Beli growth plan
Neel Naik · for Beli's Growth Lead, Product role · November 2026
Who I am
- A product manager at Tesla doing 0-to-1 work, including the Semi service and ownership experience. I own a scheduling platform that grew from 120 to 508 service centers.
- An ML engineering and data science background (MS Operations Research, Georgia Tech). I design experiments, read the data myself and build prototypes.
- A serious foodie and a heavy Beli user. I know the product as a customer, not only as a PM.
Why I'm the user this role is about
- I spent over $20,000 eating out last year, mostly on new restaurants and cafes.
- Beli ranks me a top 4% diner in San Francisco, and I still find the product hard to use. I trust its content, but logging is work and the best features sit behind invites.
- At Wingnuts, the wing spot outside Georgia Tech, my college goal was to eat my weight in chicken wings, about 210 pounds, by graduation. I hit it in my junior year. I kept a spreadsheet ranking all 30 flavors, lost it, and have wanted a better version since.
- I send places through Google Maps because I share one-on-one or in small groups, and a Maps link is easier to open.
- I am the designated food person: friends text me "where should I go?" because asking is easier than downloading Beli.
The diagnosis
Beli's highest-intent users hit the most friction. Features are gated behind invites, logging a meal is manual, and the best content is found on TikTok, not in the app. Churn concentrates in the segment Beli most wants to keep.
Facts from my own Beli profile
My Beli profile as of November 10, 2026.
| Restaurants ranked, last 30 days | Cuisines | Rank among San Francisco diners |
|---|---|---|
| 11 | 17 | Top 12% |
| 23 | 24 | Top 6% |
| 35 | 34 | Top 4% |
Inference: a power law. Most users log very little, and a very long tail logs a lot.
Scorecard
| Category | Pain point | Metric I own | Guardrail |
|---|---|---|---|
| Acquisition | Curious visitors from TikTok or a shared link hit a wall | Visitor-to-install; install-to-signup; both split by source | Cost per activated user |
| Activation | A new user with no network has nothing to rank | Time to first recorded session; first ranking in 7 days; first friend in 7 days | Rating quality |
| Engagement | Logging is manual, so sessions are rare and thin | DAU and MAU; time to record a session; photos and dishes per rating | Prompt opt-out rate |
| Referral | Friends have no reason to join; invite gates tax the foodie | Invite-to-activated-user conversion; retention of gated vs ungated users | Spam and unsubscribe rate |
| Revenue and data | No ads, small revenue, unused ranking data | Supper Club conversion; revenue per seated booking; data product revenue | Retention and trust |
Ideas I would test
- Invite reframe. Free by default, with friends unlocking shared lists, taste match and group picks. Measure gated vs ungated retention first.
- Open the wall for visitors. Show full ratings to TikTok and shared-link visitors. Gate save and follow, not the content.
- Dwell-time prompt. "Did you go here?" after 5 minutes at a cafe or 20 to 30 at a restaurant. A/B the timing.
- History import. Calendar import exists. Extend it to email receipts and to Resy and Tock, so a new profile fills in one pass.
- Menu scan. Scan the menu in the same session as the photo upload, so people rank dishes, not just places. Extracted text also feeds Beli's internal dish database.
- Comparable rankings. Rank within cuisine and type, across cities. Add "not a relevant ranking" instead of forcing a score.
- Taste-matched reservations. Book a table with people who rank like you, in a social format such as two pairs sharing a table of four, so nobody dines with a lone stranger. Restaurants pay for demand, never for placement.
- Ranking collabs. Top restaurants and top users in a city earn a Beli-exclusive menu item.
- Premium for heavy loggers. Dish-level logging, a year-in-review and data export, with the free app untouched.
What growth earns
Revenue today is small and optional: a $74.99 a year Supper Club, streak restores and creator guides. The founders have ruled out in-feed ads to keep trust, so growth has to carry the business. Three levers fit that constraint:
- Social reservations. Match users with similar rankings and book a table for the group. Restaurants pay for demand, such as a fee per seated booking, never for placement in the rankings. The founders have said they are exploring reservation partnerships.
- Ranking-driven collabs. Rank both restaurants and users. The top restaurant in a city earns a Beli-exclusive menu item, and top users get first access.
- Aggregated data. Dish-level trends for restaurants: what is rising, what people love or dislike by dish. No individual user data and no paid placement. A student case study argues for the same direction.
What I would measure, and how
- Funnel by source. Visitor, install, signup, first ranking, first friend. Tag links, keep the source through the install, split by source and city.
- Time to record a session. Taps and seconds to a saved rating, plus photos and dishes per rating.
- DAU and MAU. The ratio shows habit. Split by city and cohort.
- Gated vs ungated retention. Users who hit an invite gate vs matched users who did not, at 30 and 90 days.
- Invite quality. Invite-to-activated-user conversion and viral coefficient, not raw invites.
- Rating quality. Rating length, rank spread and later edits.
- Revenue and data. Supper Club conversion, revenue per seated booking, dish-level coverage per city.