Pulling back the curtain on location intelligence
Our interview with Josh Anton, CEO of X-Mode, takes up the state of location, what marketers need to look for in a data partner and, of course, privacy.
There are perhaps 20 companies offering location data or location analytics. X-Mode is less well-known than many others but says it’s one of the few, primary sources of “first party” location data in the market. We caught up with Josh Anton, founder and CEO of X-Mode, to get his take on the current state of location intelligence, what marketers need to look for in a data partner and some of the changes coming with stricter data-privacy rules.
ML: What does X-Mode do?
JA: X-Mode was founded by the people behind the widely popular campus safety app Drunk Mode. We work with app developers and data buyers to offer the highest quality location data that meets all current regulatory standards including the GDPR and CCPA.
X-Mode has one of the most accurate location data panels in the industry and receives the majority of its data directly from mobile app publishers through XDK, its proprietary location-based SDK. With over 300 apps on its platform, X-Mode licenses a high accuracy (70% accurate within 20 meters), dense data panel that includes mobility metrics (speed, bearing, altitude, vertical accuracy), near real-time GPS, and other detection capabilities (IoT, Wi-Fi, and Beacon).
X-Mode provides this anonymized user panel to hundreds of clients across multiple industries including Mapping and Location Services, AdTech, MarTech, FinTech, Smart Cities, Real Estate and InsurTech.
ML: What was Drunk Mode and how did it evolve into X-Mode?
JA: Drunk Mode was the living map for when you went out partying. Drunk Mode stopped you from drunk dialing your friends, allowed you to find your drunk friends, and showed you where you went last night. A lot of our focus was how we leverage location to make the college student’s night a bit safer.
To continue start generating revenue Drunk Mode began monetizing location data in 2015, as display advertising opportunities were limited and low value. During this time our team saw two major opportunities for disruption in the location data industry: 1) data licensees’ need for high-quality location data and 2) publishers looking for incremental monetization. Due to Drunk Mode’s investment in accurate location technology for college safety (breadcrumbs), we already had established contracts monetizing location and understood the pain points of having users opt-in to sharing location. We were in a unique position to solve numerous issues and generate a “win-win-win” for publishers-consumers-X-Mode.
Thus, the X-Mode Location Data Network was born in Q2 2017. We leveraged our new XDK 1.0 that was built off the core underlying technology of our Drunk Mode application and now powers X-Mode’s location platform. Having grown our network to 65M+ global users in less than 2 years, we realized early on that location-based use cases we built around Drunk Mode had a much larger impact than we ever imagined.
Instead of creating a network around college safety, we can now help optimize emergency routes. Rather than just offering drunk food discounts to users after a long night out, we can now power fortune 500 companies’ ability to better optimize their ad-spend to target around location-based moments at scale. In the past, we gave Uber/Lyft discounts to help college students get home safely, and now at X-Mode have the power to help optimize transportation routes for the masses. We realized there was a much bigger world outside college and Drunk Mode sobered up to what people know as X-Mode today.
ML: You made the statement that X-Mode was one of a small number of “first party” location-data providers in the U.S. You also suggested there’s only a finite supply of available location data in the market. Please elaborate.
JA: If a location company wants to build an audience or measurement product focused in the U.S. that their end clients use, that company would typically need to combine bid-stream data, low-quality aggregator data, their own always-on SDK data and/or first-party SDK companies. They would need to get to 30M DAUs/75M MAUs for measurement and 30M DAUs/250M MAUs for audience retargeting.
Even if one combines the top three companies in the location space, 70% of the true first-party data in the market (from an “always-on” location SDK), you only see ~30M DAUs in the U.S. (accounting for overlap). If you go downstream, there are only about 5,000 apps with over 2,000 DAUs that have appropriate permissions to run “always-on” location, with about 40% of that number monetizing location.
The reason why it seems like there is an infinite supply of location data in the market is because there’s still a huge number of companies taking low-fidelity data from ad-based SDKs (bid-stream) and creating derivative products to achieve artificial scale. This approach isn’t useful for measurement. Even worse, this approach does not have the privacy permissions needed to navigate a privacy conscious world.
Almost every location intelligence company has some sort of SDK. However, the real questions that people should be asking when it comes to understanding location data licensing are the following:
- What percentage of users come from first-party SDK data vs. bid-stream or aggregated ad-based SDK data? Can you name the suppliers that make up your feed under a non-solicitation? Most importantly, how do you really know it’s coming from an SDK?
- Do you know whether this data is coming directly from an app and not just recycled data from another 3rd party? Ask for the app categories and redacted screenshots of some of the larger apps’ privacy policies under a non-solicitation.
- Is the data being collected directly from an app? Ask questions about collection methodology; a clean panel will have a pretty standard methodology across the board.
The best panels on the market for measurement or audiences will have 60%+ of their data sourced from their own SDK or from other first-party SDK companies like X-Mode. However due to economics, and buyers not realizing that there are only a handful of companies that control first-party data, companies default to building their data products off of low-fidelity, low-cost data. Companies buying location data often think of it as a commodity, without thinking about data quality and privacy implications, which will occur in the coming months as it becomes much harder to sell data where neither the publisher nor users know their data is being monetized.
ML: You said 60% or 70% of X-Mode data has 20-meter (or better) accuracy. How is that accomplished?
JA: We use high-accuracy GPS settings at a specific cadence and machine learning to understand when the best time to trigger location may be, around a visit/movement. Then we layer beacon trilateration to enhance our collection methodology, which helps when mapping locations in a mall or a dense city block.
ML: Many companies get location from the bid-stream but claim to clean it up and discard inaccurate data. Are you skeptical? And what role, if any, does bid-stream location data have to play in the ecosystem?
JA: I am skeptical because there are actually two issues with bid-stream data. The first is data accuracy (already discussed). The second is a lack of persistent collection. With bid-stream data, you are only capturing location when someone views an ad. It’s limited to online behavior.
ML: What impact do you believe CCPA will have on location data, in terms of its availability and quality?
JA: Right now, there is a lot of fluff in the market. Only three main first-party suppliers of location data in the USA (X-Mode, Cuebiq and Foursquare) control 70% of the first party supply of background location in the market. These companies not only work with publishers directly, but also have a quality SDK and control the relationship with the publisher to pop-up the proper opt-ins needed to navigate CCPA or the other 30+ states that are passing some sort of legislation requiring explicit consent.
Third-party aggregators out there, getting data either through ad-based SDKs (where publishers may not know their data is being monetized) or through the bid-stream, will have issues not only sourcing data at scale, but also providing that data with the proper consent mechanisms needed by agencies and brands.
Privacy is a good thing because it gives consumers more control over their data. At the same time wipes out a lot of the “fluff data” in the market coming from 3rd party aggregators. Companies building location-based solutions will have to rely on first-party SDK data companies like X-Mode, Cuebiq, and Foursquare to power their solutions so they can stay on top of privacy and control quality. In the next 18 months, I expect to see both consolidation of first-party SDK players like ourselves and folks that are low quality aggregators to pivot or evolve into analytics or other location-based tools upstream.
ML: What are the most important things brands and marketers need to understand about working with location data?
JA: The most important two things brands and marketers need to ask themselves when looking at targeting and attribution are:
- How did you calculate a visit and dwell time?
- How was the data sourced to create this visit/dwell time and how confident are you about each visit/dwell time? What filters do you have in place for outliers (i.e., what was the cadence of collection, filters for speed when someone’s driving, etc.)?
- What is the average number days seen across your panel (i.e., DAU to MAU Ratio)? 2+ weeks is the gold standard but anything above 5+ days is not bad.
- In terms of how you map location data to a visit, what’s the high-level black box that you used to do this (i.e., polygon, check-ins, point radius, etc.)? Polygons and check-ins are going to be much more accurate than point radius, which is what most companies use today. Foursquare and Safegraph were ahead of the game here.
Transparency and privacy:
- Where was this data sourced and how was it collected: server to server vs. SDK vs. directly from an app?
- How are you determining or mandating you have the legal right to use this data for your defined use-case — that is: the contract terms and privacy policies of some of the larger contributors to your panel?