Data Science & Freelancing – an Honest Opinion (with Concrete Tips)

One of the most common questions I get from aspiring/junior data scientists is this:

Is it possible to get freelance gigs and jobs as a junior data scientist?

In this article, I’ll answer this question and I’ll get into the details:

  • When can you start to think about freelancing?
  • What should you do now to become a freelance data scientist in the long-term?
  • How to get clients? How to get references?
  • What types of data tasks do companies outsource?

What you’ll read here is based on my and my data scientist friends’ experience. Some parts of it won’t be the things you want to hear… but definitely the things you need to hear.

Let’s start with the main question:

Is it possible to get freelance gigs and jobs as a beginner data scientist?

Not really.
I mean, yes — there are always exceptions and rare cases. But even if you get a freelance gig as your first data science job, I do not recommend taking it. And just in general: I do not recommend looking for freelance data gigs/jobs as a junior.

Why?

Because the first years during your career should be about learning. Even if you have done the prep-work perfectly:

…even then, it’s very unlikely that you are ready to work on your own.

In your first years, you’ll run into many, many, many issues that you haven’t met before and that will require some level of seniority.

I know from personal experience that even the most talented juniors have these issues all the time in the first years:

  • They are not aware of all the common bias types that can affect their day-to-day job and the results of their data projects. And you know, a biased data project is worse than no data project at all!
  • They don’t have enough experience to prioritize and navigate between the different business cases. What’s important or not? What’s doable and realistic — and what’s not?
  • They don’t dare to say no to a project that shouldn’t be done.
  • They don’t know how to communicate with stakeholders who are less experienced with data-driven methods… and they don’t know what to do in the (very common) case when an organization is not ready to be data-driven because of the missing data culture.
  • They don’t have enough best practices under their belt.
  • They don’t have a network they can count on.

You can read books and articles about these things. But nothing will prepare you as much as working a few years as a full-time employee at a company. As a junior data analyst or scientist nothing is more valuable than having a good mentor, a data-minded manager or senior data scientists on your team — who you can learn from in the first years.

Becoming a freelance data scientist… What’s the first step?

This might sound paradoxical but:

To become a freelance data scientist on the long-term, your first step now is to get a full-time junior position.

I’ve already mentioned the main advantage of it: having colleagues, seniors and even a mentor you can learn from.

But it’s also important to add that as a full-time employee you are in a much safer environment than as a freelancer.

For one, there is a lower risk that you’ll mess up something real bad.

Doing data science comes with making mistakes — and learning from them. That’s normal. Especially when you are a beginner.

If you make huge mistakes as a freelancer, in the best case, you won’t get paid by the client. In the worst case, you’ll also burn your reputation and clients will start to avoid you.

If you make huge mistakes as a full-time employee… well first, you won’t make huge mistakes at all, because you’ll get code reviews and your mentor will fix all your statistical or business thinking mistakes. So that’s already much better. But even if you make mistakes, all you’ll have is a discussion with your manager on how you can avoid these in the future.

Having a full-time job is the safe environment that you need in your learning years.

data science freelance patience

I know what you think! Like it would be so easy to get into a junior data science position in the first place. If you are an aspiring data scientist right now, check out my free video course first: How to Become a Data Scientist.

What should you do now to become a freelance data scientist in the long term?

Okay, you had the pep-talk. I guess you get the point that, if you are a beginner data scientist, don’t take freelance gigs now. Be patient, wait a few years, learn, etc.

But I have to add that if you really want to become a freelance data scientist in the long-term, you can start to prepare with a few things.

Let’s say that you took my advice — and you work as a full-time junior data scientist at a company.

Here are three things you should do now that you’ll profit from in the future:

1. Make yourself visible! Build your self-brand and marketing!

As a freelancer is not enough if you are a great data scientist. Your potential clients also have to know that you are a great data scientist!
You don’t have to be a marketing guru… but you’ll have to learn how to make yourself visible.

Note: That can be useful if you are looking for full-time jobs, too, by the way.

Building your brand and marketing is never easy! (After all, it’s a different skill set than doing data science projects.)

But there’s a nice organic way to let people know who you are and what values you can bring to their businesses.

It’s knowledge sharing.

freelance data scientist knowledge sharing

The medium can be whatever you prefer:

  • a Youtube channel
  • a blog (like this one you’re reading)
  • a podcast
  • Linkedin posts
  • conference/meetup presentations
  • or anything else you can imagine…

But the point is that you should share your knowledge, your experience, the things you learn on your data science journey. Or in other words: you have to provide value.

It’s the best and easiest if you simply talk/write about the projects that you work on.

You’ve done a hobby project others can learn from? Great, just write an article about it!

You’ve met and overcome a unique challenge in your job? Awesome: give a presentation about it. For instance, in a local meetup group. (But don’t forget to ask your employer whether it’s okay to share the things you want to share.)

Even if you are a junior, publishing your case studies always attracts your audience… And it’s also a good way to show what you are good at.

Why start this now? Because it’s a slow process. Getting recognized can take years of consistent work.

My example:

I started Data36 the same way. In 2016, I had an average of 500 visitors per month. In 2017, this grew to ~5,000 per month. (Which brought me enough projects to quit my full-time job.) As I write this article, in 2020, I have ~100,000 visitors per month — and I get job/gig offers almost every week. (Although I don’t really take freelance gigs anymore.)

That’s 4 years!

So take your baby steps towards marketing today! It will pay off in a few years.

2. Start to build your network!

A great network of data professionals can be useful in many ways:

  1. It’s easier to get help when you get stuck.
  2. You can refer each other to different gigs or jobs.
  3. You can make friends and you can work together, inspire each other and learn from them.

There are quite a few ways of building your network.

  1. Go to meetups!
  2. Go to conferences!
  3. If you are brave enough and if you have a great project to present (a hobby project, or a case study from your full-time job), you can even present these at meetups or conferences.
  4. You can simply look up people on Linkedin and ask them to have a coffee and a chat.

All these network building strategies are much easier as a full-time employee. If you are a freelancer, you’ll deal with the stigma that you “want to sell something,” even if you don’t.

I’d add one golden rule here: always be honest! Meet people who you would like to meet and not those who are “useful” to meet. Go to meetups that you are interested in anyway. And as always: try to give value! Sometimes the value is sharing your experience — sometimes it’s only listening carefully to an expert.

3. Pro tip: Use the reputation of your workplace!

If your full-time job is at a well-known company, you can use its reputation in several ways! It sounds selfish… but it’s a win-win, really. If you grow as a professional, the company you work for will profit from it, too.

Here’s a trivial example:
When you work for a well-known company, it’s easier to get speaking opportunities at conferences/meetups. Event organizers love to have presenters from great companies because they can advertise their event with these names… Also, these speakers usually bring more cutting-edge case studies.

It’s also easier to sit down for a knowledge sharing session with other data scientists from other companies. (Again, having a great network is important!)
My managers always encouraged me to do these 1-on-1s with senior data scientists. They even helped me to get in touch with relevant people. (At Prezi, for instance, we had the same investors as many big well-known startup companies in the Silicon Valley, so I even had the chance to go and talk to senior data scientists at Evernote, Eventbrite, Zendeks, Soundcloud, etc… as a junior!)

And most importantly: when you quit your job and get started as a freelance data scientist, your previous workplaces will be your first and only references!

So if you are already working for a company as a junior data scientist right now, think how you can take advantage of your position other than just work and learn… and if you don’t see any opportunities, consider changing workplaces and going to another (more supportive and higher reputation) company that you can use for the above mentioned things, too.

How will you get your first clients?

This is another frequently asked question I get:

How will you get your first clients as a freelance data scientist?

This is why I prefer to think in the long-term… By building your network and your marketing — and providing value consistently for years —  you won’t have to look for projects: clients will find you!

I say this from personal experience: that’s a very good position to be in. 

Of course, if you don’t have years, you can try out these “shortcuts”:

  1. You can promote yourself with paid ads on Facebook, Twitter, Reddit
  2. There are well-known freelance job portals:
  3. Or you can simply get lucky and get an intro through a friend…

But these are short-term tactics and often, they don’t work very well.
Playing the long game (building your network, building your marketing and providing value) is more stable, more satisfying and more profitable.

What type of data science tasks do companies outsource?

Okay, so let’s say, you got there, you spent years learning, networking and building your marketing.

At this point, it’s good to know that there are certain tasks that companies will never outsource to freelancers — and there are other tasks that they almost always do.

Here are three typical projects that you can expect to work on as a freelancer:

One-off data projects

Smaller companies do not require (or can’t afford) full-time data scientists. Also, working with data is not their main focus yet. But even so, they might need to run occasional data projects.

A typical example for a one-off project is: conversion rate optimization for e-commerce companies.
I did many of these between 2014 and 2016. And it was quite easy to standardize it. I analyzed these online businesses’ data sets, looking for unusual patterns and issues to be fixed. And I ran a few qualitative research rounds, too. We mapped out potential usability issues where customers churn in the shopping process. Then we set up A/B tests to validate these ideas.

These projects took ~3-6 months and the value was pretty clear to the businesses I worked with. Before the project they had an x% conversion rate — after the project it was x + y %. Everyone’s happy.

Other typical one-off projects could be (depending on your specialization):

  • setting up leadership dashboards
  • building data-based products/features (e.g. a recommendation engine, face-recognition, etc.)
  • setting up data collection

Projects that need seniority

Even if a company has a data team, they don’t necessarily have all the skills they need on board. If you are specialized in a niche topic within data science — or simply are a senior — they might need your expertise to expand their team’s current knowledge.

I have a few friends who’ve been in the data field for 10+ years… They started as freelancers, now, they get so many of these seniors-only projects that they’ve even built small data science agencies around it.

Typical projects they work on:

  • predictive analytics projects (forecasting, planning)
  • production optimization projects in factories
  • image recognition projects in agriculture
  • and many other interesting things…

The good thing about these projects is that they always work on something new and advanced. The downside is that being able to provide real value in these projects really needs their 10+ years of experience. Also, they had to learn over the years how to pick up the domain knowledge pretty fast in all their different projects.

Well-defined projects that are cheaper to outsource

And the third type of project is something that most freelance data scientists wouldn’t think of… The projects that are actually cheaper to outsource for a business than hire a full-time employee for.

These are typically small, well-defined — and often not-so-exciting tasks.

For instance:
They give you the input data and the expected output… write a Python script for it, and automate it!

As I said, these projects: won’t require too much creativity or business thinking.

Regardless, these smaller tasks can be the exception to what I’ve said before in this article. You can actually do some of these during your junior years, too.

Note that it’s really hard to make a full-time living from these, but they are great for:

  • practicing coding
  • practicing how to talk to a client
  • starting to build your portfolio
  • making some extra money on the side

If you want to check and find some of these, just go to the freelance portals I mentioned above.
Note: the number of these gigs fluctuates over time — sometimes you’ll find more, sometimes less.

How will clients continue coming in? (The long game.)

Once you start your freelance career and get your first clients, everything becomes much easier.

Three reasons:

  1. The things you set into motion will still work. Your network, your marketing, etc, that you built for years will serve you for much longer. (And ideally, you will continue to work on these, as well.)
  2. You can start to build your portfolio and add real clients next to your ex-full-time workplaces. 3-5 big names here will give you enough credibility for the next few decades.
  3. If you are crushing it with your projects and delivering value from time to time, your clients will start to recommend you to others.
    Note that this is a long-term strategy, too. In the last two weeks, two new potential clients reached out to me. One of them was referred by a client who I worked with 4 years ago. The other one by a student who took one of my courses 2 years ago.

    The point is: word of mouth is a powerful thing. And the best marketing is always the satisfied clients!

Building a freelance career is pretty similar to compound interest… Things will start slowly. But if you start to put in the work now, and you stay consistent, over years, these efforts will add up — and you’ll have a great, great return on investment.

Conclusions

The take-away is simple:
Yes, it is possible to get freelance gigs and jobs as a data scientist. But don’t expect that it will happen in your junior years because:

  • getting enough experience and learning everything you need to be actually successful in these freelance projects takes years!
  • Building your marketing and self-brand takes years!
  • Building your network takes years!

Becoming a freelance data scientist is a long game. But in this article, I covered my best practices. Just give it a try, take action… And you can thank me later — let’s say when you can first afford to turn down a client, just because you already have so many cooler ones! 😉

Cheers,
Tomi Mester

Cheers,
Tomi Mester

BONUS — FAQ about freelance data science projects

I got a few questions on Linkedin and Twitter that didn’t fit the flow of the original article, so I’ll answer them here:

A lot of data science projects fail, how to handle it as a freelancer?

When you are senior enough, you’ll be able to predict whether your data science project will go through or not. The truth is that in 99% of data science projects that fail, it’s because of the company culture. You can spot that at your first meeting… And you can say no to a project when you see that the company is simply not ready for it.
Also, working with data, by default, comes with trial and error. So as a rule-of-thumb, the minimum timeframe to work with a client on a complex project is ~6 months to make sure that you’ll be able to deliver results.

data science freelancer trial error

Do you need to know/use more tools as a freelance data scientist?

Yes.
When you work for one company, they’ll have a more or less fixed stack. If you work for 3-4 companies, they might use different Python libraries, different dataviz tools, different types of SQL, etc…

Luckily, if you know one data visualization tool (e.g. Tableau), it’s much easier to learn another one (e.g. Google Data Studio). And that goes for everything.

But even so, as a freelancer, you have to know more tools than as a full-time data scientist.

So, never stop learning, go to workshops, enroll in online courses (hey, if you are an employee, your company might even pay for these… ;-)) and get more tools into your data arsenal!

Cheers,
Tomi Mester

Cheers,
Tomi Mester

The Junior Data Scientist's First Month
A 100% practical online course. A 6-week simulation of being a junior data scientist at a true-to-life startup.