Yahoo has been in the news lately for a change in its practice of remote work. An abundance of commentators have taken this opportunity to draw conclusions on everything from feminism to global warming, with stops along the way for class warfare.
I see a story about how data and Operations relate–and no one else, in my observation, has told it.
What we know we don’t know
Recognize first how little those of us outside Yahoo truly know about what has happened and is happening. Kara Swisher, who broke the story on 22 February 2013, was characteristically precise and articulate. The shift begins in June, and applies only to several hundred customer-service representatives and others who work full-time from home. Or perhaps it also applies to those who split their hours between remote and in-office. While some outlets label Swisher’s sources as “leak”, and the memorandum is topped with “… PROPRIETARY AND CONFIDENTIAL … DO NOT FORWARD”, we have no way of knowing whether Yahoo executives minded the “leak”, or perhaps even were counting on it as part of their planned cultural adjustment.
Yahoo chief executive officer (CEO) Marissa Mayer had no documented role in this first round of news. Pundits have more-or-less consistently labeled the decision hers, at the same time as Yahoo itself is saying as little in public as possible. We’re at a phase in popular culture which can’t resist taking the synecdoche too literally: Yahoo is Mayer.
This weekend, Mayer made her first explicit appearance in the affair: Swisher quotes a (different?) anonymous source who heard Mayer tell “an employee meeting last week that VPN logs showed work-at-home staff did not sign on enough“. Shortly after, Nicholas Carlson of BusinessInsider quotes an anonymous source to the effect that “the only way Mayer is comfortable making any decision [is] with the help of data“.
Even if all these quotes are verbatim and rendered with perfect fidelity, they might be conveying an impression far different from the truth. We don’t know who decided to look at the virtual private network (VPN) logs–or if the logs had been under scrutiny for longer even than Mayer’s tenure, but only now served a purpose decided on some other basis. “VPN logs showed …” can mean anything from, “an entry-level sysad brought up the records on three particular employees for the current day”, to “the company runs the past decade of events through a data warehouse with a million-dollar analytics front end”. Mayer competed on debate teams at Wausau West High as the ’90s began. In that sort of training, “help of data” can refer to support of a conclusion reached without, or before, the data.
I’m not criticizing Mayer. My point is to emphasize how little certain knowledge the current flurry of attention on Yahoo has brought us.
Late-night log reading
My own experience with logs and remote work have taught me a few things about which I’m personally certain:
- Study of logs is always an education. Every time I look through logs, with remarkably few exceptions, I learn. The most reliable signal I’ve seen in VPN logs in particular is not that employees aren’t “signing on” enough, but that some people have enormous capacity for on-line shopping.
- Pulling employees back into the home office doesn’t in itself solve problems.
- If the overt narrative is true–if the CEO learned that she has employees who aren’t working because the logs told her so–then the biggest failure is on the part of managers, not the customer-support reps who have been identified for “correction”. Is “log on” the most advanced metric Yahoo has for its front-line employees? I doubt it; the story was probably simplified for the purposes of public relations. If that is the way Yahoo operates, its problems are even bigger than any of us on the outside realize.
I’ve written before about how ambiguous real-world “data-driven” initiatives are. Mayer cultivated a reputation at Google for her engineering insight and lack of sentimentality. Yahoo’s remote-work policy is not a simple matter of “the data made her do it”, though, I’m sure: as with any other interesting question, humans choose which measurements they make and feed into the algorithms.
Later this month, “IT Ops” will look at the details of a few clear “wins” for data-centered work.