Separating internal and external audience data is essential to our understanding of usage in our web estate.

Deprecating network values in analytics

On February 4 2020, Google deprecated the ‘service provider‘ and ‘network domain‘ dimensions within Google Analytics.

Seer Interactive blog on deprecation of network domain and service provider

A lot has happened in the world since then, so I thought it would be useful to analyse that change and how we now advise to visualise this data.

We formerly created segments with network information to distinguish internal University traffic and isolate it in our reporting. We knew which network providers provided the University estate, so could easily identify – then optimise – for appropriate audiences.

Why remove network information from reporting?

In retrospect, it was an early step towards greater user privacy within analytics: why should I, as an analyst, know which company you work for, or which company provides your internet connection?

You might also argue that the physical address of visitors isn’t as definitive as it once was: in the age of the VPN and remote working, identifying ‘University people’ isn’t as easy to define as it once was.

Maintaining internal/external data reporting

Given our high level of usage, we had two possible solutions to this change.

The first option, which is the way we advise you now conduct such reporting, was to create 2 new ‘views’: one limited to internal IP addresses only, the other with internal IPs excluded (we also retained our combined view). You can easily change between views in the top left of analytics, and choose either the “External IPs only” or “Internal IPs 2020” view as desired.

Any segments created in the main view can be used on either of these views, so the flexibility of reporting is maintained.

If you need to perform reporting like this, contact our support and let them know you need to gain access to these new views.

IS Helpline contact form

Alternative solution

The second choice we had was to use behavioural indicators (such as viewing internally-oriented pages, or logins to our portal, MyEd) as a ‘proxy’ to being based in the University. While not a terrible option, it can be difficult to assess the accuracy of this approach.

Separating smaller, specific groups of users

If you are seeking to separate out specific users from your reporting, create a page address with additional parameters (like www.ed.ac.uk/?Exclude_Duncan), then get the people you want to separate out to visit it. You can then exclude this data from your reporting using a segment excluding users who visited that specific address.

Annotations on our account

As with any such change, I annotated our analytics account so that if our users noticed a sudden large change in their reporting, they could understand why that happened.

Loves Data blog on how and when to use annotations

Future changes to analytics guidance

We’ve been struggling to resource updates to our full analytics guidance document.

If you want to shape the future of analytics provision and support at the University, please join as at the next Web Publishers’ Community session.

Web Publishers’ Community session, July 2021 (University login required)