How to become a data analyst

"Data analyst" is not one job

The first thing that goes wrong is the search term. People decide to "become a data analyst", type that into a job board, and then try to reverse-engineer a curriculum from whatever comes back. What comes back is a blend of several different jobs that happen to share a keyword, with wildly different entry conditions.

We group the roles we actually teach toward into five families. Each one carries a mandatory counter-case, because a family presented without its caveat is a sales pitch. Here is the one closest to what most people mean by the term:

That fragmentation is not a technicality. It is the reason the published demand and wage data you find will not agree with the job titles you actually see posted, and the reason a course that promises you "the data analyst career path" is flattening something that is not flat. You can see all five families, the roles inside them, and which week of our syllabus backs each taught skill on the roles and market data page.

The title most people aim at is the wrong one

When someone tries to put a number on "data analyst" demand in Canada, they usually reach for one occupation code. Here is what that code actually says about its own entry conditions:

We publish that on our own market page, and we are repeating it here rather than quietly leaving it out, because it is the single most important thing a career-changer should know before choosing a target. A short program does not turn into a bachelor's degree, and no amount of enthusiasm changes what an employer has written in a job requisition.

The honest version of "how to become a data analyst" is therefore: aim at the work, not the word. Get good at the thing every one of those families actually does — turn an ambiguous question into a query, a model, and a recommendation someone can act on — and stay flexible about which title finally lets you in.

The job boards are overstating the market

Before you calibrate your plan against a job-board result count, you should know how those counts are built. We checked, wrote down what we saw, and dated it:

We are not telling you this to be discouraging. We are telling you because the inflated number is the one you will find on your own, and calibrating your expectations against it is how people end up concluding they are uniquely unemployable when the real problem is that the denominator was fiction.

What is actually load-bearing

Strip away the tool lists and the same small set of capabilities does the work across all five families: getting data out of a relational database, reshaping it in code, versioning what you build, putting it in front of a non-technical reader, and being able to say what you are unsure about.

Grade your own skills honestly

The useful mental model is to sort every skill on a target job posting into four buckets, and to be strict about the top one:

• Verified — you have server-checked, reproducible evidence you can do it. Not "I did a tutorial". Something a stranger could run.
• In track — it is genuinely adjacent to something you have built, and you could get there quickly.
• Declared — you have said you can do it. That is all that means.
• Gap — you cannot do it yet, and pretending otherwise fails at the technical screen.

Proof versus paper

Here is our thesis, stated as a thesis: we believe a small number of real, defensible, independently checkable projects does more for a career-changer than another certificate, because a certificate proves attendance and an artifact proves capability.

We built a company around that belief. And we are not going to tell you it is an established fact, because when we went looking for evidence, this is what we found:

A plan you can actually run

The failure mode for self-teaching is breadth: a long trail of half-finished tutorials, none of which you can defend under questioning. The alternative is fewer, deeper artifacts built on real data with real mess in it.

That last step is the one people skip, and it is the one that decides interviews. We wrote a separate guide on how data-analyst interviews have changed, including the rubric we grade against.

Where this advice stops

We would rather name the edges of what we know than let you assume we know more than we do. Two more things we looked for and could not establish:

The full list of what we checked, what we found, and what we failed to verify is published on the roles and market data page — including the wage bands, with the caveats their publishers attach to them.