BetaCurrently London onlyFairWhere is in beta. Plans currently work for London only.
Transparent by design · Beta · London first
How FairWhere decides
FairWhere ranks places by how well they work for the whole group, not just by distance, popularity or sponsorship. Under the hood, FairPath turns London's network into a travel-time model that improves with use. The product is in beta; planners currently work for London only.
Proprietary routing intelligence
FairPath: travel times that improve with live checks
Pinning a midpoint on a map is trivial. Ranking a shortlist fairly for a multi-origin group is not. FairPath is FairWhere's estimate engine: a hierarchical, privacy-safe model of London door-to-door times that learns from every live check, so free previews stay sharp and scarce TfL / Deep Search calls are spent where they teach the most.
Mechanism
A travel-time field that gets sharper where live checks teach the most.
Privacy
Training uses ~100 m origin cells and station pairs - never street names.
Hierarchical residual shrinkage
Station-pair, category and time-bucket residuals are learned from live labels, then James–Stein / empirical-Bayes shrunk toward parents when data is sparse, so a noisy hop cannot wreck the ranking.
EMA cache + confidence blending
Each hop keeps an online Bayesian mean (EMA) with effective sample size n_eff. Cached seconds are blended with geometric heuristics by hop confidence. 1 − e^(−n_eff/n₀), so unproven corridors stay cautious.
Graph hop inference
Missing OD pairs are filled from the rail graph: inferred hops, access/egress factors for parks and museums, and reliability quantiles (p80 bands) so “kindest” mode can score worst-case journeys, not just means.
Acquisition under a live budget
Uncertainty scores decide which corridors to explore next. Demand-weighted origin–destination sampling from privacy cells seeds coverage without burning user fair-use quotas.
Ranking-regret priors
When a force-check flips the shortlist, FairPath rebuilds venue and category residual nudges: multi-fidelity labels that pull disputed places back into contention on the next free preview.
Privacy-safe learning
Training uses ~100 m origin cells, station pairs, venue categories and timings. Never street addresses or names. Live TfL and Google remain authoritative when you verify; FairPath makes the free rank trustworthy earlier.
Spreadsheets, “meet in the middle”, and a handful of Google Directions tabs cannot maintain a calibrated multi-origin field, shrink sparse residuals, or optimise a limited live-check budget for group fairness. That is the point of FairPath.
How we measure estimate error (organic MAPE vs live TfL, what we exclude, why Google is not the teacher): FairPath accuracy.
Why Google Maps alone is not enough
Why this hasn't been done before
Fairness for a group is not one Directions search: it is every person's journey to every candidate, then a score across the set. That product of people × places is why nobody has shipped this as a casual consumer tool: brute-force live routing does not fit in a chat window.
Scale
FairPath ranks the field first, then live-checks only the shortlist that matters.
Example: 4 people considering 100 places (a typical “best of” list).
Option 1 · Google assessment basis
Look up every journey by hand
- 4 × 100 = 400 one-way Google Directions lookups (one per person per place).
- At roughly 45 seconds each to open, paste, wait and note the time, before you have compared fairness, returns or trade-offs.
~5 hours
400 × 45s ≈ 18,000 seconds of manual work
Option 2 · FairWhere model
Estimate the field, live-check the shortlist
- FairPath ranks all 100 from the calibrated travel-time field in under a second.
- Live TfL then verifies the fairest ~20 (not all 100) for the group. 4 × 20 journeys, not 4 × 100.
~15–20 seconds
Instant shortlist + ~15s live check → ranked results
That gap: hours of tab-switching versus under half a minute: is why “just use Google” never became a fairness product. FairWhere only works because FairPath collapses the search space first, then spends live routing where it changes the answer.
Fairness is more than the midpoint
The fairest option is not always the geographic middle. FairWhere compares everyone's journey, return route, cost, convenience and venue fit to show which places work best for the group, using FairPath estimates, then optional live verification.
Problem
A map centre can still leave one friend with a much longer journey.
Prefer a side-by-side comparison? Meet in the middle vs FairWhere. Already driving a through-route and collecting friends? Pick up on the way. For a practical playbook, read how London groups actually meet, or start from a curated Placelist and rank it for your group.
The FairWhere Score
Every result receives a score out of 100. A place can be fastest overall but still unfair if one person carries most of the travel burden.
Score
Balance, total time, and the longest trip all pull the ranking - not crow-flies alone.
Journey balance
25%How similar each person's travel time is.
Total group travel time
35%Sum of everyone's journey to the venue.
Longest individual journey
15%Protects the person with the hardest trip.
Return journey fairness
10%Late-night and reverse-route sanity check.
Convenience
10%Changes, walking time and step-free access.
Venue or destination fit
5%Suits the plan, group size and time.
Availability & bookability
gateOpen at the right time; bookable where relevant.
Accessibility constraints
gateRespects any step-free or mobility needs.
Walking time
gateTotal minutes on foot at either end.
Weights shown apply in Best Blend mode. Other fairness modes reweight the score.
Research sibling · Content factory
London Journey Fairness Index
The FairWhere Score ranks venues for your group. The Fairness Index is the city-scale research view: how journey burden falls across London neighbourhood pairs, built from the same precompute matrix that powers crawlable /meet/{a}/{b} pages - not crow-flies midpoints.
- Area origins are primary stations (Clapham Common, Walthamstow Central), never home addresses.
- Primary signal is the minute gap between the two door-to-door estimates to the winning venue, plus FairWhere Score balance.
- Meet-pair drafts stay private until admin Approve; the Index CSV ships once the matrix is stable - not two pages a day.
Fairness modes
Pick the definition of fair that matches your group.
Best Blend
The default. Balances all fairness factors.
Fastest Overall
Minimises total group travel time.
Most Equal
Keeps everyone's journey as similar as possible, even if total travel time is slightly higher.
Kindest to the Worst-Off
Protects the person with the longest or hardest trip. FairPath uses journey reliability (upper-tail times), not just averages.
Fewest Changes
Favours direct routes with the fewest interchanges.
FairWhere shows the trade-offs
FairWhere does not hide trade-offs. It shows why one place ranks above another.
- LabelFastest, but less fair
- LabelFairest overall
- LabelBest for the longest traveller
- LabelLower venue prices, weaker journey fairness
- LabelGreat venue, weaker journey fairness
Partners do not buy the fairest result
FairWhere may show Partner Venues, Verified Placelists or Sponsored Placelists. These are clearly labelled. Partner status does not automatically improve the FairWhere Score.
Partner venues are ranked using the same fairness logic as other eligible places.
Travel times are estimates, then verified
Routes, travel times, prices, opening hours and availability can change. FairWhere shows a FairPath estimate first, then optional live TfL or Deep Search verification. Always check before final booking or travel.
Verify
Scarce TfL calls are spent on the shortlist - and can flip a disputed place.
We track free-preview error as organic MAPE versus live TfL: mean of |estimated − actual| / actual × 100 on product traffic. Seeder runs, soft-arrival taps, and retired Google teacher labels are excluded so the number cannot be optimised by synthetic load.
Full definitions, exclusions, and the public snapshot contract: FairPath accuracy.
Data sources for journey overviews
FairWhere may use TfL Open Data and optional Google Routes to estimate and verify London public-transport journeys. FairPath learns only from privacy-safe station pairs, venue categories and timings - never from street addresses or names. Live TfL is the persisted accuracy teacher; Google Deep Search may still appear as optional live UX but is not written as a durable training label. FairWhere is not endorsed by or affiliated with TfL. Journey overviews are simplified and are not intended for navigation.
Your location data matters
People may enter home, work, airport or travel locations. FairWhere should only use this information to create and share Plans, unless the user agrees otherwise.
Privacy Policy·Terms
Frequently asked questions
- What is FairWhere?
- FairWhere is a London-beta group meeting planner that ranks venues by journey fairness for everyone coming (travel time, balance, longest trip, return journey, changes, cost, and venue fit), not by geographic midpoint, popularity, or sponsorship.
- How is FairWhere different from meeting in the middle?
- A geographic midpoint ignores Tube lines, walking, changes, return journeys and cost. FairWhere Score compares real journey burden across the group, so the top result can be fairer even when it is not in the middle of the map.
- What is the FairWhere Score?
- The FairWhere Score is a 0–100 ranking for each candidate venue. In Best Blend mode it weights journey balance, total travel, longest individual journey, return fairness, convenience and changes, and venue fit, with quality and efficiency checks applied.
- Does FairWhere work outside London?
- The product is in beta and planners currently work for London only. More cities are planned later.
- Do partner venues buy a better ranking?
- No. Partner Venues and Sponsored Placelists are clearly labelled. Partner status does not automatically improve the FairWhere Score; partners are ranked with the same fairness logic as other eligible places.
- Are travel times exact?
- Free previews use FairPath, FairWhere's estimate engine trained on prior live checks. Optional TfL live checks and Google Deep Search verify times. Always confirm before you book or travel.
- What is FairPath?
- FairPath is FairWhere's proprietary journey estimate system. It improves free rankings from every live check, fills in missing station hops, adjusts for venue access quirks, and spends live-check budget where times are uncertain.
- How does FairWhere measure travel-time accuracy?
- Primarily as organic mean absolute percent error (MAPE) of FairPath estimates versus live TfL actuals on product traffic. Seeder runs, soft-arrival taps, and retired Google teacher labels are excluded. See /fairpath-accuracy for definitions.
- What is the London Journey Fairness Index?
- A research view of how journey burden falls across London neighbourhood pairs, built from the same precompute matrix as FairWhere’s meet-between pages. It is not a crow-flies midpoint map. See /fairness-index.