# Town Monetization Memo **Pricing, activation, and the credit model — written from the outside** Darsh Shah · June 2026 · Sources: town.com/pricing (read 2026-06-12), Fortune (June 3), a16z investment announcement, and my own account. --- Your Business Operations posting lists "owning the operational infrastructure for monetization — billing, operations, and pricing" as the first real project. This memo is that work, started before the interview. It uses only public data and a week inside the product, so every claim here is falsifiable with internal data in about a week — Section 5 is the plan for that. I'd rather be wrong specifically than impressive vaguely. ## 1. The one number that matters Fortune reported it alongside your Series A: **99% two-month retention among users who've built at least one custom automation.** That isn't a retention stat. It's an **activation definition.** It means Town's entire growth model compresses into one funnel — signup → context connected → **first automation** → retained, approximately forever. Every monetization decision should be subordinate to one question: *does this move more users through "first automation," or fewer?* Held against the pricing page, three things stand out. ## 2. Finding one: the activation event is paywalled Per the public pricing page, after the 14-day Pro trial expires, a non-converting user lands on Free: 30 chats a month and **no custom routines at all**. Starter ($15) allows two routines, daily triggers only, hard-capped. The full automation surface — the thing that produces the 99% — opens at Pro, $49/month billed annually. So the only free path to the activation event is the 14-day trial window, and that window is **calendar-gated, not progress-gated**. A user whose aha moment would arrive on day 16 falls onto a tier where the aha is structurally impossible. The funnel's most valuable event sits behind the moment of maximum skepticism: pay first, discover the magic after. The 14-day Pro trial is the right instinct — it's a reverse trial, and reverse trials are the correct pattern for products whose value needs demonstration. But its clock is wrong. The trial's one job is to manufacture activated users, and it expires on a schedule that ignores whether it did. ## 3. Finding two: metered usage meters the moat Your own FAQ: *"Credits are the usage unit consumed when the assistant does work for you — chats, workflow runs, and more."* Your lead investor's stated thesis: the durable advantage is **accumulated personal context** — an edge that compounds with use. Put those sentences next to each other. If context compounds with use, and every unit of use is metered, then the pricing model charges a toll on the exact behavior that builds the company's defensibility. Usage-based pricing aligns price with cost; for Town it *misaligns* price with moat. The comparison set makes the psychology worse. The prices a prosumer anchors on are flat: ChatGPT and Claude at $20, Superhuman at $30, Fyxer at $30–50. Flat means *never think about the meter*. Town asks users to think about the meter on every chat — the taximeter effect, applied to a product whose pitch is "stop managing your tools." Lindy and Motion run credit models too, so this is a live industry-wide question; the prize is that **whoever makes proactive-agent pricing feel flat wins the prosumer segment**, the same way unlimited data plans ended the per-minute era. ## 4. Finding three: the $0 cap turns success into silence Overage exists from Pro up at $0.032–0.044 per credit — roughly **4.5–6.4× the bundled per-credit rate** of each plan. That's penalty pricing, which is a defensible way to push upgrades instead of pay-as-you-go. But the overage cap **defaults to $0**, so out of the box, a user who exhausts their credits doesn't get a bigger bill or an upgrade — their assistant **stops working mid-month**. For a proactive product, that's the worst possible failure mode: the routines go quiet, the habit loop breaks, and the user has to notice the silence to fix it. Your heaviest, most successfully-activated users are the ones most likely to hit it. Bill-shock protection is right; *silence* as the default behavior at exhaustion is a churn event wearing a safety feature's clothes. ## 5. What I'd change **(a) Gate the trial on activation, not the calendar — do this first.** Trial ends 14 days *after first automation is built* (with a reasonable outer bound, say 30–45 days). Users racing toward activation get the runway they need; users who activate on day 2 still get two weeks of post-activation habit formation before the paywall — which now arrives at the moment of demonstrated value, when willingness to pay peaks. Cheapest experiment on this list, and it directly attacks the one number that matters. I ran exactly this pattern as a pricing project at a B2B SaaS company (reverse trial, subscription conversion) and it's the highest-leverage change available without touching the credit model at all. **(b) Meter execution, not exploration.** Make chats free or near-free (a generous flat allotment); meter **workflow runs and lookups**. The COGS logic supports it — conversation is the cheapest action class, while runs carry real tool-execution cost — and the value logic is cleaner: *work done while you weren't there* is the thing worth charging for, and it's also the thing users will happily pay for. Crucially, this un-taxes context accumulation (the moat) and lets the Free tier include one routine with a small monthly run allotment, so the funnel can produce activated users again instead of holding them in a 30-chat waiting room. **(c) Replace silence with a ramp.** Default the overage cap to something small but non-zero (~$15) with threshold alerts at 50/80/100%, and make credit exhaustion trigger a one-tap tier upgrade offer instead of a quiet stop. Overage becomes a bridge between tiers rather than a penalty or a dead end. Recommendation: ship (a) now, evolve toward (b), and do (c) regardless of the other two. ### Illustrative economics (assumptions flagged, to be replaced with real data) Take ~10,000 users and assume 1,500 new signups/month, a 25% trial-to-activation rate today, and blended $60/month ARPU on converts. If activation-gated trials lift activation by 5 points — plausible when the binding constraint is the calendar, not the product — that's ~75 additional activated users/month who retain at 99% over two months instead of bouncing. At even modest 12-month survival that's roughly **$50K+ of new ARR per monthly cohort**, against a COGS cost of extending free inference for slower activators that is bounded by the cheapest action class. Every number in this paragraph is a placeholder; the structure is the point. Week one inside, the placeholders die. ## 6. The first 30 days inside 1. **Metering truth.** Pull credit consumption distributions by action class and tier. Confirm what actually draws credits (the FAQ's "and more" is doing a lot of work) and what a routine costs to build vs. run. 2. **Funnel truth.** Time-to-first-automation distribution vs. the 14-day window. What fraction of eventual activators would the calendar have cut off? 3. **Dormancy audit.** How many paid users hit exhaustion with a $0 cap, and what does their churn curve look like vs. everyone else's? 4. **Run the trial experiment.** Activation-gated vs. calendar-gated, holdout, two cohorts, four weeks. 5. **Segment definitions.** The JD asks for segment analysis; the segmentation that matters here is usage mix (the 70/30 work-personal entanglement), automation density, and assistant-to-assistant network density — these predict both retention and willingness to pay. 6. **Billing ops hygiene.** Metering pipeline integrity, spend-cap UX, dunning, sales-tax/nexus as revenue scales — the unglamorous infrastructure the JD names, owned end to end. --- ## Appendix A — My week inside Town *[Populated from 03_usage_appendix_capture.md — onboarding timeline, the credit moments, path to first automation]* ## Appendix B — The automation I built, in Town *[Populated from capture file section D — what it does, what it cost in credits, screenshot]* --- ### Caveats — what this memo can't see from outside Whether *creating* a routine consumes credits; whether proactive/background actions meter; Motion-vs-Town credit-unit comparability; your actual COGS per action class; conversion and churn baselines. All of it is week-one knowable inside. ### Who's writing this Darsh Shah — monetization and strategic-finance operator, San Francisco. Launched the live-audio monetization product at ShareChat (Google-backed, 180M users): +50% creator retention, $14.4M revenue impact. GTM for its ads product to $60M ARR. Strategic finance in the CEO's office through a $50M raise. Chartered Accountancy exams cleared first attempt (top 1%); Kellogg MBA. This memo was researched and drafted in an AI-native workflow I built — which felt like the only honest way to apply to Town.