What to measure in marketing analytics, and how to know what to track
Most marketing teams I have seen do not have a data problem. They have a decision problem. There is a dashboard, there are plenty of numbers, there may even be a weekly report, and yet when someone asks what we should do next, the room goes quiet.
That gap usually opens at the very start, when the question is what should we track. Tools like GA4, Power BI and a social media scheduler will happily collect hundreds of metrics for you. The skill is deciding which five or six deserve a place on the page, and being able to explain why each one is there. This is how I approach it.
Start with the decision, not the dashboard
Every metric you track should exist to inform a decision. Before adding anything to a report, I ask one question: if this number goes up or down, what will we do differently? If the honest answer is nothing, it does not belong in the report. It might be interesting, but interesting is not the same as useful.
This is the cleanest filter I know for cutting a bloated dashboard. It also changes the conversation with stakeholders. Instead of asking what do you want to see, you ask what do you need to decide, and the metrics follow from the answer.
Work backwards from the goal
A reliable way to choose metrics is to move through four steps in order: the goal, the question, the metric, the decision. The goal comes from the business, not from the marketing team. The question turns that goal into something you can check. The metric is the number that answers the question. The decision is what you will do with the answer.
Take a newsletter as an example. The goal is to grow readership among the people who matter most to the organisation. The question is whether the right people are actually reading it. The metric is click-through rate by audience segment, not the headline open rate for the whole list. The decision is what to change next: the subject line, the layout, the length or the call to action. Notice that follower counts, total sends and a dozen other available numbers never needed to appear.
Skipping straight to the metric is the mistake that fills dashboards with noise. If you cannot write the goal and the question in one sentence each, you are not ready to pick the metric.
Reach, engagement and outcome
Almost every useful metric falls into one of three groups, and it helps to know which group you are looking at.
Reach tells you whether people saw the work: impressions, unique reach, email deliveries, sessions. It is the cheapest number to collect and the easiest to inflate.
Engagement tells you whether people did something about it: clicks, comments, saves, shares, replies, time spent, engaged sessions. This is where you learn whether the message and the format worked.
Outcome tells you whether it mattered to the organisation: sign-ups, enquiries, applications, sales, retention. This is the group stakeholders care about most, and the one marketing teams most often fail to connect back to their activity.
A healthy report has at least one metric from each group and shows how they relate. Reach without outcome means busy but ineffective. Outcome without reach usually means you are leaving growth on the table.
Vanity metrics, and when they are not
Follower counts get called vanity metrics, and often that is fair. A large audience that never clicks, replies or buys is not an asset. But the label is not a rule. Follower growth is perfectly useful when it is tracked next to engagement rate and when the audience you are building is the one you actually want. The test is whether it connects to a decision, whatever it is called.
What to track on each channel
Here is the short list I start with, then trim. Treat them as starting points.
Social media: engagement rate per post, reach versus impressions, saves and shares, follower growth in the right audience, and which formats and topics repeatedly outperform the rest.
Email and newsletters: delivery rate, click-through rate, unsubscribe rate and replies. Treat open rate with caution. Privacy features in some email apps open messages automatically, which inflates it, so clicks and replies are a more honest signal.
Website (GA4): engaged sessions, traffic by source, the performance of key landing pages, and conversion events that you define yourself, such as a form submission, a download or a click on a contact link. GA4 will not know what a conversion means for you until you tell it.
Paid campaigns: click-through rate, cost per result, conversion rate and return on spend, always compared against a baseline.
Brand: harder to measure directly, so use proxies: branded search volume, direct traffic, mentions, and what audiences say when you ask them.
Five questions to decide whether to track something
When someone suggests a new metric, run it through these five questions.
1. Does it connect to a goal? If you cannot name the goal it supports, leave it out.
2. Can we act on it? A number that cannot change a decision is trivia.
3. Can we trust it? Check how it is collected. Inconsistent tags and campaign naming quietly break more reports than bad analysis ever does.
4. Can we compare it? A number on its own says very little. You need a baseline, a previous period or a benchmark.
5. Will anyone read it? If nobody will look at it, the cost of tracking it is higher than its value.
If a metric clears all five, track it. Keeping the final list to somewhere between five and seven is a good discipline. When a new metric earns a place, something else should leave.
Numbers need context
In one project I analysed a campaign dataset of 200,000 rows and found an overall click-through rate of 9.98%, against a typical benchmark of 1 to 5%. On its own, that sounds like a victory. But the number only became useful when I broke it down: which campaigns drove it, which channels converted, and which cities generated the most impressions. The headline figure raised the question, and the breakdown answered it.
That pattern holds almost everywhere. A single number starts a conversation. Segmenting it, comparing it with a baseline and looking at it over time is what turns it into a recommendation.
Report it so people act on it
A report that is accurate but ignored has failed. When I built a social media tracker at First Icon Group, the aim was to get the same numbers in front of the managing director and the staff running each brand, in a form both could read quickly. The tracker covered 23 staff across five platforms and cut the time we spent on stakeholder reporting by 30%, because nobody had to rebuild the picture each time someone asked how we were doing.
Whatever tool you use, keep three things on the page: the number, what changed and why you think it changed, and what you recommend doing next.
Common mistakes
The usual mistakes: tracking everything because the tool allows it, mixing reach and outcome metrics in one headline number, changing definitions between reports (which makes trends meaningless), reporting without a recommendation, and treating tracking setup as an afterthought, so that conversion events and campaign tags are only fixed after the data is already wrong.
A simple measurement plan to start with
If you take one thing from this article, make it a one-page measurement plan. For each goal, write down the question, the metric, where the data comes from, who owns it, the baseline and how often it will be reviewed. It takes an hour to write and saves months of confusion. Then build the dashboard from the plan, not the other way round.
The point of measuring is to know, with some confidence, what to do next. Start with the decision and the right numbers tend to follow.