Technology · Analyzing & Problem-solving
Data scientist: what the job actually is
Turning messy records into an answer somebody can act on, then explaining honestly how sure you are.
Published Last reviewed How we research and correct this
The work
What this job actually is
- Taking a pile of data that was collected for some other reason and working out whether it can answer the question somebody is asking.
- A large share of the job is cleaning: finding the duplicates, the blank cells, the column that changed meaning halfway through the year. The modelling is a small slice of the week.
- The question you are handed is rarely the question that needs answering. Part of the skill is asking what decision the answer will feed.
- You will often have to say the unwelcome thing: the data does not show what the team hoped, or it is too thin to say either way.
- It sits on top of statistics and programming. You need to be comfortable with both, and comfortable being checked.
A Tuesday
What a normal day looks like
Not a good day or a bad one. The ordinary version, which is what you would actually be living.
- 9:00 — Open the analysis from yesterday. The numbers have shifted because somebody corrected a source file overnight.
- 9:30 — Find out which rows changed and whether the shift is real or an error. Spend an hour on it.
- 10:45 — Meeting with the person who asked the question. They realise the thing they asked for is not the thing they need.
- 11:30 — Clean a new data set. Dates in three formats, names spelled five ways, a column that is mostly empty.
- 12:30 — Lunch. Someone shows an impressive chart that is almost certainly measuring the wrong thing.
- 13:15 — Build a first rough model. It works suspiciously well. You go looking for the leak.
- 14:30 — Find the leak: the model could see something that would not be known at the time of the decision. Remove it, and the result gets much less exciting.
- 15:30 — Write up what you actually know, what you do not, and how sure you are. This is the part that gets used.
- 16:30 — Review a colleague's work and ask the awkward question about their sample.
- 17:00 — Leave a note for tomorrow, because you will not remember why you made that choice.
Honestly
Who loves this, and who should not do it
The second list is the useful one. Nearly every careers page leaves it out, which is why nearly every careers page reads like an advert.
- You are happy being told your answer is wrong and going back to check.
- You like finding the pattern, and you are equally interested in whether it is real.
- You can stay with dull tidying work because you know the result depends on it.
- You enjoy explaining a result in plain language to someone who will decide on it.
- You do not need the answer to be dramatic.
- You want to build the exciting model all day. Most of the job is preparing data and explaining caveats.
- You dislike statistics or being careful about uncertainty. This work is built on both.
- You need the people you work with to be pleased. You will often deliver news that was not wanted.
- You get restless sitting alone with a screen. Much of the time is spent that way, with meetings in between.
- You need a guaranteed route in. This is a competitive, not a plentiful, corner of the labour market.
The route in
How you actually get there in Ontario
Know what the outlook says before you commit
Job Bank rates the 2025-2027 outlook for data scientists in Ontario as Limited, and says employment is expected to remain relatively stable with not many positions becoming available through retirements. That means openings are thin and competition for them is real. It is a reason to prepare well, not a reason to stop.Build the maths and statistics base
There is no licence: Job Bank says the occupation is not regulated in Canada. Job Bank lists a bachelor's degree in statistics, mathematics, computer science or a related discipline, or a college programme in computer science, with a master's or doctoral degree in machine learning, data science or a related quantitative field for some roles. Take the strongest maths you can in high school, because it keeps these doors open at 18.Get paid work with data before you graduate
Co-op and internship placements are the most direct way to show an employer that you can work with real, messy data. Ask any programme you are considering how many students get a co-op term in a data or analyst role, and ask to see where they end up.Show a finished project, not a certificate
A project where you chose a real question, found the data, cleaned it, answered it and said honestly what you could not conclude is what an employer can judge. A data set analysed carefully beats a fancy model nobody can explain.
Where this comes from: Job Bank — Requirements for data scientists in Ontario
The money
What it pays in Ontario
Job Bank's own figures, quoted rather than summarised, with the period they describe — so you can tell how old they are and go and check them.
- $31.25
- Low (hourly)
- $47.69
- Median (hourly)
- $71.79
- High (hourly)
Ontario, NOC 21211. Reference period 2023-2024, last updated by Job Bank on 19 November 2025. Wages vary a lot by region inside Ontario and by how long you have been doing it — the low figure is roughly where you start, not what the job is worth.
Where it is heading
The outlook for the next three years
Job Bank publishes a three-year outlook for every job in every province. It is the closest thing to an honest forecast, and it moves, so read it as a weather report rather than a verdict.
Limited2 out of 5 stars
- Employment is expected to remain relatively stable.
- Not many positions will become available due to retirements.
What that rating means: how many openings Job Bank expects in Ontario for this kind of work against how many people are likely to be looking — not whether it is a good job. A “limited” rating means more competition for each opening; it does not close the door, and it can change before you are ready to walk through it.
- 23,450
- people work in this job in Ontario
- 30%
- are women (48% across all jobs)
- 6%
- are self-employed (15% across all jobs)
Source: Job Bank — Data scientists job prospects in Ontario (NOC 21211). Outlook updated by Job Bank on 10 December 2025; read by us on 5 October 2026.
Before you leave school
Things you can do now
- Take the highest maths available, and statistics or data management if your school offers it. Dropping maths at 16 closes the main doors.
- Learn spreadsheets properly. Pivot tables, filters and honest charts are real data work, and they are free to practise on any data you have.
- Do one project that starts with a question you actually care about, such as a sport, a game or your school, and finish it.
- Ask whether your school or board offers computer science or a co-op placement in an office that handles data.
Test it
Do this before you decide anything
Reading about a job tells you almost nothing. This is designed to be done alone, this week, for free.
Answer one question with a week of your own data
About 90 minutes- Pick a question about your own life that data could answer: how much sleep you get against how you feel, how long your commute takes, how many minutes you spend on something.
- Record the numbers for a few days in a spreadsheet. Then break your own data on purpose by leaving a few days blank and noticing how that changes what you can say.
- Make one chart and write one sentence with your answer. Then write a second sentence saying what you cannot tell from such a small sample.
- Show it to somebody and ask whether the chart makes the point clearly or whether you have to explain it.
What it tells you. If the second sentence, the one about what you cannot conclude, was the part you cared about getting right, you think the way this job asks. If you found the sentence about the limits annoying and just wanted the answer, you may prefer a field where answers are firmer. Either result is useful.
Better than any article
Questions for somebody who does this job
Most people say yes to a teenager who asks good questions. These are the ones that get real answers rather than encouragement.
- “What proportion of your week is cleaning data compared with building models?”
- “How did you get your first data role, and what did your portfolio contain?”
- “How do you handle it when your result is not what the team wanted?”
- “What do you wish you had studied more of?”
- “How competitive was it to get started, and what made the difference?”
- “What does a bad question look like, and how do you push back on it?”
Could you own one of these?
The part other career sites never mention
Sometimes. Independent analysts and consultants exist, working for organisations too small to hire one full-time. It usually follows years of experience, because clients pay for judgment about their data, not for the ability to run software. A student should treat it as a later move, not a first one.
Not sure this is the right list to be reading
Start from yourself, not from the job.
Future Finder is free, takes about ten minutes, needs no account and stores nothing. It will tell you which kinds of work you lean toward, and which of these pages are worth your time.
Parent of a teenager weighing this up? Career coaching for teens gets them talking to people who do the work, anywhere in Ontario.
If this appeals, look at these too
software developer cybersecurity specialist accountant marketing specialist
Read next
University vs college vs apprenticeship Careers that match your strengths Career aptitude tests for teens All careers guides
Wages and training requirements change. Every figure on this page is linked to the body that publishes it, and the review date above says when we last checked. If you find something out of date, tell us and we will fix it.