What does a data analyst do?
A data analyst answers questions with evidence. Which ad brought in customers who stayed? Where do people leave the checkout? Did the launch move sales or just views? The analyst gathers the data from analytics platforms, ad accounts, sales systems and surveys, cleans it until it can be trusted, runs the analysis and presents the result as a chart and a sentence a decision-maker can use. They build the dashboards a marketing team looks at every morning and the one-off studies that settle an argument.
In an agency the analyst works across clients and channels, measuring campaigns and feeding what they learn back to the creative and media teams. In-house they tend to sit closer to the business, connecting marketing numbers to revenue. Either way the tools are similar: spreadsheets, SQL, a visualization tool and whatever the ad platforms report.
Why it matters
Why a data analyst matters
Marketing generates more data than anyone can read, and most of it is noise. An analyst finds the signal and, just as important, says when there is none. Without one, teams make decisions on the metric that happens to be on the screen, usually views or clicks, and optimize toward it whether or not it matters. With one, the question becomes what actually changed and why, and the next campaign is briefed on an answer rather than a hunch.
In practice
How a data analyst works in practice
- 01
The analyst starts by agreeing the question and the metric that answers it, before any data is pulled. Then the plumbing: tracking in place, naming conventions consistent, sources reconciled so the ad platform and the analytics tool are counting the same thing. Most of the job is this cleaning, and it is why the first report on any new account takes longer than the client expects.
- 02
With clean data they analyze, which means comparing like with like, controlling for what else changed and resisting the chart that tells the story everyone wanted. The findings go into a report that leads with the answer, shows the evidence and ends with what to do differently. Then the dashboard is updated so the next question can be answered faster than the last.
At RIOT
Data analysts at RIOT
The Meta ad engine we ran for Homebase from February to April 2025 was built around reading the numbers: what had just run, what it did and what to make next. Creative made on that rhythm depends on someone reading the results properly, and the reading belongs in the same conversation as the edit, not in a report sent afterward.
Watch out
What people get wrong about data analysts
Asking them for a dashboard instead of an answer. A dashboard with forty tiles is a way of avoiding the question. The other mistake is treating the analyst as the person who proves the campaign worked. A good analyst will tell you when it did not, and a team that punishes that answer will soon have a dashboard that only ever goes up.
