Dispersing a city with small good deeds.
The strategy behind Questable's city-wide load-balancing pilot with Amsterdam Municipality. How a network of rewarded deeds steers residents and visitors toward neighborhoods that want more footfall — and away from the ones drowning in it.
Popular cities are being loved to death.
Amsterdam receives 27M annual day visitors across 850K residents. They concentrate in a handful of central attractions, straining infrastructure and resident quality of life.
Dam Square, Rood Light District, Nine Streets.
Residents are being priced out of their own neighborhoods. Local shops turn into waffle chains. Housing becomes short-term rentals. Anti-tourism sentiment is growing — and with it, political pressure to cap or redirect.
Noord, Oost, Nieuw-West.
Neighborhoods outside the centre have capacity, culture, and businesses that would thrive with more footfall. But visitors never reach them — they don't show up in the guidebook.
32 previous experiments, no actionable insight.
Amsterdam's municipal dispersal programs failed not from lack of effort but from lack of attribution — rising tourist numbers masked center-reduction effects, proxy data (Flickr posts, City Card taps) couldn't separate correlation from causation.
You can't improve what you can't measure.
Previous data tools identified WHERE problems exist, not WHETHER solutions work. Annual reports close the feedback loop months after the intervention ends — by which point the next summer's crowds are already booked.
Small rewarded deeds can redirect foot traffic at the neighborhood level.
Not with signs. With incentives that show up where you already look — on the map in your phone.
residents and visitors can earn small, real rewards for doing simple deeds at specific places and times,
we can measurably shift footfall from hot zones to cold zones, peak hours to off-peak, without coercion.
The lever is the reward, not the rule.
Reward amount and availability are both variables. Turn both up in a cold zone, both down in a hot zone, and watch the flow change.
The map is already where people decide.
We're intercepting visitors at the moment they're asking 'where next?', not trying to change their mind at a kiosk.
The right time beats the right place.
A coffee-at-11am deed redirects just as much as a café-in-Noord deed. Time-based shifts are easier to ask for than neighborhood-based ones.
How the toggle works.
Four moving parts. Reward intensity is the dial the city actually pulls.
Zone the city.
Partition the city into zones based on congestion goals. Hot (disperse from), warm (hold steady), cold (pull toward). Draw by hand on the operator map.
Seed partners in cold zones.
Walk the streets, recruit local businesses as network partners. Each sets a simple deed (visit before 11am, bring a reusable cup) and a small reward (free coffee, 10% off).
Toggle by zone in real time.
City operator opens the dashboard. Amplify rewards in cold zones, throttle them in hot ones. Change on the hour as crowd data comes in.
Measure the shift.
Compare footfall pre/post toggle, control vs. treatment zones, peak vs. off-peak. Dashboard shows redemption rate, crowd-index delta, partner NPS.
The program we're running with Amsterdam.
€10K seed from Gemeente Amsterdam. The grant buys causal proof on one arm; the rest of the network scales on partner-funded rewards and the intrinsic reward of doing a good deed. By design, no permanent subsidy.
€10K · Gemeente Amsterdam
Disbursed. Funds one arm with fully-paid rewards against a control — enough for directional causal proof at €5–7 per completion.
Their marketing budgets
Café comps a coffee, shop gives 10% off — tiny per-reward cost, big footfall value. This is where the bulk of completions come from.
The deed is its own reward
Badge/status arms test whether people will disperse for pride alone. Zero variable cost. If this works for a segment, it scales forever.
Months 1–3 · ~1,500 completions
Spend the €10K on one focused arm with fully-paid rewards. Control group + treatment in adjacent zones. Directional causal signal on incentive effectiveness.
Months 4–8 · ~15K completions
Partner-funded rewards carry the majority of arms. Voucher, cultural, social, transit — near-zero marginal cost. Run concurrently under the statistical framework.
Months 9–12 · toward 28K
Optimize winning combinations. Pure-intrinsic arm (badge / status) tests self-sustaining motivation. Network runs on partner fees + city data value.
The six incentive types we'll test.
Not all rewards are equal. The research is built to find which ones move which segments — not to assume.
Control (no incentive)
Baseline completion for the same quest with no reward. The only way to attribute lift.
Discount voucher (€5)
Instant local-business value converts price-sensitive segments.
Cultural reward
Museum or concert passes attract repeat-visit NL explorers.
Gamification
Badges and leaderboards move Gen Z more than cash.
Social reward
Instagram-worthy experiences convert the 18–41 multi-day visitor.
Transit pass (€9 value)
Lowering the friction to leave the centre is its own reward.
Each incentive is crossed with 3 difficulty levels (Easy / Medium / Hard) and 3 reward amounts (€3 / €7 / €15), stratified by 4 user segments and 8 interest categories.
How the math holds up.
28,000 users isn't arbitrary. It's what the power analysis demands to detect the effects we care about.
Statistical design
- Baseline completion rate assumed at 20%
- Statistical power: 80% · α = 0.05 · two-tailed
- Mixed-effects logistic regression with partial pooling
- 192 cells: 6 incentives × 4 segments × 8 interests
Attribution
- Control group per arm — the only way to attribute causally
- GPS check-in verifies visits (not proxy signals)
- In-app completion = ground truth for primary DV
- Each visitor randomly assigned — no self-selection
Sample sizing
- Full factorial (5pp detection): 210K users — infeasible
- 10pp detection: 57,600 users — beyond our reach this year
- 15pp detection: ~28,000 users — our chosen operating point
- Hierarchical model borrows strength across cells
Effect size targets
- Main effects (Incentive): ≥5pp detectable · 1,096/cell
- 2-way (Incentive × Segment): ≥10pp · ~300/cell
- 3-way (Incentive × Segment × Interest): ≥15pp · ~150/cell
- Interest-matched vs. generic: H2 target ≥10pp (p < 0.01)
The strategic moat: attribution.
We're in the early stages of partner onboarding and haven't run the arms yet. Here's why we believe the program will produce answers — not just effort.
32 experiments. Zero causal answers.
Previous dispersal efforts couldn't separate correlation from causation — proxy data like Flickr posts and City Card taps can show WHERE people go, but not whether an intervention made them go there. Rising tourist numbers masked any center-reduction effect.
Direct control + measurement.
GPS verification for each visit. Randomized control groups per arm. A/B testing as the operating model, not an afterthought.
Self-sustaining by design.
The €10K municipal seed buys causal proof on one arm — not the ongoing program. Everything else rides on partner marketing budgets (they pay because they get the traffic) and intrinsic motivation (the deed itself is the reward).
0.20% of visitors is all we need.
Of 14M unique annual visitors, converting 0.20% produces 28K research participants — enough statistical power to hand the city a validated recommendation. On €10K of seed funding, that's a survey that pays for itself in behavior.
Previous efforts vs. Questable.
Who should talk to us.
If your city has a congestion problem and is tired of signs that don't work.
Tourism boards
You want to keep your brand without killing the districts that built it. We give you a dial instead of a campaign.
City planners
Dispersal policies need a feedback loop. Our pilot gives you real-time footfall response to reward intensity.
District BIDs
You want footfall to your neighborhood, specifically. Ride the network; don't build your own.