h

HAAT · Competitive Radar

Market Intelligence
LIVE Wolt + 10bis · pulled today
BDBD & Expansion

Cross-platform intelligence — Tel Aviv (Rothschild)

Real restaurants resolved across Wolt and 10bis — the same venue, matched by name + geography, with live ratings, review volume and dish prices. The HAAT overlay that turns this into an acquisition list activates once HAAT connects its catalog.

Wolt ↔ 10bis — resolved restaurants

Click a row for cross-platform detail. Every row is a real venue on both platforms; HAAT column pending catalog.
RestaurantPlatforms Wolt ★10bis ★Reviews (W/10b) Match conf.Status
Wolt (live) 10bis (live) HAAT (pending first-party) ★ Wolt on 10-pt scale · 10bis on 5-pt scale (native)

Dish-level price deltas — real menus, both platforms

Matched dishes pulled live from Wolt + 10bis menus. HAAT column is a modeled floor (no live HAAT data yet).
RestaurantDish Wolt10bisWolt vs 10bisHAAT*

Real finding: for these restaurants Wolt and 10bis prices are near-identical — restaurants set one delivery menu, not per-platform pricing. So the HAAT wedge is not an existing Wolt/10bis gap; it's HAAT's lower commission (~23% vs ~27%) letting a partner price below both and still net more. HAAT* = modeled 3% floor for illustration, not observed data.

◍ How the match works LIVE

  • Wolt restaurant list + menus from its public consumer API (name in EN, geo, rating, review volume, dish prices).
  • 10bis restaurant list + menus from its NextApi (name in HE, rating, reviews, dish prices) after resolving a real Tel Aviv address context.
  • Entity resolution: Hebrew→Latin transliteration + English-substring extraction, fuzzy name similarity, and geo-distance agreement (same origin point) → a calibrated confidence score.
  • HAAT is not scraped — it's mobile-app-only. In production HAAT is the first-party spine; scraped-HAAT vs true-HAAT then gives a real precision benchmark.

⚠ Honest limits of this pilot

  • 10bis sample = top ~70 by review volume within 1.4km of Rothschild — not the full ~750, so "absent from 10bis" is not yet reliable.
  • Match confidence is real but imperfect: rows < 65 include genuine false positives (e.g. a hummus place mis-matched to a pizza chain by geography) — that's the human-review queue, working as intended.
  • No HAAT data at all — every HAAT figure is pending / modeled and labeled as such.
  • Prices are live for the flagship restaurants on the Price Deltas tab; other rows carry ratings/reviews only.

Match-confidence distribution (this pilot)

<60
60–69
70–79
80–89
90–100

Operating threshold set at 65: above it auto-accept, below it route to human review. In production this curve is tuned against HAAT's own catalog as ground truth.

Provenance: Wolt public consumer API + 10bis NextApi, both pulled live for the Rothschild/Center area of Tel Aviv on the build date. Ratings, review counts and the flagship dish prices are real. Cross-platform matches are computed by the transliteration + geo engine described under Method. HAAT has no live data source here (mobile-only app) — its column is pending first-party integration. This is a discovery prototype, not a production dataset.