Diagnostics
Why your RPM chart is lying to you
On a site we run that gets under a thousand page views a day, daily RPM ranged from $2.93 to $49.26 in a single month. Almost none of that was a signal.
By Mario Bailey · Published
RPM is the number every publisher watches, and on a small site it is the number most likely to fool you. Below about a thousand page views a day, a single click can move your whole day. Here is a month of that on a site we run, and the rule we use so the swings stop driving decisions.
A month that looks like a story, and isn't
Over 22 days on one site in our network, daily page RPM had a median of $11.45, a standard deviation of $8.86, and a range from $2.93 to $49.26. The high day was not a breakthrough. The low day was not a collapse. Both were the same site, the same content, the same ad setup, days apart.
| Daily page RPM, 22 days | |
|---|---|
| Median | $11.45 |
| Standard deviation | $8.86 |
| Lowest day | $2.93 |
| Highest day | $49.26 17x the low |
AdSense page RPM by day, one site in our network, under a thousand page views a day. The spread is what a handful of clicks does to a small denominator.
Why one click moves everything
Page RPM is revenue per thousand page views. On a day with only a couple hundred views, each click is worth several dollars of RPM by itself. On a 212 page view day, a single extra one dollar click moves page RPM by $4.72. That is the whole mechanism. A day with seven clicks and a day with four are not a trend. They are a coin landing differently.
We have watched a single premium click account for more than 40 percent of a week's revenue on a site this size. We once nearly shipped a "revenue down 27 percent" conclusion from a week over week read. Dropping one outlier day flipped the same comparison to up 5 percent. Neither number was real.
The rule we use
Never read RPM week over week on a site under about a thousand page views a day. Read trailing 28 day revenue, and before you believe any change, check whether one day carried more than a quarter of the window. If it did, that day is the story, not the trend. That single outlier check is one line of arithmetic, and it reverses the sign more often than you would expect.
What to watch instead
Watch the numbers that do not depend on click luck. Ad impressions per page view, and viewability, both move with how your setup is built rather than with who happened to click today. Page RPM is still worth tracking, but over a trailing 28 day window and against itself, not day to day. At real volume, RPM becomes a metric again. Below it, treat a good day and a bad day as the same day seen twice.
How we measured it
AdSense page RPM by day over a trailing window, on one site in our network that runs under a thousand page views a day. One site is an illustration, not a proof. The point is narrow and it holds: at low volume, daily and weekly RPM is dominated by click luck, so reading it as performance will have you chasing noise. The numbers are real and our own.