How many e-mails did you send today? What time did you swipe in with your keycard? Did you IM your friends? Did you make changes to a shared document? When you went to lunch, did you bring your BlackBerry? If your boss walked in the door right now, could you say how productive you've been? You may not have a ready answer, but those digital breadcrumbs you've been leaving can answer for you.
Cataphora is a Silicon Valley company that tries to model what an "effective" employee looks like based upon her electronic trail. The company began in the e-discovery field, dealing with the massive corporate databases that are now routinely subpoenaed as trial evidence. Imagine you're a lawyer tasked with going through a few terabytes of e-mail to see if there's anything damaging to your client. A keyword search isn't going to cut it; the linguists and programmers at Cataphora (and similar companies) specialize in sifting through electronic records to figure out what's useful and relevant.
Over the years, Cataphora has helped out in many cases where it's useful to know whether an employee is thriving within the company. This may indicate whether he will be a cooperative witness. Or take the example of a whistle-blower. While it's against the law to conduct a witch hunt and fire the whistle-blower, it's very advantageous to know, before you get into court, who the whistle-blower may be (i.e., is it someone in a position to give a lot of information to the government?). When dealing with these kinds of issues, Cataphora started with the basic tradecraft assumption that a happy employee is unlikely to cause problems.
But how do you tell if an employee is happy and working in the company's best interest? I discussed this with Elizabeth Charnock, the CEO of Cataphora and a mathematician by training. Often, she and her co-workers start by determining the normal level of electronic activity for individuals at a company. Then they go hunting for any deviation from the norm. Cataphora looks for who's using all-caps (typically a sign of high emotion) or who is communicating with people on a distant part of the org chart—a relationship that makes no organizational sense. Companies typically have "shadow networks" of people who consistently message one another. They can be harmless, or not.
They also look for "call me" events, when something being discussed electronically is taken onto the phone. Another sign of "something very personal going on" (and potentially damaging) would be if two employees who share a language, like Russian, switch from English to their native tongue midcorrespondence. Cataphora can build filters that find out when people are revising their résumés. They also look at the first drafts of employee reviews on hard drives, which are often much harsher and more accurate than the versions filed to HR. Another filter looks for threads where an employee contradicts herself—instances in which she e-mails "Great idea, boss!" and then one minute later IMs her pals: "Can you believe the idiocy around here?"
This kind of bad-apple analysis leads to the inverse question: What does an effective employee look like electronically? Let's say you were a big company about to buy a hot new startup. Wouldn't it be nice to know which executives in your target company actually make the decisions?