what3words Map App

Audit what addressing actually does

I started by looking at how addressing works in practice. Street addresses were often inaccurate or ambiguous. Road names repeated. Homes and businesses sat a long way from the centre of their postcode. Whole places had no address at all, from informal settlements to the park where people had arranged to meet. Around four billion people live without a formal address.

Bring the field research back to the room

I ran the research on site and then took it into internal workshops, where we went through the findings together and agreed what the real user problems were. Search, refine and save were not separate features but the three core scenarios from which the product needed to be designed.

Build around those three scenarios

Search accepts a three word address or an ordinary street address, so nobody has to learn a new habit before they can start. Refining is what turns a rough tap into the exact square, and it is the step the research said people needed most help with. Saving turns a square into a place with a name, which is what makes an address worth keeping. I worked alongside the engineers on the Google Maps integration, so the map underneath stayed familiar while the layer on top was ours.

The grid appears at zoom 19, not before

The map reveals detail in stages: continents around zoom 5, cities at 10, streets at 15, the three word grid at 19, individual buildings at 20. The alternative was to draw the grid whenever the map was open, which is the most direct way to teach what the product is. We did not, because at city zoom the squares are smaller than a pixel and the grid turns into a grey wash that hides the map underneath. The cost is that a first time user does not see the idea until they have zoomed a long way in, so the rest of the interface has to carry the explanation until then.

The marker changes shape rather than carrying a status label

A location is floating while it is only selected, locked once it settles on a square, saved when it is kept, and named when it has a label. Each of those is a different marker rather than the same marker with words beside it. The alternative was one pin and a status line, which is cheaper to build and far cheaper to translate. Four markers cost more design and engineering time and one more thing for the design system to hold. What they bought is a map that can be read at a glance while it is still moving.

The address resizes itself instead of being given a fixed size

Three word addresses are published in more than twenty languages, and the same address is a different length in each. The alternative was to set the type once and cut whatever did not fit, which keeps a layout stable. Cutting an address destroys it, because all three words are needed to find the square, so the type scales to the space and to the number of characters instead. The cost is that the address is not always the same size on screen, which makes it a weaker visual anchor for everything designed around it.

Look unlike Google Maps while being built on Google Maps

We integrated Google Maps through its APIs, so the tiles underneath were the ones people already knew. The alternative was a different base map provider, which would have given us control of the cartography itself. Staying with Google kept the map familiar and the integration proven, and it cost us the ability to change the base map at all. The difference had to come entirely from what we put on top: colour, typography and iconography.

What I would change is the order. We treated explaining the system as something the interface would do on its own, and the strongest signal after launch was people asking for that explanation directly. Teaching should have been designed as part of the product rather than left to the map to imply. The ideas I did not get to build were about turning a saved place into something worth sharing: lists that can be private or public, locations that carry a story, and a link to the 3wordPhoto app so a photograph and a place travel together. Those are my own ideas rather than the company's plans.

The prototype covered the three tasks end to end: find and save an address, navigate to one, and share one. It ran as a connected flow rather than a set of separate screens, so the movement from finding a place to correcting it to keeping it could be judged as one gesture instead of three.

The work moved through staged delivery, with fixed points for review, iteration and sign off, then validation before release, and then it repeated. Research, analysis, design, build and test each had a place inside a stage rather than a phase of their own. That is what let a feature go out, come back for revision and go out again without stopping everything around it.

Some of this work is not shown. What is withheld is what what3words treats as confidential, under the Data Protection Act 2018 and the Computer Misuse Act 1990. What is shown is used for portfolio purposes under fair dealing, sections 29 and 30 of the Copyright, Designs and Patents Act 1988. Every address in these screens is a placeholder from design review rather than a real saved location, because a three word address is precise enough to identify where somebody lives. The views here are mine and are not the company's positions.