Hacker News Reader

Comment by exhilaration | Hacker News Reader

I'm confused, throughout the article it mentions contractors being in oil facilities. But it also says "Orbital receives location data without any personal identification information and aggregates it."

So how do they know these people are contractors? Or are they just measuring the raw number of people in the oil facilities?

View on HN

Seemingly a contractor would be a person who is not typically at the oil field but stays for an entire work shift. Other people would be on tours or probably execs.

They are buying this data off ad providers and getting an ad id and geo coords, this is what leads me to believe they are just going off volume of users in the location

If you have lots of location data say in Houston, and then suddenly in a refinery that's not in their normal day-to-day location data you could assume it's a contractor being called into investigate an on-site problem. or if you detect increased activity from devices that don't have a history on-site. Not 100%, but then they're not going for 100% either.

Step 1: Get list of recent past problems at different refineries from news or people there after the fact or any number of ways.

Step 2: Get past location data that includes these periods (but spans a much larger time).

Step 3: Analyze what's different about the problem times compared to normal operation.

In some cases, it might actually be as simple as X% of work force doesn't come in for normal shift on what is otherwise a normal day.

The discoveries through big-data analysis and visualizations can be endlessly counter-intuitive, mindblowingly unpredictable but also dynamic and powerful, even visceral when you have massively voluminous data that can be fed into systems like Tableau where you can actively tweak data queries, aggregations, transformations and visualization types.

The result can often be something akin to: "wow, who could have ever predicted that this data would extrapolate out to show such a unique visualization or trend"

The article doesn't specify, they could be just looking at patterns of "usual" vs "unusual" ad ids showing up.

However, re-identification of "anonymous" ad ids is extremely common, and they could easily be using a service that gives them, for instance, a home address and email account associated with the phone via re-identification methods.

Contractors might have specialized skills related to refineries, so you could look for phones that spend small-to-medium amounts of time at refineries belonging to different companies. You'd probably catch some inspectors and the food truck guy and whatnot, but for the most part it should only be contractors moving around between companies like that.

You don't really need to know they are contractors, they are looking for a jump in the number of people onsite because access is controlled. A major turnaround or outage will bring in hundreds of extra contractors for days at a time.

They could get gps traces for each unique phone within the confines of the plant but labelled by random uuid. Then you can tell if a new one shows up, but with only locations inside the refinery you can't really say more.

Or the weather app is selling their location data to one or a hundred data brokers every 5 minutes. Mobile monetization via third party SDKs is a big thing. Likely 100k different SDKs to choose from (a number that came from a data aggregator).

Add in mobile operators, etc, and recombine the anonymous ids with the help of LiveRamp et al and it’s just a simple data problem.

Reply on news.ycombinator.com