Your first data hire: analyst, data engineer or analytics engineer?

The leadership team asks how much margin a sales channel really makes. Finance has one figure, marketing has another, and sales has a third sitting in its own spreadsheet. Someone spends two days stitching exports together by hand, and at the end of it nobody quite trusts the answer. That is usually the moment someone says: “we need a data person”.

Fair enough. The trouble is that “a data person” can mean three very different profiles, and whoever joins first will be working on their own. There won’t be a team around them to cover the gaps, so getting the profile right matters far more than it will for your fifth data hire.

The three profiles people usually consider

A data analyst turns data into answers. They write queries, build dashboards, spot trends and, above all, explain to business people what is going on. To do that well, they need the data to be reasonably easy to get at.

A data engineer builds the plumbing. They pull data out of each tool, load it into a central warehouse, automate those pipelines and make sure it all runs every day without surprises.

An analytics engineer sits somewhere in between. They take the centralised data, model it so it is easy to work with and set shared definitions, so that “active customer” means the same thing in every report.

None of them is better than the others. The real question is which one tackles the problem that hurts most right now.

Find the bottleneck first

When we work with a company in this position, the first thing we do is pin down where things are stuck. It tends to fall into one of these scenarios:

• The data exists and you can get at it (your tools already produce reports, or you have a fairly tidy database), but nobody has time to ask it the right questions. A self-sufficient data analyst fits well here.
• The data is spread across the CRM, the ecommerce platform, the marketing tools and the ERP, and there is no way to join it up. You most likely need to start with someone from the engineering side.
• You already have a warehouse, but every report gives a different number. What’s missing is order and shared definitions, which is analytics engineering territory.

That said, for a lot of mid-sized companies the hire that works best first is a hybrid. Someone with a solid analytical grounding who can also set up simple infrastructure with current tools. They won’t design you a complex architecture, but you don’t need one yet.

Before you open the search

A good data candidate loses interest quickly when they sense the company isn’t sure what it wants them for. We’ve seen candidates ask, in the interview, something as simple as “what would you like to know that you can’t answer today?” and get no clear reply. It shows, and the strong ones end up choosing a different process.

It’s worth arriving with three or four concrete business questions you can’t answer right now. The real profitability of a channel, say, or the step in the funnel where customers drop off. That alone lets a candidate picture the job.

A rough map helps too: which tools you use, where the data lives and who grants access to what. So does having someone in leadership who backs the role, helps set priorities and opens doors in other departments. Without that support, your new hire spends their days chasing permissions.

Set aside some budget for tooling. There are affordable options, but they’ll need some freedom to choose their stack. And be realistic about timing. The first months go on understanding, cleaning and connecting. The impressive analysis comes later.

Interviewing beyond SQL

Because this person is the whole data function on day one, how they work without structure matters as much as their technical skills. Ask them about a project they had to build from nothing, with no team or processes in place, and pay attention to how they decided where to start.

Ask what stack they’d suggest for a company your size, and why. If they describe exactly what they used at a large corporate, without adapting any of it, be a little wary. A short SQL test or an exercise on a small dataset tells you a lot, and it doesn’t need to take hours.

Then ask them to walk you through one of their analyses as if you were on the board. You’ll see straight away whether they can turn numbers into decisions. And ask what they do when the data doesn’t add up, and how they stop it happening again.

For us, the best sign is a candidate who asks about your business. Someone who is only curious about the technology may be very good, but they’ll struggle to make a difference early on.

Where it tends to go wrong

The mistake we see most is hiring a highly specialised profile first. A data scientist focused on predictive models, with no clean data and no infrastructure, will spend months doing work you didn’t hire them for. And they usually leave.

Another classic is expecting them to be an expert at everything: infrastructure, modelling, analysis, visualisation and, while they’re at it, training the rest of the team. Decide what comes first and what can wait.

Treating data as an IT project doesn’t help either. The value is in the decisions people make with it, so the role has to sit close to the business. And if every tool access request takes weeks, don’t expect much progress.

How you’ll know it worked

Don’t count dashboards. A dashboard nobody opens is worth nothing. Six months in, dig out those three or four questions you wrote down before hiring and see how many now have an answer you trust. That’s the measure that counts.

If you’re still unsure which of the three profiles fits your situation, we can help you work that out at Mimmic Talent before the search even starts. Get in touch and we’ll look at your case.