Germany $0.0550), and service messages bill at that same rate — making a per-reply cost meter a pure arithmetic problem over webhook data — https://sleekflow.io/blog/whatsapp-business-price — observed 2026-08-29
• Signal 3 (economic): BSPs layer $0.003–$0.010 per-message markup on top of Meta’s card and reseller models run 75% gross margin plus 8–15% on message-wallet topups — money is moving per message, and the incumbents profit from the message count they’re supposedly helping you cut — https://richautomate.in/blog/white-label-whatsapp-bsp-india-2026-reseller-playbook — observed 2026-08-29
• Signal 4 (corroboration): Asksuite tells hotel customers its first optimizations alone save “40% of the new cost”; Intelli models 10,000 conversations/month at four replies each and flags that per-country rates are unconfirmed — https://asksuite.com/blog/new-charges-whatsapp-api-asksuite-hotels/ — observed 2026-08-29 Category: Platform shift

3. The opportunity

The gap is a conflict of interest, and it’s a clean one.

Your BSP — Wati, AiSensy, Interakt, 360dialog, Twilio, Gupshup, whoever — is the only party today with the data to tell you that your agents average 6.4 billable replies per resolved ticket, that your returns flow burns 11, and that Priya on the late shift sends four bubbles where Rahul sends one. They will not build that report with any teeth, because a large slice of the BSP business model is a $0.003–$0.010 markup on every message you send, plus 8–15% on wallet topups. The reseller playbook the Indian BSP channel circulates openly is ₹999/mo wholesale resold at ₹3,999/mo — 75% gross margin — plus the topup margin. Asking a per-message reseller to help you send fewer messages is asking them to shrink their own invoice.

So what they ship instead is the upsell-shaped version of cost control: “deploy our AI agent,” “buy our Flows builder,” “move to our economy mode.” All of which may genuinely help — and all of which require you to buy more of them before you can find out whether you had a problem. Nobody sells you the measurement first.

That’s the wedge. ReplyMeter is a read-only meter that sits beside whatever BSP you already use and prices your actual traffic. It ingests your message webhooks, reconstructs conversations, classifies each outbound message into its billing category, applies Meta’s published country rate card, and hands you three numbers nobody currently has: what October will cost you at current behaviour, which flows and which agents generate that cost, and what it drops to if the average reply count falls by one.

Then it does the part that makes it sticky: it keeps metering after you change things, so the saving is provable rather than claimed. When your BSP’s account manager tells you their new AI tier will cut your bill 40%, ReplyMeter is the referee that tells you whether it did.

The 10× isn’t AI cleverness. It’s alignment. A tool billed at a flat monthly fee that gets more valuable the more money it saves you is structurally a different product from a tool billed per message. You cannot get here from a BSP’s P&L.

4. Target market

Primary customer: Head of Customer Support / Head of CX at a consumer brand running human-staffed WhatsApp support at 3,000–50,000 customer-initiated conversations per month, headquartered in or serving a high-rate country — Germany, UK, UAE, Saudi, Indonesia, Netherlands, Nordics. Typical shape: 30–400 employees, 5–40 support agents, e-commerce / travel / hospitality / education / clinics. They already pay a BSP ₹1,500–₹15,000/mo (or $49–$300/mo) in platform fees, and a separate, larger Meta message bill on top.

Secondary: agencies and BPOs running WhatsApp support for 5–50 client brands. Their cost exposure is multiplied and their clients will demand line-item explanations in November. One agency seat covers many WABAs — better ACV, faster reference loops.

Why they buy — in their own words: the ecosystem quotes are all variations of the same sentence. Chakra’s guidance: know “your actual exposure rather than guessing at it.” Intelli, on July 2026: “Specific per-message rates for each country have not been confirmed.” Asksuite, to hoteliers: Meta confirms rates “just 1 month before the change.” The buyer’s felt state is not “I want analytics.” It’s “my November invoice is going to have a number on it I cannot explain to my CFO, and I have four weeks.”

Rough TAM reasoning: Meta reports the WhatsApp Business Platform is used by millions of businesses, but the relevant slice is much smaller and knowable: businesses running the API (not the free Business App) with human agents in high-rate countries. BSP customer counts give the shape — Wati alone publicly markets to tens of thousands of businesses, AiSensy similar, and there are dozens of BSPs. A defensible working estimate is 150,000–400,000 API-active businesses globally with real human support volume; the high-rate-country, 3k+ conversations/month band is plausibly 20,000–60,000 of them. At a $99–$399/mo price point, capturing 0.5% of the lower bound is comfortably past $1M ARR. I don’t need the number to be right to within 2× — I need it to be an order of magnitude bigger than the customer count that gets me to $2M ARR, and it is.

Why now for them: their cost structure changes on a fixed date they didn’t choose, the rate card lands 30 days before, and the behaviour driving the cost is invisible in every tool they own.

5. Product sketch (MVP)

• Connect your WABA in one step — read-only webhook subscription or BSP API key. No migration, no number porting, no touching your inbox. You keep Wati/AiSensy/360dialog exactly as-is.
• October forecast — replays your last 60 days of traffic against the post-1-October billing rules and Meta’s published country rate card, and shows the single number the CFO will ask for: your bill would have been $X higher. Split by old rules vs new rules so the delta is unarguable.
• Cost per conversation, cost per flow — every resolved conversation gets a price tag. Sorted descending, so the returns flow that burns 11 replies to do what a Flow could do in one is at the top of the list on day one.
• Agent reply-habit board — billable replies per resolved conversation, per agent, per shift. Not a surveillance tool — a coaching number, framed as “resolved in fewer turns,” with resolution time and CSAT-proxy shown alongside so nobody optimizes into worse service.
• Bubble-splitting detector — flags outbound messages sent within N seconds of each other to the same customer by the same agent. This is the single biggest cheap win: three bubbles collapsed into one is a 66% cut on that reply, with zero change to what the customer is told.
• Window-waste alerts — messages sent just after a customer service window closed (now a template charge), free-entry-point windows from Click-to-WhatsApp ads left unused while they were still free, and utility templates fired inside a window that could have been a service reply or vice-versa, whichever is cheaper in your country.
• Before/after proof — pick a change (new macro, new Flow, a coaching push), and ReplyMeter reports the measured cost delta for 30 days after. Screenshot-ready for the board deck.
• Weekly cost email — spend to date, run rate against forecast, top three cost regressions. The artifact that keeps a support lead logging in after the October panic passes.

6. AI angle — what’s load-bearing

Strip the AI out and this product still exists as a spreadsheet — a worse one, but it exists. So I’m going to be honest about where AI is genuinely doing work and where it’s decoration, because the rubric punishes bluffing here.

Load-bearing, genuinely:

Reply consolidation rewriting. Given an agent’s actual three-bubble sequence, generate the single-message version that says the same thing, in the same tone, in the same language. This is the product’s core saving mechanism, it runs over messy multilingual real-world chat (Hinglish, Portuguese, German, Bahasa, Arabic), and it produces a concrete artifact — a suggested macro — not advice. Doing this with rules is hopeless; doing it with a small model over conversation transcripts is trivial and cheap.

Intent-level conversation clustering. “Where is my order,” “wrong size,” “cancel my booking” — grouping thousands of conversations by what the customer actually wanted, so cost can be attributed to problem types rather than to agents. This is what turns the meter into a roadmap: “your top three intents are 41% of your service-message spend and all three are Flow-able.” Embeddings + clustering, off the shelf, but there’s no non-AI version that works across languages and phrasings.

Flow candidate detection. Reading a cluster of expensive conversations and proposing the specific interactive Flow that would collapse it — which fields to collect, in what order. The output is a spec a human implements in their BSP’s builder.

Not load-bearing, and I won’t pretend otherwise: the metering, the rate-card arithmetic, the forecast, the agent leaderboard. That’s deterministic accounting over webhook data, and it’s most of the perceived value in month one. The AI is what stops the product being a dashboard you check once and abandon — it converts observation into a change you can ship.

7. Localization angle (if any)

Inverted from the usual India-first instinct, and deliberately so.

India is the worst launch market for this despite being the biggest WhatsApp market: at $0.0014 per service message, a brand doing 10,000 conversations a month at four replies each faces about $56/month of new cost. Nobody buys a $99/mo tool to manage a $56/mo problem. The India pain is real at very high volume only.

The product wants high-rate, high-labour-cost countries: Germany ($0.0550), Indonesia ($0.0250), UK ($0.0220), UAE ($0.0157), Saudi ($0.0107). Germany is the standout — 39× India’s rate — and German mid-market e-commerce is both WhatsApp-heavy and famously cost-disciplined. Indonesia is the interesting one: a high rate ($0.0250) attached to a market where WhatsApp is the commerce channel and support conversations run long. A Jakarta seller doing 20,000 conversations at six replies is looking at $3,000/month appearing from nowhere.

Localization that actually matters: the rate card is the localization. The product must ship every country’s rate correctly on 1 September, handle multi-country customer bases in one account (a UAE brand serving Saudi, Kuwait and Egypt pays three different rates in one inbox), and reason over conversation transcripts in German, Bahasa, Arabic and English. Currency display in local terms and invoicing in EUR/GBP/AED matter more than any UI translation.

So: global product, deliberately sequenced by rate card. Germany/UK/UAE first, Indonesia and Brazil second, India last and only for the 100k+ conversation tier.

8. Business model — path to $1M–$5M ARR

Pricing — flat, never per-message. The whole pitch collapses if my invoice grows when your volume does. Tiered by conversation volume, priced as a fraction of measured savings:

Tier	Conversations/mo	Price
Starter	up to 5,000	$99/mo
Growth	up to 25,000	$249/mo
Scale	up to 100,000	$599/mo
Agency	10 WABAs, pooled	$899/mo

The sanity check that makes this an easy yes: a UK brand at 25,000 conversations × 5 replies × $0.022 = $2,750/month of new service-message cost. Asksuite’s own published figure says optimizations save ~40%. That’s $1,100/month saved against a $249 bill. A 4.4× return, and the customer keeps the receipts because ReplyMeter measures the before and after. In Germany the ratio is absurd — the same brand faces $6,875/month and pays $249.

ACV: blended ~$3,400/year. Agency accounts pull it up; Starter pulls it down.

Path to $1M ARR: ~295 customers at blended ACV. Realistically: 120 Growth ($249) + 60 Scale ($599) + 25 Agency ($899) + 200 Starter ($99) ≈ $1.06M. That’s a bootstrapped-scale number, not a venture number, which is the point.

Path to $5M ARR: needs ~1,400 blended customers, which means the product can’t stay a one-shot October panic-buy. Two things must be true: (a) the meter becomes the standing system of record for WhatsApp spend — the weekly email, the QBR chart — so it survives into 2027; and (b) the agency/BSP-reseller channel works, because 50 agencies at 20 client WABAs each is 1,000 brands acquired through 50 relationships. Plausible, not assured. I’d underwrite this at $2–3M and treat $5M as upside.

Expansion path: WABA count → conversation volume tiers → RCS and Instagram DM metering (same billing-shift logic, same buyer) → a “spend guarantee” tier where you pay a % of measured savings above a floor. The last one is where ACV really moves, but it needs a year of trustworthy measurement first.

9. Go-to-market wedge — first 100 customers

The date does the selling. I have a 33-day runway before the change and a ~60-day window after where every support lead in a high-rate country is being asked “why did the WhatsApp bill go up.” The playbook is built around that.

• Free “October Shock Calculator,” shipped by 5 September (four days after Meta publishes the rate card). Public page: enter your country, monthly conversations, and average replies per conversation → get your projected new monthly cost and a per-reply breakdown. No signup for the number; email required for the PDF with a saving plan. This is the highest-leverage asset in the plan because it is the exact calculation the entire BSP ecosystem told people to do and then didn’t build. Seeded into the ~15 BSP blog posts and comparison sites already ranking for “WhatsApp service message pricing October 2026” via direct outreach to their authors — they need a tool to link to and they don’t want to build one. Target: 3,000 calculator runs, 600 emails, 40 paying in the first 45 days.
• Scrape the buyers directly. Public WhatsApp Business API directories, Shopify app-store review pages for the major WhatsApp apps (Wati, Interakt, AiSensy, Bitespeed, Zoko each carry hundreds of public reviewer store names), and Click-to-WhatsApp advertisers visible in the Meta Ad Library filtered by country. Build a list of 2,000 brands in Germany, UK, UAE and Indonesia running WhatsApp support. Cold email with their own estimated October delta computed from public signals (country + observable ad volume + category benchmark) in the subject line: “Your WhatsApp bill goes up ~€4,100/mo on 1 October.” A specific number in a cold subject line is a different animal from a pitch. 2,000 sends × 8% reply × 25% demo × 40% close ≈ 16 customers. Repeat monthly with fresh geos.
• Agencies and BPOs first, because they multiply. ~200 identifiable agencies running WhatsApp CX for multiple brands (BSP partner directories list them publicly). Pitch: a white-labelled October forecast for every client, sent under the agency’s logo — turning a bad-news platform change into a proactive client touchpoint they look smart for. Free for 30 days, $899/mo after. 200 outreaches × 15% trial × 50% conversion ≈ 15 agency accounts covering 200+ brands.
• Be the neutral referee in the incumbents’ own comment sections. Every BSP is publishing “here’s what changes in October” content and getting operator comments underneath asking “how do I know what mine will be?” That question has no honest answer from a per-message vendor. Answer it, publicly, with the calculator link. Cheap, fast, and the conflict-of-interest framing is genuinely true, which is why it lands.
• A single benchmark report, published mid-November. “What 300 brands actually paid for WhatsApp service messages in October” — median replies per conversation by industry and country. Nobody else can publish this because nobody else is metering across BSPs. It’s the credibility asset that carries acquisition through Q1 2027 when the panic has faded.

10. Build complexity — justification

Low. The MVP is a webhook listener, a conversation reconstructor, a rate-card table, and a reporting UI. The WhatsApp Cloud API already emits every inbound and outbound message with timestamps, message IDs, and pricing metadata; reconstructing conversations and window state from that stream is bookkeeping, not research. The rate card is a published table. The AI pieces (consolidation rewriting, intent clustering) are off-the-shelf model calls over short text.

The genuinely fiddly parts are three: correctly modelling window state and billing category under rules that Meta finalizes on 1 September; supporting ingestion from several BSPs whose webhook relay behaviour differs (a couple don’t forward outbound events cleanly, which will force a per-BSP integration matrix); and being scrupulously accurate — a cost meter that’s wrong by 15% is worthless and the reputational recovery is brutal.

Estimate: 6–8 weeks to a paid v1 for one or two BSPs plus direct Cloud API, by 1–2 people. The calculator ships in a weekend and can go live the day Meta publishes rates.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Read-only use of the customer’s own WABA data with their consent. Standard processor terms; GDPR-relevant since chat content is personal data, so EU hosting and a DPA from day one.
Ethical — no harm / dark patterns	✅	One live risk: the agent leaderboard can become a surveillance stick. Mitigated by design — report at team and flow level by default, per-agent gated behind an explicit admin toggle, and always paired with resolution-quality metrics so nobody is rewarded for terse, unhelpful service.
Market exists (evidence above)	✅	Dated, mandatory platform change affecting every API business; entire BSP ecosystem publishing about it; a competitor publicly claiming 40% savings.
1–5 person team can build this	✅	1–2 people, 6–8 weeks.
Launchable with <$50K / ₹40L	✅	Well under $10K. Hosting, model inference on short text, and a landing page.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	15/20	Real, dated, and lands on a P&L line — but it’s a cost problem, not an outage or a fine. A German brand facing $6,600/month is hair-on-fire; a US brand at $0.0034 is mildly annoyed. Intensity is concentrated in a subset of countries, which caps this below 17.
Demand evidence	15	12/15	Strong indirect evidence: mandatory dated change, a dozen-plus BSPs publishing urgent guidance, Asksuite quantifying 40% savings, Meta’s own 3→5 charge example. Docked for the honest gap — I found no verbatim buyer complaints, because the change hasn’t hit a bill yet. Demand is inferred from the mechanism, not observed in the wild.
Build feasibility	15	13/15	Webhooks, arithmetic, a rate table, light model calls. 6–8 weeks for 1–2 people. Docked for per-BSP webhook variance and the accuracy bar.
Distribution clarity	15	12/15	Named lists (Shopify app reviewers, Meta Ad Library CTWA advertisers, BSP partner directories), a genuinely linkable free tool answering a question the whole ecosystem is asking, and a date forcing the conversation. Docked because cold email to support leads is unproven for this category and the calculator’s SEO window is short.
Revenue mechanics	15	11/15	Pricing is easy to justify — $249 against $2,750 of new cost is a trivial yes, and flat pricing is a differentiator not a concession. Docked because ACV is modest, churn risk after the October panic is real, and $5M requires the agency channel to work.
Time to first revenue	10	8/10	Calculator live ~5 September, paid v1 by mid-October, first invoices within 6 weeks. Not instant only because the product must exist before the panic peaks.
Defensibility	10	3/10	This is the weak axis and I won’t dress it up. The metering is copyable in a month. The only durable assets are cross-BSP benchmark data (which compounds but slowly) and the structural conflict that stops per-message BSPs from doing this credibly — real, but it doesn’t stop a neutral third party cloning it. Execution and speed moat only.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · content-heavy

Technical because it’s an integration and accuracy product. Content-heavy because the entire wedge is a free calculator plus a benchmark report ranking against a dozen BSP blog posts — distribution is content here, and it has a 60-day peak.

Key assumptions to validate (3–5)

1. Assumption: Brands in high-rate countries will pay a flat monthly fee to a third party for visibility their BSP nominally already has. How to test: 40 cold outreaches to German/UK support leads in the first two weeks of September with a manually-computed forecast; measure how many book a call and how many pre-pay before the product exists. Pre-payment is the only real signal.
2. Assumption: Average billable replies per conversation is high enough (≥4) that consolidation produces a saving worth a subscription. How to test: get read-only access to 5 friendly brands’ 60-day message history and compute the actual distribution. If the median is 2, the saving is too small and the product is dead.
3. Assumption: BSP webhooks reliably expose outbound agent messages with enough fidelity to price them. How to test: technical spike against Cloud API direct plus three named BSPs before writing any UI. If two of three don’t forward outbound events, the addressable market shrinks to direct-Cloud-API brands.
4. Assumption: The buyer keeps paying past November, once the shock is absorbed. How to test: can’t be tested pre-launch — instrument it. Track 60-day and 120-day retention of the first cohort as the primary kill metric.
5. Assumption: Meta’s 1 September rate card matches the widely-reported structure (service = utility/auth rate, no volume tiers). How to test: read it on 1 September. If Meta softens the change — caps it, grandfathers it, or bundles service messages back into a window charge — the urgency evaporates and this becomes a VALIDATE at best.

Risk flags

1. Platform dependency — total, and this is the big one. Meta owns the rules, the rate card, and the data source. They could publish a native cost-analytics tab in Business Manager and vaporize half the product in a release note. They could also delay or soften the October change; as of today Meta’s own public pricing page still shows service conversations as free, and the entire thesis is built on BSP reporting plus Meta’s commitment to publish rates by 1 September. If the change slips, so does the business.
2. Event-driven demand with a short half-life. This is a fire-alarm product. The peak is roughly 1 September to 30 November 2026. If it doesn’t convert into a standing spend-management habit by Q1 2027, revenue decays. Everything about the product design — weekly email, before/after proof, benchmark report — is aimed at this risk, and it may not be enough.
3. BSPs close the gap out of self-defence. The flat-fee BSPs (360dialog-style, no per-message markup) have no conflict of interest and could ship this as a free feature to win competitive deals. That’s the most likely competitive death, and it’s more likely than a startup clone.
4. Weak defensibility (3/10). Nothing here is hard. The bet is entirely on being first, being neutral, and accumulating benchmark data nobody else holds across BSPs.
5. Ethics of the agent leaderboard. Handled in design, but a support org that misuses it into a “fewest words” contest will churn angrily and say so publicly. Ship the guardrails in v1, not v2.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical solo founder or pair, comfortable with messaging-platform
                        integrations and capable of shipping content fast; ideally has run
                        or sold to a WhatsApp-heavy support team.
Time to revenue:        5–7 weeks (calculator ~5 Sept, paid v1 mid-October)
Capital to launch:      <$10K / ₹8L
Top 3 assumptions to validate first:
  1. Median billable replies per conversation ≥4 — measure it against 5 friendly
     brands' 60-day history before building the UI.
  2. High-rate-country support leads will pre-pay in September — 40 cold outreaches,
     count pre-payments, not compliments.
  3. Outbound agent messages are reliably exposed via Cloud API and the top 3 BSP
     webhooks — technical spike, one week, before anything else.
Kill criteria:
  - Abandon if Meta's 1 September rate card delays, caps, or grandfathers the service-
    message charge such that the median target customer's new monthly cost is under $500.
  - Abandon if median billable replies per resolved conversation across 5 sampled brands
    is under 3 — the saving is then too small to fund a subscription.
  - Abandon if fewer than 3 of 40 September cold outreaches will pre-pay $249.
  - Abandon if 120-day retention of the October cohort is under 50% — that proves it's
    a one-off panic purchase, not a product.

15. Next step — 1-week validation sprint

The clock is the constraint. Meta publishes rates on 1 September; this sprint runs the week before.

• Day 1–2 — Prove the data exists. Technical spike only. Connect a test WABA via Cloud API direct and via two BSPs (one per-message-markup, one flat-fee). Confirm outbound agent messages arrive on the webhook with timestamps, message IDs and category metadata sufficient to price them. Binary outcome: can I meter, or can’t I.
• Day 3–4 — Prove the number is big. Get read-only 60-day message history from 5 brands (offer a free manual forecast in exchange — this is an easy trade in late August). Compute median billable replies per resolved conversation, per brand and per intent. This single distribution decides whether the product has a market.
• Day 5 — Prove someone pays. Send 40 cold emails to German, UK and UAE support leads with their manually-computed October delta in the subject line. Offer: $249/mo, first invoice 1 October, product delivered mid-October, full refund if the forecast is off by more than 20%.

Go/no-go, falsifiable: proceed only if (a) outbound messages are meterable on Cloud API plus at least 2 of 3 BSPs, (b) median billable replies per conversation across the 5 sampled brands is ≥4, and (c) at least 3 of 40 cold prospects commit to pre-paying $249. Miss any one and the sprint has told me something true and cheap — which is the point.