Tuesday, October 13, 2009

Those wonderful statistics are getting better all the time

The strategy people made it quite plain to me that numbers provide great arguments when you want to see things change. The WMF is eating its own dog food because they have in Erik Zachte someone who provides us with the most wonderful numbers.

Erik recently introduced the "score cards", they are the numbers the WMF uses to formulate its strategy. These numbers are a condensed presentation of many charts. The traffic statistics are my favourite at the moment.



These statistics have seen several incremental changes that made it much more useful. It started with me finding several small things; the name of the Afrikaans language shown in Zulu, several languages missing and some sorting order problems. These were quickly dealt with and as a consequence these traffic numbers became meaningful to me.

The latest innovation I noticed is that a projection is given of what the likely numbers are going to be for the next month. When you follow the stats, you find that the order of the languages changes. It becomes clear that the gaps between certain languages is quite big; it will take Chinese quite some time before it overtakes Dutch as the 10th language on the list.

When you are interested in being part of the strategy planning, spending time getting to grips with the statistics will help you to be effective. Realising our potential is the name of the game. Understanding where a  potential exists is helped by understanding our statistics.
Thanks,
      GerardM

1 comment:

Erik Zachte said...

Thank you Gerard for your appreciative words.

One small thing, the row of projection figures is not a new feature. It is only shown when the month has progressed far enough to make predictions dependable.

It is actually a very old feature that was introduced for other xml dump based tables, but was no longer relevant there as wikistats now only processes complete months from the dump, or better said after a full month has been added.

To explain: if I would process a dump that completed on the 17th which had run for two days, some article would only have history till the 15th. That's why only full months are processed and prediction is no longer needed there