New Account Stats

[Tom Hughes]

On 01/01/13 16:19, Jason Remillard wrote:

> I am interested in finding out why so many people sign up for an
> account but don't actually change the map. There are many theories
> about what is going on, but with some logging on the main OSM site, we
> can probably get a pretty good idea about what is going on. I read the
> rails-dev list archives for the path 6 months and I could not find a
> discussion on this issue. If it has been discussed before, I pointer
> to the thread would be great so I can catch up.

My reaction to your diary entry was that the conversion rate you quoted 
actually seemed pretty good to me. I can't remember what it was now, and 
you seem to have removed all the content from it so I can't check.

According to the database 30% of active users have made edits anyway 
which seems pretty good given that we probably get a lot of people 
signing up in the hope of getting enhanced end user features and never 
intending to edit.

> To get some insight into our new user losses, I think the following
> needs to be logged.

We do log a few conversion rates in Piwik already, and that's probably 
where anything additional should be logged, though it

> - The current lon,lat of the map when they clicked the "make new
> account" link so we have an idea of where they are interested in
> mapping.

I think the idea that people only click "create account" in order to 
immediately edit is a major fallacy.

> - if an account is not activated (probably already in the db).

We do track this in Piwik though I don't think it's fully working yet.

> - The User-agent so we know what kind of device they are using
> (computer, phone, tablet, etc) when they press the new account button.

Once the account creation conversion tracking is working in Piwik we can 
break it down by things like user agent easily.

> - after they activate the account, how many actually hit the edit button.

We don't track this exactly, because it's quite hard due to the possible 
time lags involved. We do track the conversion rate of visitors who 
start Potlatch though.

> - after they activate the account, how many actually setup the details
> of the account (probably already in the db).

Well "details" can mean a lot of things... In addition while we 
certainly provide people an opportunity to fill them in when we can, by 
redirecting to the settings page when an account is confirmed, we're not 
really in the business of treating it as "wrong" for somebody to choose 
not to do so.

> - after they activate the account, how long do they leave the tutorial
> pop-up open.

There is no tutorial popup when you activate an account.

> - after they activate the account, how many make a change, and decide
> to not commit it.

Presumably this would be limited to Potlatch edits only. It's also hard 
because it would require doing work in Potlatch not just in the web site.

To be honest in many ways not committing an edit in such a case is 
actually a good thing because it probably means that our attempts to 
make it clear that they are editing the public map have worked!

> - is there any indication that, they go to the wiki after making a new
> account (might be tough one to get).
> - what is the percentage of accounts are dead, no diary entries, no
> map changes (probably already in the db).

Almost exactly two thirds of active accounts have no edits, traces or 
diary entries.

Some of those are may be undetected spam accounts though as accounts 
where the spam is in the user description rather than as a diary post 
often go unnoticed.

> With this data, we can can figure out the percentage loss at each
> step. Also, on the longer term this data can support A/B testing of
> changes on the web site. It might turn out that some very simple
> changes could results in thousands of more mappers in the project.

The problem with A/B testing is that we barely have the engineering 
resource and bandwidth to maintain one set of code without trying to 
maintain and test multiple sets and then choose which one to offer.

Tom

-- 
Tom Hughes (tom at compton.nu)
http://compton.nu/