Don’t reduce the locality to product cards.
FromCatalogue→ToCommercial environment
A physical market is itself a discovery interface: you walk a street, see who does what, and step inside. A catalogue flattens that. An agent could instead assemble a “digital street” of relevant providers around each intent, used where the mission benefits from it and skipped where it doesn’t.
What the catalogue removed
Offline, you work out what you need and go to the right market. You walk past shops whose owners compete for your attention and tell you what they sell and why. You step into one, inspect things, ask questions, negotiate if appropriate and buy. Along the way you stumble on things you didn’t know existed. The environment itself carries information about identity, specialisation, reputation and craft. E-commerce compressed all of that into search, catalogue, product page and cart, and most of that information went with it.
Market, street, shop
The original concept maps the physical sequence onto three levels of digital navigation[CN].
| Level | Question | Interface |
|---|---|---|
| Agentic navigation | What am I looking for? | A conversation with the agent |
| Digital street | Which provider do I prefer? | A short vertical stream of relevant providers |
| Digital store | Let me look inside. | Walkthroughs, past work, products, services, credentials, reviews, availability, ways to transact |
The street is not every business within 2 km. It is the relevant environment generated around one person’s intent. “I want to renovate my kitchen” might bring up kitchen designers, modular-kitchen specialists, carpenters, stone suppliers, hardware specialists, appliance dealers, electricians and installers. “I want a traditional Indian kitchen under ₹2 lakh” brings up a different street of local carpenters, traditional hardware suppliers, stone specialists, regional manufacturers and independent designers. The physical geography stays the same, but the commercial environment changes with intent.
A mode, not the default
The evidence review is precise about the order of operations: agentic filtering first, then a bounded set of relevant providers, then narrative evidence, and finally a structured storefront or service brief before any transaction.
The agent discovers. Structured data filters. Trust validates. Video communicates. The storefront transacts.
Video is the human-facing discovery layer, never the underlying discovery mechanism. That keeps the idea from becoming another social-commerce product. The evidence supports video for explaining specialisation, showing past work and introducing the people behind a service. It does not support making a reel feed the default for every mission. Someone who needs a replacement fuse, an urgent plumber or a specific medicine is better served by direct, structured matching.
So the street is a mode. Design-led, trust-heavy and provider-dependent missions open onto the street (kitchens, furniture, tailoring, interiors, events). Urgent or specific missions get a structured shortlist, with video available but optional.
Go deeper
Discovery beyond known demand
The agent reduces complexity, but it shouldn’t collapse the world into one supposedly optimal answer. Local commerce offers something conventional search tends to destroy: serendipity. You might find a craftsman you didn’t know existed, a specialist whose work you prefer, or an approach that fits the mission better than the one you had in mind. Search starts from I know what I want. The street allows I roughly know what I want to accomplish, so show me what exists.
Open commerce should not digitise the neighbourhood as a catalogue. It should digitise it as a navigable commercial environment.
Not new by itself
Video-first local commerce already exists in pieces. Closebuy is verified[1]. The early research also surfaced Fyndra, Sabu Live, BazoraX, BigBox, Nearcom, BharatSetu and VDOlocal as possible precedents. The evidence review could not verify most of them through durable public sources, so they are treated as leads, not evidence[CN]. The claim here is about composition: an environment generated around intent, filtered by capability and trust, and running on open rails.
What could go wrong
- Friction.A browsing layer adds steps, and for many missions a plain shortlist may simply be faster.
- Preference for lists.Even for design-led missions, users may prefer a comparison table to a stream of providers.
- Attention drift.An environment built for browsing can turn into an infinite scroll optimised for engagement rather than completed missions.
- Weak precedent.None of the adjacent products has yet shown public evidence of traction.
How we’d test it
No lift from the street over a plain directory, or a lift that comes with worse outcomes.
| The test | What we would measure | What would end the idea |
|---|---|---|
| Street versus directory (Experiment 3) | Same providers, randomised formats: enquiries, quotes, conversion, decision time, confidence, outcome quality | No lift, or a lift that comes with worse outcomes |
| Mode fit | Street usage and skip rate by mission type, design-led versus urgent | Users skip the street even for design-led missions |
How strong is the evidence?
- Established
- Adjacent local-discovery products exist, at least one of them verified.
- Evidence-supported
- Video works best as a consideration layer after intent filtering.
- Hypothesis
- A curated street outperforms a directory for design-led, capability-heavy missions.
- Unknown
- Which mission types benefit, and whether browsing drifts into engagement for its own sake.
Sources
- [1]Closebuy: AI Local Shopping, Google PlayChecked; the listing showed 50+ downloads
- [CN]Voice-session working notes (internal)Provided
Spotted a source we missed or a rule we misread? Tell us what we got wrong.
A street shows who is there. It cannot tell the customer whom to believe.