Can we disable the ping that does this?
We’re getting a bit off topic here. Would you mind starting another thread about WordPress core privacy issues?
Oh man, this thread is going into déjà vue territory! Believe me, the whole privacy issue was dealt with in deep conversation a few years ago where the end result was basically just an educational lesson on how it all works. It’s in the tavern forum archive, I’ll have to dig some posts up.
I was very excited about this change! One small thing to note, the version stats only seems to work to one decimal place as in 0.x or 1.x.
Where will this becomes somewhat of an issue/annoyance? It depends on what versioning convention an author uses. Mainly if an author wants to see how quickly his security release is being implemented.
I’ll not get into any privacy discussions!
does anyone know how frequently the Active Installs figure is calculated and how reliable is it?
When it was first introduced, our plugin had +3000 active installs. After few weeks it went up to +4000 and ever since it stayed at +4000.
I kept an eye on it recently and I don’t think it is consitent with the number of downloads and the % of each version number.
For example, we released version 1.5.0 on august the 4th.
After 10 days and approx 2500 downloads, the % of active install using 1.5.0 is around 22.4%. That should be approx 11k active installs.
I’m sure not all the 2500 downloads kept the plugin active after testing it, but I’m also sure most of them were updates and defintely not all new users. In any case that number doesn’t make much sense to me.
Is anyone sharing my same doubts?
I’m not following this. How are you arriving at 11k? What does the download count have to do with it?
Which plugin are you talking about, by the way?
The number is calculated once a day. It is simply the sum of the number of active installs from the plugin-update requests that each WP install makes.
In that sense, the number is accurate. However, the number varies daily because not all WP installs are active every day. Some therefore might not check for updates. This is in addition to the normal variation caused by people activating or deactivating or installing or uninstalling plugins. To avoid confusion, the number is rounded to the first significant digit. So, a number of 4000+ = between 4000 and 5000.
Download counts are not an accurate representation of real-world usage. Remember that inactive plugins receive update notifications as well, and we’re only counting active installs. Additionally, download counts are raw data. If I download a plugin manually 10 times, the count increases by 10. Thus they are subject to manipulation (and yes, people have tried to do that before… we shut them down when we detect it).
The active install count given is accurate to the best knowledge we have.
The plugin @paolo is referring to is “geodirectory”. I took the liberty of taking a look at the raw numbers we have and checking the active install count history for this plugin. The numbers are consistent for each day, and slowly increasing. The plugin broke 4000 installs on May 27th, and given the current trend I’d estimate it to break 5000 by the end of the month. The rise is fairly steady. So, good job there then.
But don’t make the mistake of looking at download counts. They’re entirely misleading.
Additional: BTW, the version bar graph has a two day delay on it. Be careful in how you interpret it.
Additional-additional: Sometimes, you’ll still see download counts in a profile page or the search results. This is especially true for new plugins. If we show download count instead of active install count, then that’s because the active install count is zero. This usually happens with a new plugin, before any installs have check for an update for it yet. There is also a day delay on the active install count data, because think about it: once a day, we sum up the results from the last 24 hours. So what you’re seeing is a snapshot back in time a day. Something to consider as well.
I go through the admin section to get plugins but in the dark ages, I would go to wordpress.org, download the plugin ONCE, unzip, then FTP upload to wp-content/plugins/ to every site. So one download, 5 FTP uploads.
The vast majority of installations happen through the wp-admin. We keep user-agent records of plugin downloads for 2 days (they get purged after that). These are kept essentially as a log so that we can see if somebody is faking downloads or just as a backup for recounting if the other systems break down. Note that this includes plugin updates as well as just first time downloads.
Note that WordPress’s HTTP methods send “WordPress” as the beginning of their user-agent. So, I can actually calculate that for any given day, as long as it was in the last two days.
Based on a quick search through that data set, over 90% of plugin downloads come from WordPress installs. Not through normal browsers using the Download button.
@otto when you said “over 90% of plugin downloads come from WordPress installs”, did you mean new installs? Or does it include updates? Also, is this stats are for all plugins? Or only for .org plugins?
I have a hypothesis that over 70% of the overall plugin downloads are .org plugins, I’m really curious if that’s the case?
Only w.org plugins. We do not save information for plugins we do not recognize from our systems.
Updates are included in the downloads.
having 2500 downloads for v1.5 and that version representing 22.4% of your active installs, it could mean 2 things:
11k active install or only about half of the 2500 downloads are actual active installs and the +4k is correct.
This is what I meant… ( I’m sorry for my bad english )
As Otto explained we are in the second case, which is good to know.
Hi @kmvij, it might help if you shared which plugin you were talking about?
It definitely sounds unusual considering how the vast majority of plugins are installed via the admin installer, but I suppose it’s not impossible if people use alternative methods of installation to deploy the plugin on multiple sites.
@kmvij People downloaded it from somewhere else or it uses the same name as some other plugin which people dowmloaded from somewhere else, possibly before the plugin existed on .org.
Thanks @leland and @otto. This might be the case that people downloaded before the plugin existed on .org. Actually I was going through wordpress most downloaded plugins and i came across clef-two factor plugin. Its showing 500,000+ active installs out of 486,175 total downloads.
Clef had a plugin for their users long before they put it on w.org. This isn’t uncommon.
Rather than start a new topic, I’ve chosen to add this to this (old) conversation since a number of questions did arise here about accuracy of the figures. I have a plugin in the repository that has shown 10+ active installs for several months. Kanzu Support Desk stats
That wouldn’t be an issue if other data didn’t say something different.
- The plugin has an opt-in setting for collection of usage statistics using Google Analytics. It is turned off by default as per plugin guidelines. The stats from users who’ve opted in alone show bigger numbers
- The Wp.org stats show that all users are using version 1.6.x (released May 27). You do point out that version stats are at least 2 days old. However, we released a major version 2.0 about a month ago and going by GA’s usage statistics, we do have a number of users on this version.
Statistics aside, I’ve provided support for a number of users and all their installations have been using version 2.x.x
Could something be wrong?
No, you don’t. You have a plugin in the repository that has shown 10+ active installs since the 6th of September. Before that it showed 20+ active installs. That’s down from the 30+ number that was shown on the 29th of August.
We do have historical data. The numbers for your plugin are consistent, and show a relatively minor decline over the last month.
This is exactly why we don’t show raw numbers. They’re too prone to misinterpretation. If you have the plugin actively running on a thousand sites, and nobody ever visits those sites on one day, then those sites never get triggered into checking for updates, and the number we see is not 1000, but zero.
Active sites, actively running the plugin, and actively doing things like checking for updates. That’s the number you will see.