o
o
o
o | 
WHAT YOUR HIT COUNTER
IS REALLY COUNTING
a ninety-day study by James Beyers, actuary
REPORT | THE NINETY-DAY LOG |
| o
o
o
o |
o
o
o
o
o
o | On Saturday morning, August 28, the counter on our family's home page went from
3,999 to 4,000. My wife, Alice, looked over my shoulder and said, "Four thousand people have read about our
kitchen?"
I'm an actuary. I work for an insurance company in Des Moines, and my job is estimating how many of
something there really are, from numbers that don't quite say so. I also happened to have ninety days of
notes on that counter, because I'd been writing it down four times a day since the first of June. So I
told her I'd get back to her. This is the answer, with the arithmetic shown, because an actuary who won't
show his arithmetic shouldn't be trusted with yours. |
| o
o
o
o
o
o |
o
o
o
o | THE SHORT ANSWER
The 4,118 on the counter at the end of the day on Sunday,
August 29, meant about 1,350 different people, give or take 300. Not four thousand. A third of
that, roughly. And more than a quarter of the four thousand was me. |
| o
o
o
o |
o
o
o
o
o
o
o
o
o
o
o
o | WHAT A COUNTER COUNTS
1. Page loads, not people. A hit counter is a little
picture with a number in it. Every time a browser asks for that picture, the number goes up by one. It doesn't
know who's asking, or whether they asked five minutes ago.
2. Returning visitors: sometimes. If a visitor goes from my front page to the softball statistics and
back, the counter may go up again or it may not. It depends on whether the browser fetches the picture
fresh or shows the copy it already has. On our Netscape at home, the Back button usually didn't add one and
clicking my "Home" link usually did. On the Internet Explorer at my office, both usually did. I tested it
fifty times on each (those loads are in my own column) and got an average of 1.35 counts per visit.
3. Not robots, mostly. The search engines send programs around to read pages. Those robots read the
words, but they often skip the pictures, and my counter is a picture. So most robots don't count, which is
one thing a counter gets right.
4. Not people with pictures turned off. Some people on slow modems browse with pictures off. They
read every word and never touch the counter.
5. A counter can start at any number. Mine started at zero on June 1; I typed the zero myself. My
old counter was a BeanCounter, and if you've read Jerry Dietz's
reviews of free counters, you know what happened to
those in May. Mine went with them, and my old number with it. His reviews also mention a service that will
start your counter at any number you type in. Remember that the next time you feel bad about somebody
else's. |
| o
o
o
o
o
o
o
o
o
o
o
o |
o
o
o
o
o
o
o
o
o
o
o
o
o
o | THE WORKSHEET
All figures June 1 through August 29, 1999, from
the log.
| Counter at end of Aug. 29 (started at 0) | 4,118 |
| less my own loads: reloads while editing 742, looking and showing people 393 | -1,135 |
| less family (Alice, the kids, my mother) | -520 |
| = loads by visitors | 2,463 |
| divided by 1.35 counts per visit | = 1,824 visits |
| divided by 1.35 visits per person | = 1,351 people |
| CALL IT | 1,350 |
|
| o
o
o
o
o
o
o
o
o
o
o
o
o
o |
o
o
o
o
o
o
o
o
o
o | Where 1.35 visits per person comes from. I can identify 41 people who visited during the
ninety days: guestbook signers, people who e-mailed, and league players who mentioned it at the ballpark.
Between them I can account for 55 visits, or 1.34 each. Call it 1.35.
Why "give or take 300." Both of my 1.35s are soft. Counts per visit could be anywhere from 1.2 to
1.5, depending on what browsers my visitors use. Visits per person could be as low as 1.15 or as high as
1.6, because 41 people is not many to go on, and the people who sign guestbooks are the friendly ones, and
friendly people come back more. Put the two worst cases together and you get anything from about 1,030 to
1,790 people. Both factors landing at the same end at once is unlikely, so I've narrowed it to 1,050 to
1,650. I haven't added the people with pictures turned off. Their number is small, and it's inside the
give-or-take. |
| o
o
o
o
o
o
o
o
o
o |
o
o
o
o
o
o
o
o
o
o | WHEN THEY COME: BY WEEKDAY
Visitor loads only, average per day. The ninety days
started on a Tuesday and ended on a Sunday, so every weekday came up thirteen times except Monday, which
came up twelve. That's why these are averages and not totals.
| Monday | | 29.8 | | Tuesday | | 27.9 | | Wednesday | | 26.8 | | Thursday | | 24.8 | | Friday | | 20.5 | | Saturday | | 21.0 | | Sunday | | 40.9 |
|
| o
o
o
o
o
o
o
o
o
o |
o
o
o
o
o
o
o
o
o
o
o
o | WHEN THEY COME: BY HOUR
I read the counter four times a day: at 6:30 in the
morning before work, at noon from my desk, at 5:30 when I got home, and at 10:30 before bed. (Every reading
from my office adds one, which I subtracted. Actuaries are thorough, not fun.) That divides the day into four
stretches. Here's the average number of visitor loads per hour in each, with Sunday evenings pulled out on
their own line.
| 10:30 p.m. to 6:30 a.m. | | 0.44 | | 6:30 a.m. to noon | | 0.95 | | noon to 5:30 p.m. | | 1.24 | | 5:30 to 10:30 p.m., Mon.-Sat. | | 1.99 | | 5:30 to 10:30 p.m., Sunday | | 4.51 |
Sunday night is the busiest
time on our page by a mile. I keep the statistics for our church softball league, which plays on Sunday
afternoons, and I post them Sunday night. I can't prove that's the reason. But I've noticed the batting
averages get looked at a lot more than the kitchen. |
| o
o
o
o
o
o
o
o
o
o
o
o |
o
o
o
o
o
o
o
o | A REQUEST
Please stop comparing counters. Yours is not a score, and
neither is anyone else's. Somebody's 25,000 might have started at 20,000, and might be half reloads, and might
be a third that many people. Somebody's 300 might be three hundred real people who read every word. The number
on a counter is a measurement of page loads, taken with a cheap instrument under conditions nobody wrote down.
If you want to know who's reading, read your guestbook.
Our family page isn't linked from here, by the way. That's deliberate. I'm keeping the log going through the
end of November, and I don't want this page's readers to throw off next quarter's count. |
| o
o
o
o
o
o
o
o |
o
o
o
o
o
o | THIS PAGE, SINCE SEPTEMBER 5, 1999:
  
(of which 141 are me. I'm keeping track.)
Corrections to the arithmetic welcome: [email protected]
Last updated Sunday, September 12, 1999 |
| o
o
o
o
o
o |
|
GeoCities.us
» WallStreet » Bureau
