Writing

Software, technology, sysadmin war stories, and more. Feed
Sunday, March 24, 2013

Using numbers to justify anything

Online services come and go. Sometimes, the justification given is that "only n% of our users were using it". This seems to be an easy way to make things really bad for people. Here's what I mean.

Let's say you're a little company. You have 100 users. You make a change to your product that really annoys 1% of them. Fortunately, 1% of 100 is just one person, so while that one person is a loss, it's probably the sort of thing you can survive.

Your company grows and now you have 100,000 users. You make another change and again annoy 1% of users. This time, 1000 people are miffed as a direct result of your actions (or inactions). That's a bit of a crowd, but you can probably pass it off as "a handful of hard-core users who don't represent the whole". Whatever helps you sleep at night.

Some years down the road, you pretty much dominate whatever sector you are in, and you have at least 100,000,000 users. Again, you annoy 1% of that population. This time, a cool million people are unhappy with you. That's basically the entire city of San Jose, for instance. Or, if you like, that's "1 RI" (Rhode Island).

This might sound familiar, because last year I wrote about a situation where 10% does not equal 10%. In all of these scenarios, you annoy the same percentage of your user population. The problem is that people aren't just some random interchangeable parts. People have connections to other people. Once you have enough people in one group, you have a relatively short number of "hops" to most of the world.

Look at it this way. If you annoy 10,000 people, what are the odds that one of them is going to be a super influencer who can turn others away from your product? If none of them are, how many of them are just one "hop" (like a friendship) away from someone who is? Just how big does that group have to get before it includes someone who is rather well-connected and will roast you over the coals for screwing things up for them?

How about if the annoyance factor doesn't have a linear relationship to the number of users? That is, as the system grows, the percentage of people who will be annoyed by your change actually drops. In that scenario, you should be afraid to use any service that's really getting massive, because it becomes really easy for them to justify whatever they want. They can just say that some tiny percentage of people actually use it, and even though that may be a bunch of individuals, they won't see that or care.

This is why justifications of "it's just n%" don't always sit well with me, even if n is a relatively small number. That could be a handful of people, or it could be massive. If they think that 10 people and 100000 people are the same just because they are both 1% of their respective groups, how can you be sure they are even human? Who thinks of people as strictly numbers or rows in a database? What company has that many users and is also that cold and calculating?

Actually, I think we know the answer to that one.