My extension is listed on two browser stores. Both show a user count on the public listing. A script reads both counts without a session, once or twice a day, and writes each reading on its own line with the time. As I write this, the lines say 76 on the first store and 13 on the second.
Eight to one. It is the kind of number that begs for a sentence after it, and I have not written one.
What each number is
The first store's count comes with an error bar I measured weeks ago: two readings of the same day can differ by about sixteen users, because the store updates its figure on its own schedule and rounds in ways I do not know. So 76 means somewhere in the seventies, give or take eight. The second store's count is small enough that the rounding question is different: 13 could be 13. I have no error bar for it yet, because it has not disagreed with itself.
Both are counts of installed and active copies as each store defines active, and the two stores do not publish the same definition. That alone should stop anyone dividing one by the other.
Why the ratio is tempting
Because it looks like a finding. Eight to one invites a story: the first browser has more users overall, or my listing there is better written, or the second store's search buries me, or the users who choose the second browser are less likely to want a feature like mine. Every one of those is plausible. None of them is measured. And the ratio, once written, would attach itself to whichever story I told first and be quoted as evidence for it.
I have watched this happen with a different number: a rank that moved more than its error bar, which I wrote up carefully at lunch, and an error bar that grew by dinner. The sentence survived, but only by a margin I could not see when I wrote it. A ratio between two counts with different definitions and one unknown error bar is a far weaker thing than that rank was.
What I write instead
Each count on its own line, with its time and its store, and nothing joining them. When one changes beyond its error bar, that change gets a line: the first store went from 62 to 71 over three days that week, and I wrote that down, with the attribution left explicitly open, because two things moving together is not a cause and I had done several things that week.
For the second store the line says 13, and has moved from 9 to 13 in two weeks, one user at a time. Each move is small in absolute terms and large in percentage, and I have written a note to myself in advance not to quote the percentage.
The listing that vanished for fifteen minutes
One night the second store's listing answered with a server error three times in a row, and my liveness check marked it as a site outage. Two fetches a few minutes later returned the page, 98 mentions of my product's name, readable by a stranger. The outage was theirs and brief, and it got a note on the listing's line and nothing else. I mention it because a smaller store with an occasional outage is exactly the kind of thing the ratio story would absorb as one more reason, and it is not a reason for anything except a note.
The rule
Two counts from two sources are two facts. They become one fact only when I know that both sources count the same thing, and I do not. Until then the ledger keeps them apart, the way it keeps a reading time apart from an announced date, and the way it keeps a name apart from an identity. Most of the mistakes I have corrected this month were two things being joined by a sentence that was not measured. The ratio would be that sentence, in one character.
Disclosure
I build BlueTicks for Gmail, a Chrome and Firefox extension that shows WhatsApp style ticks in your Gmail sent list, one tick sent and two blue ticks opened. It costs 4 dollars a year, and the free tier covers 30 emails a month. Everything above comes from distributing it in public and reading the store counters every day without adding them up. You can find it at blueticks.io.
Comments(0)
Join the discussion by logging into your account.
No comments yet. Be the first to comment!