Analytics hygiene: What to measure in GA4 (and what to ignore)

But the role of think tanks, policy and research organisations was never about chasing clicks. As one communicator at one of our recent WonkComms events put it: “We’re here to inform, not just increase PDF downloads.” The real goal is reaching the right audiences and shaping debates — and that calls for smarter, more goal-aligned measurement.

This guide breaks down what matters most in GA4 for the policy and research sector — and what to ignore.

What’s changed (and why your numbers look weird)

Teams across research, think tanks and policy publishing are seeing similar patterns in their web analytics:

• Different metrics, eroded trust: GA4 defines engagement differently and filters more bot traffic, meaning your dashboards might show a 5–30% drop. Year-on-year comparisons to Universal Analytics don’t hold up, and confidence is slipping.

• Mysterious “Direct” traffic: GA4 often mislabels bot and privacy-blocked visits as Direct. Big spikes with near-zero engagement or traffic from data centre hubs (like Ashburn, VA) are red flags.

• AI is taking your clicks: Generative tools like ChatGPT and Google’s AI Overviews serve answers directly, reducing organic traffic even when your content ranks. This doesn’t mean your work isn’t useful — just that users are consuming it differently.

• Engagement is harder to prove: The long tail — like the policymaker who downloads your report three months after release — is still valuable but harder to surface. GA4 can help if you set it up intentionally.

The takeaway: it’s time to contextualise numbers, focus on quality and stop expecting your dashboard to tell the whole story on its own.

What to track (and how) in GA4: A smarter approach for research and policy comms

You don’t need to track everything. But in a sector where communications success is measured in influence, not impressions, you do need to track the right things — and set up GA4 in a way that reflects what success looks like for your organisation.

This section lays out the metrics we’ve found most meaningful across research and policy clients — along with the practical setup steps that ensure your dashboards capture real engagement, not just noise.

1. Focus on real engagement (not just reach)

Engaged sessions and engagement rate are more relevant than total visits. GA4 counts a session as “engaged” if it lasts more than 10 seconds, has 2+ pageviews, or triggers a conversion (like a report download). This is a much better proxy for serious readers — like the policy analyst who’s digging into your evidence — than a bounce rate ever was.

Average engagement time shows how long your content was actively being read — not just how long the tab was open. This gives you a more accurate view of attention, especially for dense explainers or long-form analysis.

Scroll depth is particularly useful for think tanks publishing detailed reports or data-rich blogs. Tracking whether users reach 50%, 75% or 90% of a page can give you a clear sign of true reading. For example, 75% scroll depth on a climate finance briefing or consultation response is a strong signal of policy-relevant engagement.

To do: Set up scroll tracking via GA4’s Enhanced Measurement or Tag Manager, and combine it with engaged sessions to build a fuller picture of how people interact with your research.

2. Track the actions that show real interest

Micro-conversions matter most in this sector. Think beyond visits to actions like:

• PDF downloads of reports, briefings or summaries
• Outbound clicks to a cited article or related policy document
• Newsletter sign-ups (especially important if growing a policy-engaged email list)
• Video plays for recorded events or explainers

Interactions within structured publications — e.g. using a table of contents, switching tabs, or navigating between chapters (like in this interactive report from NatCen).

To do: In GA4, define these as conversion events. This lets you shift the internal conversation from “traffic” to “impact”: who actually engaged with our content, and how?

3. Look at combinations of behaviours…

…not just individual actions. For one think tank, an “engaged user” might be someone who scrolls through 75% of a page, then downloads the full report. For another, it might be return visits plus a newsletter sign-up. Your definition should reflect your own comms strategy and theory of influence.

And, understand what your audiences actually want:

Internal site search queries offer direct signals about user needs. If policymakers or journalists are searching your site for terms like “refugee statistics 2024” or “net zero policy tracker,” that’s gold for both content strategy and IA.

To do: Enable site search tracking in GA4 and check reports regularly. Use findings to improve how you surface key research and spot content gaps.

4. Look at navigation paths (user journeys)

These help you see how users move through your site — e.g. from a blog post to a data tool, or from an event summary to a report download. For research orgs with lots of overlapping content types, this can help you optimise journeys across topic areas or policy themes.

And, attribute outreach efforts accurately:

Campaign-tagged links (UTMs) are essential for understanding what’s working. Whether it’s a new report launch via newsletter, a paid LinkedIn campaign, or a partner’s press release, UTM tags help you trace traffic back to the source.

To do: Be consistent with naming: all lowercase, standardised terms like “newsletter” or “linkedin”. Set up a shared sheet or tool to manage tags. GA4’s Acquisition reports will then let you filter by campaign and source — showing which outreach channels are driving truly engaged users.

5. Maintain clean, honest data

To do: Exclude internal traffic from staff — especially if your organisation is small or your content team views pages often. Internal sessions can distort your data significantly.

To do: Watch for bot and spam traffic. Spikes in Direct traffic from places like Ashburn, VA, or sessions with zero engagement, often point to bots. GA4’s bot filtering helps, but it’s not foolproof.

To do: Ensure GA4 is implemented everywhere. This includes microsites, subdomains, and asset-heavy pages like publication libraries or download pages. Gaps in tracking can make your performance look worse than it is.

And what to ignore (or treat with caution)

• Bounce rate: Now just the inverse of engagement rate. Use it as a light diagnostic, but prioritise engagement rate and scroll.
• Raw traffic totals: Cookie opt-outs, bots, and blockers distort totals. Use traffic as a backdrop, but focus on quality interactions.
• Demographics and interests: GA4 only tracks users who opt into ad personalisation — likely a small, skewed subset. Don’t overinterpret.
• Session duration: GA4 now tracks only active time. Lower numbers don’t mean lower interest — just better accuracy.

• Cross-domain tracking: Without a technical setup, GA4 won’t join journeys across multiple domains. Don’t sweat it unless it’s a high priority.

Final thought

In a privacy-first, AI-driven web, success isn’t just about traffic — it’s about clarity of purpose. Use GA4 to measure progress toward influence, not impressions. Focus on engaged users and meaningful actions tied to your mission.

Want help tuning your GA4 setup to reflect your organisation’s goals?

Soapbox supports think tanks, research and policy organisations to:

• Audit and configure GA4 for clarity
• Define metrics that match your theory of influence
• Build intuitive dashboards
• Identify tools to fill GA4’s gaps

Get in touch for a quick health check or full strategy support — and turn analytics into a tool that actually helps you communicate impact.