Why bother with What Three Words?

by @edent | # # # # | 34 comments | Read ~51,515 times.

I'll be wording this post carefully as What 3 Words (W3W) have a tenacious PR team and, probably, have a lot more lawyers than I do.

W3W is a closed product. It is a for-profit company masquerading as an open standard. And that annoys me.

A brief primer.

  • The world is a sphere.
  • We can reference any point on the surface of Earth using two co-ordinates, Longitude and Latitude.
  • Long/Lat are numbers. They can be as precise or as vague as needed.
  • Humans can't remember long strings of numbers, and reading them out is difficult.

W3W aims to solve this. It splits the world into a grid, and gives every square a unique three-word phrase.

So the location 51.50799,-0.12803 becomes ///mile.crazy.shade

Brilliant, right?


Here's all the problems I have with W3W.

It isn't open

The algorithm used to generate the words is proprietary. You are not allowed to see it. You cannot find out your location without asking W3W for permission.

If you want permission, you have to agree to some pretty long terms and conditions. And understand their privacy policy. Oh, and an API agreement. And then make sure you don't infringe their patents.

You cannot store locations. You have to let them analyse the locations you look up. Want to use more than 10,000 addresses? Contact them for prices!

It is the antithesis of open.


W3W refuses to publish their prices. You have to contact their sales team if you want to know what it will cost your organisation.

Open standards are free to use.


When an earthquake struck Japan, street addresses didn't change but that their physical location did.

That is, a street address is still 42 Acacia Avenue - but the Longitude and Latitude has changed.

Perhaps you think this is an edge case? It isn't. Australia is drifting so fast that GPS can't keep up.

How does W3W deal with this? Their grid is static, so any tectonic activity means your W3W changes.


Numbers are fairly universal. Lots of countries use 0-9. English words are not universal. How does W3W deal with this?

Is "cat.dog.goose" straight translated into French? No! Each language has its own word list.

There is no way to translate between languages. You have to beg W3W for permission for access to their API. They do not publish their word lists or the mappings between them.

So, if I want to tell a French speaker where ///mile.crazy.shade is, I have to use ///embouchure.adjuger.saladier

Loosely translated back as ///mouth.award.bowl an entirely different location!

You're not allowed to know what word lists W3W use. They take a paternalistic attitude to creating their lists - they know best. You cannot propose changes.

Anecdotally, their non-English word lists are confusing even for native speakers.

Cultural Respect

Numbers are (mostly) culturally neutral. Words are not. Is "mile.crazy.shade" a respectful name for a war memorial? How about ///tribes.hurt.stumpy for a temple?

How do you feel about ///weepy.lulls.emerge and ///grouchy.hormone.elevating both being at Auschwitz? Or ///klartext.bestückt.vermuten - "cleartext stocked suspect"?

This is a classic computer science problem. Every sufficiently long word list can eventually be recombined into a potentially offensive phrase.

Open Washing

W3W know that the majority of technical people are not fooled by their attempts to lock down addressing.

They include this paragraph to attempt to prove their openness:

If we, what3words ltd, are ever unable to maintain the what3words technology or make arrangements for it to be maintained by a third-party (with that third-party being willing to make this same commitment), then we will release our source code into the public domain. We will do this in such a way and with suitable licences and documentation to ensure that any and all users of what3words, whether they are individuals, businesses, charitable organisations, aid agencies, governments or anyone else can continue to rely on the what3words system.

I don't know how they propose to bind a successor organisation. They don't say what licences they will use. If they go bust, there's no guarantee they'll be legally able to release this code, nor may they have the time to do so.

There's nothing stopping W3W from releasing their algorithms now, subjecting them to scrutiny by the standards community. They could build up a community of experts to help improve the system, they could work with existing mapping efforts, they could help build a useful and open standard.

But they don't. They guard their secrets and actively promote their proprietary product in the hope it will become widely accepted and then they can engage in rent-seeking behaviour.

This is not a new argument

My mate Leigh wrote about this three years ago. Lots of people have criticised W3W.

But W3W have a great PR team - pushing press releases which are then reported as uncritical news.

The most recent press release contains a ludicrous example:

  • Person dials the emergency services
  • Person doesn't know their location
  • Emergency services sends the person a link
  • Person clicks on link, opens web page
  • Web page geolocates user and displays their W3W location
  • Person reads out their W3W phrase to the emergency services

Here's the thing... If the person's phone has a data connection - the web page can just send the geolocation directly back to the emergency services! No need to get a human to read it out, then another human to listen and type it in to a different system.

There is literally no need for W3W in this scenario. If you have a data connection, you can send your precise location without an intermediary.

What Next?

W3W succeeds because it has a superficially simple solution to a complex problems. It is a brilliant lesson in how marketing and PR can help a technologically inferior project look like it is a global open solution.

I'm not joking. Their branding firm says:

Edelman helped what3words frame their story to be compelling by tapping into human emotion.
We also created a story for CEO Chris Sheldrick about how having an address can drive social transformation and business efficiency, securing profiling and speaker opportunities.
Through paid social campaigns we re-targeted these stories, getting through to the decision makers that mattered most.
We articulated their purpose narrative and refined their strategy to engage investors and excite the media.

It takes too much time to refute all their claims - but we must. Whenever you see people mentioning What3Words, politely remind them that it is not an open standard and should be avoided.

34 thoughts on “Why bother with What Three Words?

  1. Sam@sammachin.com says:

    Personally I really like w3w and their solution is certainly novel, I take your point about an open standard but equally these people have worked to create a novel and effective solution so why shouldn't they be able to profit from their labour?
    We could apply the same argument to say UK Postcodes, despite the creation of the system having been funded out of public money (when the Royal Mail was state owned) they still want to charge people for access to the dataset? At least w3w have a nice API and is free for basic use, more than can be said for Royal Mail

    1. @edent says:

      I agree that W3W should be allowed to profit from their labour. I just think they should be more honest about the fact it is proprietary and that they should stop spinning it as an globally useful open standard.

      I also agree that RM's postcodes are also a closed system - and have been fighting against that for ages 🙂

      1. This seems like a false dichotomy. An open standard can be completely compatible with income.

        For example, what3words could sell services around the standard, akin to SUSE selling support for Linux (and doing very well at that). Or, perhaps more immediately comparable, look at the OpenStreetMap ecosystem which has plenty of businesses like Klokan Technologies and Geofabrik.

        From a business point of view the openness could actually lead to adoption and increased demand for the services.

    2. Dan says:

      Thinking GPS coordinates needs an equivalent of DNS does not seem terribly novel, and how is the solution effective?? Go to another country and suddenly nobody will understand. Tell someone you are at location "one.two.three" and they will probably also respond with three words: "WTF"

  2. Fabian Homborg says:

    As a native german speaker:

    Or ///klartext.bestückt.vermuten - "cleartext stocked suspect"?

    This example seems benign to me - it's "suspect" as in "guess" as a verb, not a criminal suspect. (And stocked like a store shelf would be)

    1. @edent says:

      Thanks for letting me know! The perils of autotranslate 🙂

    2. Wacokjacko says:

      Which actually illustrates the problem of translation issues beautifully, what is benign is one language can be loaded with innuendo in a translation.

  3. Ervin Ruci says:

    Another problem I see is its lack of relation to actual geography. People have already named places around the world in their own ways. We should adapt to them, not the other way around.

    The quest for the optimal addressing system goes on.

  4. Jonty says:

    FWIW there's a standard called OLC/Pluscodes, which achieves a similar thing and is actually open: https://en.wikipedia.org/wiki/Open_Location_Code

  5. Jacek says:

    Here's the thing... If the person's phone has a data connection - the web page can just send the geolocation directly back to the emergency services! No need to get a human to read it out, then another human to listen and type it in to a different system.

    Well, but then you need a special webpage made by the emergency services. They didn't have that; they just had what3words. You don't get to disparage things that exist in favour of things that don't exist.

    You could of course try to build that page for emergency services, sell it to them, and have them maintain it. Or, you know, the dispatcher could just tell the person to open what3words...

    1. @edent says:

      You are correct. I took their press release to mean that they had a special page for emergency services. Their videos make it clear they're just sending a generic text - https://what3words.com/mapsitecontrol/

      As it happens, I did build a similar site many years ago when working for a telco. It was half-a-dozen lines of JavaScript, POSTing back to a server. I'll see if I can drag out the code.

      1. Tim says:

        Actually wouldn’t the link be even more useful if it just auto-responded with the GPS Geo Location like the phone already has? That way you wouldn’t need to rely on a data download of graphics for a map, or someone who might not be able to speak, as opposed to a few bits being uploaded to a web-server and that doing all the back end translation from a compressed format? Assuming the 51.50799,-0.12803 example was easier to send compressed as opposed to plain text or in hex/binary?.

  6. When people meet me for the first time they often say "Oh, you're the guy who hates What3Words", because I blog and tweet about them a lot, e.g. at http://grcdi.blogspot.com/2017/11/when-is-implemented-code-not.html . It may seem that they are the only ones in my firing line, but that's because they spend so much on public relations that they seem to be constantly in the media who, as you say, then report uncritically what has been spoon-fed to them. It would be great if once, just once, they would ask the opinion of address experts before publishing... But anyway, there are actually a lot of these algorithm-based systems around, too many of which are looking not to solve any real addressing problems, but to make a quick buck. If you're interested, I analysed those systems I know about and compared them to traditional street addressing. You can view it free here: https://www.grcdi.nl/Address_encoding_versus_traditional_addressing–the_state_of_play_20181030.pdf

    1. Mr. Rhind’s white paper is excellent. I recommend it for those in this area.

  7. Andi says:

    Open Location Codes solve this problem. It works out much better with scale, and it's open source.

    1. Thanks for posting this. I hadn’t heard of Plus Codes before. Still not as elegant as W3Ws.

      I do find it quite elegant, but there’s no reason why someone couldn’t riff on these ideas in a way that was more elegant. We remember words better than random letters. I’m sure someone could generate short phrases that map to a similar grid. Might also be able to deal with things like elevation (what floor of the building are you on).

      It is totally worth while getting compensation for the work that you’ve done to push out an idea, but on the other hand if it is going to be used as an essential service then releasing the algorithm is critical. You have to engage the locations independently from the W3Ws service.

  8. Satan Claus says:

    I mostly agree with the OP's gripes, but at the same time acknowledge that some form of projection from decimal degree coordinates to an equivalent tokenized form, whether it be 3 words or something else, can be useful. For instance, in some places in Dubai (my family have recently moved there) there are simply no street addresses. Directions are very much like "down the hill, to the right, near the clump of trees, 50m onwards and near the green bin outside the iron gates opposite the ditch". Clearly this sort of arrangement could benefit from some sort of simplified/tokenized geocoding of a location for addressing purposes - but yes, again I don't think W3W is particularly useful or clever for that matter. Conceptually at a high level it's intentions seem good, but the way it's implemented is not particularly useful - to my mind anyway. Promoting it as something useful to emergency services... uhm... no. Location of fire hydrants or trees or something, or just a rough address of a property - yes perhaps. But it should not be so guarded and shrouded in proprietary secrecy - as the OP says, if it's of any use, open source it. Stop this possessive behavior and attitude of "I've reinvented the world and it's mine mine mine"...

  9. someone says:

    Having experimented with the W3W api while teaching myself, I couldn’t help but notice that the words changed even when I opened them from the same exact place from time to time. Granted, this could be related to the browser or ISP but not always. Still, I admire their tenacity for trying to map out the world in a more humane manner. There is something about the idea of not having to memorise many different addresses that is inherently appealing although I agree with each and every point you’ve made here.

  10. Of course, now with your blog post which is trending on HN now, you are helpin W3W to become even more prominent 🤔

    PS: the German example is weird but not offensive

    1. Babalak says:

      Far from that. In fact he’s exposed the many problems to many people like myself who initially thought this was a brilliant idea. It turns out on closer inspection not to be such a great idea, even idiotic

  11. AlisonW says:

    W3W is an atrocious waste of space and effort. A couple of years ago I was developing a gis/location system for emergency services and, very briefly, looked at their offering alongside a wide range of tools and methods. It fell down on every point, especially that of having multiple ways to describe a single location, and that of having no inherent indication of distance between two points. The very randomness of the wordlist makes it impossible, indeed it also requires a live connection at all times which is frequently unavailable or of low bandwidth in the field. A service which has no unique usecase.

    1. Marco says:

      Actually, it doesn’t need a live connection. Read their website. It uses the GPS in your phone and as such available offline.

      Everyone moaning about the fact it’s not open source. Is Apple open source? Windows? Word? Excel? But people still use them (and pay for them). Why do people feel that in the Internet-age everything should be available for free? That’s just BS.

      1. @edent says:

        You are correct. If you already have the app, it doesn't require a live connection. If you don't have the app, you'll need to download and install a 50MB file.

        Apple does produce lots of open source - https://opensource.apple.com/

        Microsoft, famously, publishes a huge amount of open source on GitHub - https://github.com/microsoft

        I think it is entirely reasonable to say emergency services should be free to use (gratis). We don't pay for making 999 / 112 calls.

        It is also prudent that safety critical software should be free (libre). It is important that we can examine the code to make sure it doesn't contain any nasty surprises.

      2. James says:

        We're back to use cases, yet again. In some use cases you'd have access offline -- but there are lots of use cases, including those W3W themselves tout (like being given a W3W location in lieu of an address, or emergency services asking for your W3W location) where there's no reason to assume the consumer would have an app pre-installed. In that case, yes, you need a live network connection to either dereference the W3W location into a lat/lon (in the former case) or look up your current W3W location (in the latter case). This is because the actual product that W3W sells is basically a database -- you can download it, or you can make lookups over the network, but because it's a closed product, it's never going to be included (especially offline) in the mapping software you're already using.

  12. Orion Edwards says:

    The thing I most dislike about w3w is that there's no spatial locality or reference

    By that I mean, given two addresses, there's no way to guess where they are or compare them and determine how far away they are.

    Addresses obviously have this property. I can intuitively determine that "42 Wallaby Way, Sydney, Australia" is next to "44 Wallaby Way, Sydney, Australia", and the same holds for GPS co-ordindinates. Even postcodes - while not deterministic - often cluster together.

    With w3w though, where is levels.nine.coins? How far is it from fork.scruples.wage? It's a total mess.

  13. P J Evans says:

    I can see where something like what3words would be very useful in situations where there’s no handy street address – like rural areas – or when you have limited communications. Lat/long, while covering the entire planet, requires a lot of numbers to get 10m accuracy. (I lived in west Texas for several years. Our address was a mailbox on a route number, and yes, we could get deliveries there from commercial companies – but we had to describe the location for them to find it (two miles west of a and two miles south of b, on road n); they’ve implemented house numbers since then, to make it easier for emergency services (such as they are in the rural areas), but it’s still necessary to tell people where it is.

  14. Nick says:

    I vote for the military grid reference system and to be done with it. Yeah it’s still numbers and letters, but I would take a combination of 11 letters and numbers to get me within 10 meters any day. Oh, and it can be overlaid on any UTM map in the world.

    1. James says:

      See Orion’s comment above about spatial locality. MGRS does have some locality but takes practice to understand, because it interleaves latitude and longitude so you have to do some mental acrobatics to figure out if two points are near one another, outside of the first 3 characters (top level grid).

      I notice that use cases don’t come up often in this discussion. What are we doing with this point reference? Reading three (2-3 syllable?) words to somebody over a spotty phone connection actually sounds pretty good, and certainly better than reading 11 character names (especially if one or both parties don’t know a common phonetic alphabet), or let’s say 12-20 digits of a decimal lat/lon. Plus Codes could arguably be better (though I think it can have the same phonetic-alphabet problems) with a useful location in about 7 characters, assuming both parties know the general area being referenced. And of course, that’s one pretty specific use case and there would be others where different systems make sense. Do I actually want the easiest system for remembering an address? Could be completely different.

  15. Prof Livingstone. says:

    https://en.wikipedia.org/wiki/Maidenhead_Locator_System – is a system where you

    can figure out distances between two locations
    you can decide yourself how granular you want to be, granting you some geospacial privacy
    culturally neutral
    works internationally
    is a free standard

    Maidenhead Locators are in all respects superior to w3w

    1. James Bromwell says:

      The most commonly advertised purpose is postal addressing, so you’d want precision over privacy. If I understand the maths correctly, it would require a 12-character string to achieve comparable accuracy. If we’re being honest, can we really say that a string of 12 random alphanumerics is more memorable as a postal address than 3 words? Is it easier to tell somebody, or write on a form? Can I read it correctly from that form, given that it doesn’t appear to remove visually-ambiguous pairs (1/L, O/0, etc)? Can an untrained person who doesn’t know a phonetic alphabet by heart read it accurately over a poor-quality phone call?

      For most of the use cases I’ve seen pitched, OLC would be “in all respects superior” to Maidenhead. Again, though, you’ve got to ask, what are you trying to do with it?

      1. Ervin Ruci says:

        A 10 digit string can have accuracy within 1 meter. Anything else is just splitting hairs.

  16. Matthew says:

    I looked at the first “uncritical” link (BBC News) and maybe it’s been updated? If so, it’s probably thanks to you:

    “What3Words has, however, faced criticism, not least for the restrictions it places on how its data is used and the fact it charges companies to incorporate its service into other products.” … etc.

    I also don’t like “W3W” because it sounds so similar to that paragon of open standards, the W3C.

  17. Phil says:

    Regarding the emergency services sending a link and the location being sent back to them. A system achieving this is successfully in use by UK Mountain Rescue (and other UK emergency responders) and has been rolled out globally to emergency responders.
    SARLoc was originally developed by a Team Member of the Ogwen Valley Mountain Rescue Organisation and records verified hits and uses multiple times a week in the UK alone.
    The only time What3Words would make work easier would be if no data connection existed and only voice or text messaging could be used. But in this instance the app would already have to be pre-installed on the user device. Even then, the only benefit would be the ability to phonetically spell words more easily than numbers via a bad line or the ability of W3W to ensure that misspelled words in a text are still pointing to a relevant location.

    All in all – as you say W3W are producing a great example of faking Openness and trying to sell people a solution to a problem that by and large does not exist in most developed western nations.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.