While evaluating new tech startups’ products, I keep my feedback on a piece of paper with two columns, creatively labeled “helpful” and “unhelpful”. It’s useful not just for organizing my thoughts, but as a reminder not to be a hypercritical ass, which can be easy for an analytical person like myself.
But recently, I’ve had to start adding in a new column: “Creepy”. Creepy tends to be:
- Helpful… but unsettling in a way that’s difficult to quantify
- Done with good intent
- Rarely used by the founders/developers
- Driven by data
Let’s look at an example of creepy:
Exhibit A: LinkedIn’s Who Else Might You Know
LinkedIn has a marvelously efficient on-boarding process - every little step is focused on getting you set up and connected with people as quickly as possible. It makes sense for a social network: I’m sure the analytics says that once someone makes 7 or 10 or whatever connections, they’re 50% more likely to stick around, so everything is funneling you towards making new connections.
Take for example the “Who else you might know” feature. This helps you quickly get a few connections by suggestions people you might be connected to; on a recent experience, it helpfully suggested that I connect with several of my housemates. Awesome, right?
The only problem was that I was using a randomly generated test account: no personal information, a clean email account, and generic login details.
Indignation flooded my thoughts as alarm bells rang loudly in my head: how did they know who I was? Are they tracking my web activity? How the hell are you getting this information?
I’m sure there are perfectly reasonable explanations. They probably cross-referenced that IP I logged in from with other members in the database. They probably couldn’t care less about my extensive web history of procrastination. They’re probably doing things that are just par for the course for any highly data-driven company. But the immediate impact I felt was that of disgust, indignation, and violation. Certainly not the things you’d like a new user to feel.
Data Driven, not Data Only
As the “Big Data” meme continues, I fear that this trend will grow. Being able to pinpoint those fractional improvements in conversion or betting on the statistical correlation between mouse position and LTV will only make this worse.
Even in my own small-scale tests, I found it hard to argue against treating your users like idiots because the data clearly shows that it makes them convert better. How could anyone at Groupon argue against sending even more daily-deal emails with millions of data points definitively linking more emails to more dollars? How could anyone at Zynga argue that they were pissing off new users with relentless game-spam when they were signing up new users like gang-busters?
How can you possibly argue against something that works but feels wrong? What is the P-value of decency?