Advice about using Visitor or Unique Visitors

[Sandeep]

So quickly I just wanted some info about which metric is more
accurate. I am interested in Unique Visitors on a monthly and weekly
basis (consider them separately). My company has been using Unique
Visitors for a long time, but when I use the Visitors, which I heard
is supposed to be the same, I get a number that is anywhere from 5 to
20% different from Unique Visitors.

So I just wanted to know which one is more accurate? And also which
one is more consistent.

Keep in mind that I am just using the Google Analytics UI, and not
using a query to fetch the data from the API.

I would like to use a query eventually (because then I can automate
the fetching), but I cannot do that if we have to stick with Unique
Visitors.

Please let me know any insight you have into this. Also if you know a
good way to automate fetching - I started using ShuffePoint and it has
been great so far to automatically fetch data every 'x' minutes within
Excel.

Thanks!

[Sandeep]

Just to be clear, when I say 'using Unique Visitors' and 'using
Visitors', I am referring to the metrics within the GA UI. Thanks

[Tim Uckun]

I have noticed that on a daily basis the numbers are the same as the
GA UI but when you add them up they are wrong. Obviously Google is
doing something other than simple addition when giving you totals.

[Nick]

Hi.

So the objective of these 2 metrics is to give you the absolute unique visitors per the date range for the selected dimensions. These values are not additive.

Today we have 2 calculations of this metric. You can explicitly see them in the custom reports selection for v5. In the API we expose ga:visitors which automatically maps to either calculation depending on the query.

For complex, ad-hoc, queries, GA goes through each hit in all sessions within the date range and counts the unique occurrences of the visitor's id across the selected dimensions. While accurate, this uses memory and tends to be slow. But we can give you unique visitors across any combos of dimensions, ad-hoc, which is nice. This value was only recently added to the product.

For any simple query which only uses the date dimensions, the method above is slow. In this case each session has the time stamp of the first hit in the session as well as the timestamp of the first hit in the previous session. So as GA goes through each session, we can increment the visitor count if the previous timestamp is outside of the start of the date range. This is fast and doesn't require lots of extra memory. But it also relies on the browsers timestamp....which can be off.

All GA's date ranges are based on the time the hit was collected by GA, server time. And this second calculation relies on the browser time stamp. When these don't match, this count can be off and you see difference between metrics.

-Nick

[Sandeep]

Perfect. Thanks Nick.

So it seems then that the 'complex' method is extremely accurate? If
so, then is there a query that, in effect, makes ga:visitors map to
the complex calculation? I really want to get rid of the inaccuracies
and variance in my data. I use the isMobile dimension quite a bit
already - does it create the level of complexity needed to force the
API to do the calculations? Because if so, it is easy to add the 'yes'
and 'no' columns up to get the total number.

-Sandeep


On Aug 22, 7:19 pm, Nick <api.nickm%google....@gtempaccount.com>
wrote:

[Nick]

It's as accurate as the visitor ID which gets set via the tracking library.

If you use date dimensions you get the fast calculation. All other queries, including isMobile, will use the unique visitor ids.

One thing to consider is that isMobile is a session level dimension, meaning either the entire session is 'yes' or 'no'. But a visitor with the same visitor id, can have sessions with both 'yes' and 'no'. So again, the visitors metric is not additive.

-Nick

[chris@shufflepoint]

Hi Nick,

What may still be ambiguous is which method is used if one had both a
date (i.e. ga:month) and non-date dimensions in the query. My
interpretation of your explanation is that in this case it would use
the slower, more accurate method since the query doesn't use only date
dimensions. Is that correct?

Thanks,
Chris

On Aug 23, 5:26 pm, Nick <api.nickm%google....@gtempaccount.com>
wrote:

[Nick]

Yup,

If you use anything besides date, we have to re-process the results, so it's the slower more accurate way.

So what's interesting is every time you apply an advanced segment, GA has to recalculate the values. So you create an advanced segment which matches all visits, you effectively force re-calculating data. So you can use date and visitors and apply a segment which matches all visits, to force the slow way of re-calculating data.

If you get subtract the number of visitors between the 'slow' and 'fast' ways, you can get a sense of the number of visitor's who have a different time on the machine vs the time on our server. ie...who have an incorrect client time.

-Nick