These posts have each focussed on one specific feature or report or set of fields in GA4. However, when performing an analysis you probably wouldn’t be using these in isolation, you would probably need to consult a range of different reports and answer a few questions.

Working out how to do such an analysis is something that takes a while to get good at, it requires a bit of experience and there’s probably no way to convey those details step by step.

But we still wanted to leave you with some sort of framework so here’s a very very basic one. It’s similar to what GA4’s automatic anomaly detection does but the more detail about this you know the better.

NOTE: Each step discusses slightly different hypothetical scenarios to better illustrate that step — it won’t be a consistent scenario because it’s not possible to have a scenario where you would find insights at each of the steps below.

Step 1: Someone notices an anomaly

This might be an automated process, it might be a dashboard, it might be you, it might be someone in senior management. A problem or question has presented itself. Let’s say for this example that conversion rate has dropped.

Note that the anomaly doesn’t necessarily need to even be a change. If your dashboard shows that mobile traffic is converting at 1/3 of the desktop conversion rate, you’re free to treat it as a problem worth investigating as much as you would if a metric changed.

Step 2: Verify the anomaly on a time scale

Let’s say your dashboard is monthly and compares data to the previous month. Your conversion rate might be about 5% each month, but imagine last month it doubled to 10%. In that case, this month’s dashboard will show a significant drop in conversion rate – 50%! But it’s a phantom effect. The actual anomaly is why did the conversion rate rise so much last month?

It’s always a good idea to get more context by extending the time scale. If the change is monthly, do a quick chart of the last 6-12 months. Do a check for same time next year in case of seasonality.

Step 3: Break the anomaly down into its constituent parts

In our example, we’re trying to investigate a conversion rate drop. But “conversion rate” is a calculated metric, it’s just a shortcut for “number of conversions / number of sessions” (or number of users depending on how you measure it). So next, you would want to find out which of these is actually responsible for the change, because depending on the answer, the cause might be something different.

• Did we get the same traffic but fewer conversions? Then maybe something about the conversion process is broken.
• Did we get more traffic but the same conversions? (Or even more traffic and more conversions but the traffic rose faster?) Then it might just be an artifact of the website getting extra, less-engaged traffic.

Step 4: Apply dimension breakdowns to pinpoint the source of the anomaly

These are the ones usually worth focussing in marketing. UX and product would be different:

• Can the anomaly be attributed to a specific traffic source or campaign?
• Or a specific landing page?
• A lot of the time, the above would cover it, but you may also need to try looking at it by geographic region, device type and browser before you try something else.

Step 5: Break down the question into more specific sub-questions and repeat

Let’s imagine that the conversion rate drop was because there was more traffic than normal which didn’t convert well (from step 3). And let’s say from step 4 that you’ve worked out that it’s a new traffic source. A followup question might be “why is this new traffic source not converting?”

Why questions may require some qualitative research including user testing, surveys, watching session replays etc. But sometimes you can still stay in GA4 and get some idea of the answer, if you can break down the question into more specific ones and follow the framework again. As an example, here are some possible sub-questions (some of which would require you to go beyond GA4):

• Do we know why this source is giving us traffic? Is there a specific page/section of their website we can pinpoint this to?
• Is this traffic not filling out the form because people are not even getting to the contact page on our website, or are they getting there but not filling the form out?
• Is there something about this traffic that has a different profile to your “standard” website visitors? (eg. location, demographics)

Once you have the hang of these basics, you can extend this framework to be more specific to the question you’re trying to answer.

Good luck!