Why did Facebook acquire Mapillary?

Why did Facebook acquire Mapillary?

Disclaimer: the views I express in my personal blog are just that: personal. They aren’t endorsed by my employer, my mother, my lawyer, or anyone else for that matter.

Photo by Joe Green on Unsplash

The big news in the geospatial world at the moment is Facebook’s acquisition of Mapillary. For those unfamiliar, Mapillary is a darling of the mapping world and one of the highest-profile geospatial startups of the last decade—launched in 2013, their mission was to create a global street-level imagery dataset to rival Google Street View.

Mapillary was the prototypical “venture-scale” business — preposterously ambitious, technically impressive, inarguably valuable for the world, and plausibly monetizable. What Mapillary accomplished in a short seven years is simply staggering. Google, with an enormous head start and untold resources at the ready to support Street View announced last year that they’ve collected over ten million miles of street imagery. Mapillary, on the other hand, crossed three million miles of mapped streets in 2018 and has more than doubled the number of images in their catalog in the years since (to over one billion!), putting them squarely in the same conversation as Google. That’s an insane accomplishment for any company, let alone a startup who, over its lifetime, raised ~$25M or about half of the annual compensation package for a typical member of Google’s C-Suite.

In addition, they’ve created some stuff I deeply admire as a member of the broader open source geospatial community. Two quick examples:

  • OpenSfM, a popular computer vision engine for stitching together overlapping images to reconstruct places in 3D.
  • Vista, a free 25K image dataset labeled for semantic segmentation — one of the largest such open datasets in existence.

Example annotated image from the Vista dataset

What is Facebook Up To, Exactly?

If you predicted Facebook would acquire Mapillary, congratulations — you are probably alone. Apart from a public image that stands in cartoonish opposition to Mapillary’s ethos of grassroots community building and radical openness, Facebook doesn’t really do maps, do they? Google, Apple, and Microsoft have invested billions into consumer mapping applications and acquired numerous companies to support those efforts. Any of those three behemoths would have felt like natural landing spots for Mapillary. But Facebook? What’s that about?

Unless you are already tapped into the seedy underworld known as the “geospatial industry,” you can be forgiven for not knowing that Facebook actually does do maps. In this incredible paper¹ released j̶u̶s̶t̶ ̶l̶a̶s̶t̶ ̶m̶o̶n̶t̶h̶ just over a year ago², researchers found that Facebook has contributed over 800,000 kilometers of mapped roads to OpenStreetMap (if you’re unfamiliar, it’s one of the largest crowdsourcing projects in history). They rank third in kilometers mapped behind Mapbox/Development Seed (1.69M) and Apple (1.64M).

A tidy summary of corporate contributions to OSM by type courtesy of Anderson, J.; Sarkar, D.; Palen, L. Corporate Editors in the Evolving Landscape of OpenStreetMap. ISPRS Int. J. Geo-Inf. 2019, 8, 232.

And beyond directly contributing cartographic features to OSM, they’ve publicly released an open source, AI-assisted road mapping tool called RapID that is an impressive thing to witness in action. They also support the OSM Foundation at the highest corporate giving level and have had a formidable presence at the annual OSM conference the past few years.

Still — the mere fact Facebook has dipped its gargantuan toes into the mappy water doesn’t explain why they would bother with acquiring Mapillary. So I’ve assembled three possible strategic reasons why it could make sense:

Justification #1: It Hurts Google

This is my leading theory. Reading through the acquisition announcement on Mapillary’s blog, the first thing that struck me was this section about halfway down the page:

Previously, you could pay Mapillary for special commercial licensing if, for example, you wanted to use their imagery to inventory street signs and sell that data to municipal governments. Now it seems that Facebook just wants to give it all away for free. If you’re like me and have 100 ideas for how this data could form the foundation of a commercial endeavor, that is rather exciting news. It raises an obvious question, though: why would Facebook want to pay a bunch of money to take a revenue-generating company and turn it into an unabated cost-center?

The answer, I think, is to hurt Google.

This is a strategy known as “commoditizing your complement,” that I first read about in an excellent article from the mysterious internet journalist/researcher Gwern. Essentially, if you’re getting your ass handed to you by a competitor with a software or data advantage you can’t replicate yourself, you can use permissive licensing as a weapon to erode that advantage by encouraging the world to collaborate with you. Google itself is no stranger to this strategy; their choice to acquire Android and support it as a free and open operating system for smartphones is an example of this tactic. In that instance, the complement was Apple and iOS.

Google is the clear leader in consumer mapping, and their Google Maps app consistently ranks as one of the most downloaded smartphone apps in existence. For a bunch of companies that sell your private data for incomprehensible sums, the value of a top-ten downloaded app that simultaneously gives you a “legit” reason to ask for the user’s location data AND provides an advertising opportunity is hard to assign a value to…I would guess it’s comfortably in the tens of billions.

So if you’re Facebook, Apple, Amazon, or Microsoft and you’re wayyy behind Google in this very valuable category…why not all collaborate to simultaneously improve your collective offerings while eroding the value of Google’s? Street View is one of the signature elements of Google’s map offering (and if you’ve ever read the laundry list of limitations on how it can be used by 3rd party apps, you’d appreciate how jealously Google guards it).

Thanks to my colleague Deborah Boyer for pointing me to this lil’ tidbit in the Street View terms of use.

Buckle up Google, because that mildly annoying upstart Mapillary suddenly has pockets as deep as yours. And judging by the high-profile names investing resources into OSM, the rest of your mapping hegemony may be next.

Justification #2: It Complements Their Augmented Reality Business

In a now famous leaked internal memo from 2015, Mark Zuckerberg laid out Facebook’s strategy for investing heavily in the emerging field of VR/AR. It boils down to an obsessive fear of missing the next great platform shift. Facebook survived the move from desktop to mobile by aggressively investing in mobile well before it was obvious that shift would make or break all social media companies. And while VR/AR may seem like a curious plaything now, the risk to Facebook of getting a slow start during another tidal platform revolution is much greater than their fear of being mocked for being too early. These are not my ideas, but I can’t remember who I’m stealing them from…probably Ben Thompson.³

How does street level imagery fit into this strategy? Well, the same week this acquisition was announced, rumors began swirling that Google is testing out adding AR objects/markers to its street view imagery:

Facebook’s $2B acquisition of Oculus showed they were serious about the entertainment side of AR/VR, but I suspect their plans are much broader. Eventually, Mapillary might become onepillar of a suite of tools and data aimed at commercial uses for the technology. The fact it’s free to develop on top of for commercial use cases would certainly make recruiting an ecosystem of 3rd party developers to build value-added applications alongside the imagery much easier (and is clearly not something Google is prioritizing with their current licensing scheme).

Justification #3: It Supports Facebook’s Place Data Generally

Lest we forget, Facebook is in the business of knowing who you are, where you are, what you like, and what you don’t. If you discover a new restaurant through Instagram, and then you leave that platform to open Google Maps and read reviews of that restaurant, Facebook (or the Facebook company or whatever they’re calling themselves now) has just lost a very valuable interaction. But what if street level imagery and supplementary data like reviews were built directly into Facebook’s products or at least resided somewhere on a Facebook property that could be linked out to? That at least would keep users in their scrupulously measured web of control. Facebook’s company pages don’t offer any kind of street level view (whereas Google’s company pages certainly do) and they link directly out to Google Maps, which I imagine feels like waving a huge white flag of surrender to whoever the product manager is for Facebook pages:

Facebook links out to Google Maps in their own product.

If you’re Facebook, I would guess that sometimes it’s easier to negotiate buying a company outright than it is to negotiate a custom licensing deal that is orders of magnitude larger than any other deal that vendor will ever see in the life of their company. Because of their scale, the value of a particular data asset to Facebook may be entirely dislocated from the rest of the market. At a certain point, you just have to ask yourself, “Instead of buying that, why don’t we just buy them?”

Read More