StreetComplete

OpenStreetMap has a dirty secret. The geometry is largely done — in most of the developed world, the roads are drawn, the buildings are traced, the parks have edges. What’s missing is everything about those shapes. Is that footpath paved or gravel? Does that street have lighting? Is the pharmacy’s entrance step-free? When does the bakery actually open? The map is wide and thin, and the thinness is why routing apps still send wheelchair users down stairs.

The reason that gap persists isn’t apathy. It’s onboarding friction. Contributing to OSM traditionally means confronting a free-form key=value tagging schema documented across thousands of wiki pages, inside editors — JOSM, Vespucci, iD — built by cartographers for cartographers. The barrier isn’t drawing. It’s knowing that “is it lit” is lit=yes, and that surface=paving_stones and surface=sett are different things.

StreetComplete, built by German developer Tobias Zwick under the handle westnordost, throws that model out. It doesn’t hand you a map and an editor. It hands you a question. The app cross-references your surroundings against OSM’s database, identifies every attribute that’s missing or stale, and drops each one on the map as a quest marker — a single, concrete, answerable prompt. You tap a button. The app translates your answer into correct tagging and uploads it to OSM under your own account. No second editor, no review queue, no wiki.

It is free, ad-free, tracker-free, open source, and distributed on both Google Play and F-Droid, and it’s an easy recommendation for any newcomer to OSM. StreetComplete is not a simplified editor. It’s a fundamentally different interaction model that happens to produce the same output — and cleaner output, at that.

• The Quest System: Missing map attributes surface as location-pinned questions with constrained answer sets. One question, one tag, one tap. The app owns the schema knowledge so you don’t have to.
• Overlay Modes: Focused editing surfaces for sidewalks, cycleways, addresses, and street parking — a step up in power for users who want to systematically complete one data class across a neighborhood rather than answer whatever the map serves up.
• Resurvey Logic: Quests don’t just target absent data. StreetComplete re-asks about aging attributes, treating the map as something that decays rather than something that gets finished.
• Offline Map Downloads: Cache an area before you walk it. Essential for basement-level cell coverage, rural surveying, and anyone who’d rather not burn mobile data for two hours.
• Photo-Assisted Address Capture: Snap a house number and enter it, rather than squinting at a plate while balancing your phone and a coffee.
• Free-Form Notes: For everything the quest system can’t model — a new business, a closed road, a situation that doesn’t fit the question format — you can drop a note for other mappers to resolve.
• Levels & Achievements: A progression system that unlocks new quest types as you contribute, gating complexity behind demonstrated activity.

The Inversion

The central design decision here is worth sitting with, because it’s smarter than the cheerful cartoon UI lets on. Every other OSM editor presents an open input surface: here is the database, here is a text field, good luck. StreetComplete presents a closed one. Each quest is a multiple-choice question with a fixed, curated answer set that maps deterministically to a known tag.

The payoff is data quality. It’s better data. The dominant failure mode for novice OSM contributors has always been well-intentioned garbage — invented tags, misapplied values, freehand strings where an enumeration belongs. By constraining the answer space, StreetComplete makes the most common category of newbie error structurally impossible. The app has effectively encoded a senior mapper’s judgment into the UI and then loaned it to a stranger for the duration of a walk. That’s a quality-control achievement disguised as a convenience feature.

The Loop

The gameplay loop is embarrassingly effective, in the best sense. You open the app. Your immediate surroundings bloom with markers. You walk twenty meters, tap one, answer “yes, it’s lit,” and watch it vanish with a satisfying pop. Twenty meters later, another. The latency between action and reward is near zero, and the reward is real — you didn’t collect a fictional token, you improved a public dataset used by millions of downstream applications.

The compulsion is the same one that powers Pokémon GO, minus the manipulation. There’s no energy meter, no timer pressuring you back, no dark pattern engineered to extract a session. The only thing pulling you down the next street is that there are markers on it. The design invites the “addictive” label without any of the guilt that usually comes with it — enjoying it this much requires no explanation.

The level and achievement layer is deliberately thin, and correctly so. Progressive unlocking of quest types functions less as a dopamine drip than as a competence gate — you graduate into questions with more ambiguity once you’ve demonstrated you’ll actually keep showing up. It’s the mildest possible gamification, and heavier machinery would have cheapened the whole thing.

The Structural Ceiling

Here’s where the honest criticism lives. StreetComplete can only ask about things that already exist in the database. It is an attribute-filler, not a map-builder. It cannot draw a road that was never traced, add a building nobody digitized, or split a way. In a well-surveyed European city, that limitation is invisible — the geometry is there, and the app has thousands of questions to ask. In under-mapped regions, where OSM’s real gaps are most severe, StreetComplete goes quiet precisely because the underlying shapes it needs to interrogate haven’t been drawn yet. The tool is most useful where it’s least needed. That’s an inherent property of the model, not a bug, but it puts a firm ceiling on what the app can do for global map equity.

The inverse problem is built into the model too. Map a dense, active area thoroughly and you exhaust it. Quest fatigue is the predictable end state for committed contributors who’ve cleaned their neighborhood and now walk streets that render as an empty map. The resurvey mechanic partially answers this — data ages, quests return — but on a cadence measured in months, not walks.

Control and Curation

Not every quest is equally worth your time, and not every mapper agrees on which. StreetComplete does offer quest toggles and presets, letting you disable categories wholesale, but the granularity stops there: no contextual rules, no per-area profiles, no priority weighting. Right now the control is coarse — on or off, globally. For a casual contributor, irrelevant. For someone running a deliberate survey campaign across a district, it’s the difference between a focused hour and a cluttered one. The overlay modes are the current best answer, and they’re good, but they cover four data classes, not the long tail.

The aesthetic is friendly without being childish — flat, high-contrast, generously spaced tap targets sized for a thumb operating one-handed while walking. This is the correct design brief, executed correctly. Multiple map themes are on offer, which matters more than it sounds: a bright daytime theme is unreadable at dusk, and half of all street-lighting surveys happen at dusk. Answer sheets slide up as clean cards with large, obvious options and clear iconography that survives glancing at arm’s length.

As a 2D vector-tile app, StreetComplete is light on hardware by design; the demanding parts are the things no interface can design around. Two costs are inherent to the concept.

Battery drain is significant by nature. Continuous GPS, an always-on screen, and live map rendering is a punishing combination, and an extended mapping walk will visibly eat a phone. Some of this is physics — no software fixes the cost of a GNSS fix — which means serious surveying is a battery-pack activity. Plan accordingly.

Large offline downloads are a strain on the renderer. Cache an aggressive area and the map has that much more to draw at once, so smaller downloads than you’d instinctively grab are the safer bet.

StreetComplete is Android-only, and that fact does more damage than a feature list suggests.

On Android, it’s superb. It’s a native app that leans on the platform properly: background location handled cleanly, real F-Droid distribution for users who want nothing to do with Google Play, and a permission story that’s easy to reason about because there’s no advertising SDK or analytics pipeline to audit. Touch is the right input model here — the entire interface assumes one hand, in motion, outdoors, possibly in the rain. Nobody surveys a street with a mouse.

But the absence of an iOS release is a strategic problem, not a preference problem. This is a network-effect utility: its value scales directly with the number of people out walking with it. In markets where iPhone share approaches or exceeds half of all handsets, a huge fraction of potential contributors simply cannot participate. No App Store release exists, and until one does, the app’s ceiling is set by a platform decision rather than by its own design.

Worth naming plainly: this is largely a single-developer project with community contributions. The output has been extraordinary and the update cadence is steady. It’s also a bus-factor risk for a piece of infrastructure the OSM community now leans on heavily.

StreetComplete is among the most quietly important pieces of open-source software on Android, and the reason has nothing to do with it being free. It’s that Zwick identified the actual bottleneck in crowdsourced mapping — not willingness, not tooling speed, but the expertise tax — and designed it out of existence rather than lowering it. Turning a schema problem into a question is the kind of insight that looks obvious only in retrospect.

The limits are real, and there’s no reason to soften them. It can’t build a map, only enrich one. It leans hard on a battery. Its filtering is blunter than power users deserve. And a project this good being locked to one mobile platform is a genuine loss, one measured in every iPhone-carrying pedestrian who would have contributed and can’t.

None of that dents the core achievement. This is software that trains you to notice things about your own street you’ve walked past for a decade. That’s not a feature. That’s a side effect of software that asks better questions than most software bothers to ask at all. Install it, walk somewhere, and try to stop.