WhatToSee Architecture board →
Design prototype · V1 · September 2026

WhatToSee, Colchester prototype

The best possible version of WhatToSee, shown through one town. Proves the model, not the database: one entity, four ways to discover it, and a route that keeps working once you're actually standing in Colchester.

The route, every visit follows it
Discover
"What could we do?" Search, AI search, social, editorial, listicles.
→
Decide
"What should we do?" Browse, filter by audience, time, price, weather.
→
Go
"Where is it, how do I get there?" Map, directions, walking time.
→
Discover again
"We're here now, what else is nearby?" Loops straight back to Discover.
↻ one connected product, not two separate websites

Two routes in, one database underneath

Route A

Build the audience

Found before you leave the house: things to do in Colchester, what's on this weekend, best soft plays, free things to do. Editorial energy, big photography, strong headlines. Premium and trustworthy, not BuzzFeed-branded, just BuzzFeed-clickable.

Route B

The digital town kiosk

Found once you're already in Colchester: QR codes on car parks, the station, shopfronts. Search "menswear," get distance and directions in seconds. Mobile-first, fast, no editorial clutter until the task is solved.

Search / AI / Social → Editorial (Best / New / What's On) → Discover a place → Map / plan
→ Visit Colchester → Scan a QR → Search nearby → Directions → More discovery → Return to WhatToSee

What's built

Real, working, not mockups: a national front door, the Colchester site, a real Mapbox map with a real drawn walking route, and a dedicated council page. Click "The site" or "The map" above to use them directly.

00
National homepage
What WhatToSee is, the town picker (Colchester live, others coming soon), how-it-works, a link through to the council pitch
Built View →
01
Colchester site
Search-led hero, discovery tiles, a real map preview near the top, What's On, New, Best of Colchester, audience row, Explore Everything. Real Colchester photography on the most visible slots
Built View →
02
The flagship map
Real Mapbox GL, custom-styled, 53 seeded places, category pins, search → results → place → a real drawn walking route → While You're Here. Content throughout the site links into it as filters (?cat=, ?tag=)
Built View →
03
Council pitch, /for-towns/
"Give your town a digital kiosk", the flow diagram, a clearly-labelled illustrative data demo, the accuracy caveat kept explicit
Built View →
04
Category & audience pages
Standalone URLs, currently just map filters
Next
05
Listicle template & place page
Real permanent URLs of their own, the content already exists as map filters and the place sheet
Next

Art direction

Nothing in the fleet's Design Bank fits a consumer local-discovery product, it's all B2B finance. This needed its own direction rather than a forced reuse.

Type: Space Grotesk on Inter. Deliberately not the fleet's usual serif-on-sans, this is consumer media, not advisory trust.

Colour: warm paper white, near-black ink, coral as the one loud accent, gold for freshness, teal for utility. A wider hue set used only functionally for map-pin category coding, never as brand colour.

Avoided on purpose: gradients, excessive pills, icon-above-heading card grids, generic SaaS rhythm, tourism-board or council-website register.

Site tree · Colchester · one database, seven sections

Every page this pilot maps to

Not all of these are standalone URLs yet, most resolve to a filtered view of the map or the directory today. Shown as the target information architecture, not current build status.

See6 topics
AttractionsMuseumsHistory & heritage Animals & wildlifeInteresting placesGardens
Do5 topics
Activities & entertainmentOutdoors & nature Family activitiesSport & leisureParks & playgrounds
Eat & Drink6 topics
RestaurantsCafésPubs BarsTakeawaysAfternoon tea
Shop9 topics
MenswearWomenswearChildrenswear ShoesGiftsToys BeautyHomeIndependent shops
What's On7 topics
TodayTomorrowThis weekend Next weekendSchool holidaysSeasonalEvents
New6 topics
Just openedOpening soonJust announced New attractionsNew restaurantsNew shops
Find (utility)5 topics
ParkingToiletsPharmacies TransportAccessibility
Cross-cut: who's goingnot a nav item, a filter
FamiliesToddlersKids TeenagersCouplesAdults GrandparentsGroups

44 topics, 7 sections, 1 database. Every place tagged into as many of these as genuinely apply, the same record surfaces on the category page, the relevant audience filter, and the map, without maintaining separate lists.

Condensed

The business & data model

The short version. Full detail (competitors, the 6 paths to data, proof gates, source docs) stays on the 49k board, this is the quick reference.

One database, three surfaces

One structured place record (category, audience, price, facilities, indoor/outdoor, opening status, events, last checked) powers the website, the kiosk/map, and the council pitch. Never separate data per surface.

The real buyer: councils, not consumers

The map/kiosk is the actual product, sold to councils and BIDs as town-wide wayfinding infrastructure. The consumer website is real and valuable (SEO/AI-search discovery, editorial, ad/affiliate revenue) but secondary, "a nice win," not the main pitch.

The 4-layer business model
Layer 1 · today

Rank4AI services

£1-5M/yr at modest scale, works today

Layer 2 · yr 1-2

Map/kiosk SaaS

£10k-£100k per venue or town per year

Layer 3 · yr 2-3

In-venue revenue

Sponsored pins, brand deals, DOOH ad inventory

Layer 4 · yr 3+

AI data licensing

£500k-£5M per platform per year, speculative

Layers 1-3 alone are a £5-15M/yr business even if Layer 4 never lands. Layer 4 is the call option, not the foundation.

Tracking, kept honest

What's real: search → results shown → place viewed → direction requested. That's genuine local intent. A direction request is not proof of a physical visit, never claimed as one.

Flagged 11 Sep: Mapbox renders the map, it does not give us footfall. Real footfall (did someone actually walk there) needs a deliberate future decision, consent-gated geolocation, QR re-scan inference, or a different vendor. Not settled by "we picked Mapbox."
Who else is here

Mapbox Movement: 700M+ MAUs, partnered with OpenAI, hasn't packaged for AI grounding yet, engage as a partner not a race. Living Map (UK, Bath): the company that inspired this, NHS/Canary Wharf/St Pancras, don't compete, partner first.

Full business case, competitors, data paths & source docs on 49k →