Cookie Policy

We use cookies to operate this website, improve usability, personalize your experience, and improve our marketing. Privacy Policy.

By clicking "Accept" or further use of this website, you agree to allow cookies.

Accept
Learn Machine Learning by Doing Learn Now
You are reading articles / Career
Data Science Consultant Cover
Jeremiah Mannings
Author: Jeremiah Mannings
Senior Data Science Consultant

How to Become a Data Science Consultant in 2018: 8 Actionable Tips

Successful data science consultant Jeremiah Mannings breaks down what you need to know to start a consulting business of your own

Intro

If you're anything like me, you've got an itch for doing your own thing, growing your own business, and having the flexibility that conventional jobs don't usually provide.

If working at a company doesn't fit your personality, and you're not quite ready to do a startup, then consulting is great way to break free.

In this interview, Jeremiah Mannings shares with me how he got started in data science consulting, what kind of things he works on, and some extremely valuable and actionable tips to get started today.

~ Brendan

TL;DR

Eight tips to get started today:

  1. Network and become a trusted advisor
  2. Build a portfolio
  3. Attend industry events
  4. Get referrals
  5. Partner with data firms
  6. Do market research
  7. Build a network of consultants
  8. Patience!

About

Tell the readers a little bit about yourself. Who are you, what’s your educational background, and how did you get into data science?

My name is Jeremiah, I am currently a Senior Data Science Consultant at the consultancy Capgemini, as well as running my own company Evolved Projects (evolvedprojects.io). I hold a bachelor of engineering in electrical and electronics (honors), as well as a diploma in project management.

Previous to my current position I was a Data Science Consultant at a smaller company, and previous to that a Cognitive Associate at IBM’s Watson. I am also one of the top 25 analytics professionals in Australia.I first got introduced to data science by being selected out of the internship queue for a place in the IBM Watson division, of which I was the only intern selected for this in APAC.

From there, I primarily learned through experience, being mentored (heavily!), and being thrown into the deep which often happens with consulting. I also had the opportunity to learn fast on many occasions, as the team was small I was treated as an equal resource, which means I wasn’t just given menial tasks. I was also fortunate to have some mentors that taught me a lot of what I know today, primarily the intersections between data science and business.

Having a technical degree with a focus on calculus and other mathematical forms helps immensely toward understanding what's actually going on with what your implementing (and helps when explaining it), however, it’s not required to get into data science, and it's something you can pick up with some practice (and effort).

The beauty of data science at the moment is it's not a career track within itself yet, which means you get a lot of diversity within the ranks. Sometimes the diversity of background can be one of your greatest strengths, particularly when that involves heavy use of soft skills which is one of the core parts of consulting on a daily basis.

Clients

How do you find your clients right now?

Clients come through multiple channels; current clients wanting more work, clients passing on results of work (leading to more engagement with new clients) as well as building portfolio and ability in various areas that leads to specific work.

There’s a fantastic book called The Trusted Advisor by Robert M. Galford which dives into the key value a consultant brings - trusted advice. Having your team become known for providing this, as well as an open and thoughtful approach helps future clients consider what you are offering.

There is also a focus on tendering and other bidding processes that are open.Marketing can also be a useful tool but more often than not I tend to rely on my network as well as contacts in industry to find this type of work, additional open tenders can be great for that as well. The issue with marketing is that this is such a niche specific industry that picking a target for marketing will often fall short, and your better off spending your time and energy becoming a trusted resource.

Writing opportunities, speaking opportunities, running your own workshops, offering free internal workshops (or paid ones), anything that gets you up in front of groups of people helps with finding opportunities.Don’t be afraid to just message conferences and workshop vendors to get your name out there, you’d be surprised how many are interested! These can be local in your area or larger, personally I focus on both as it has a different effect, being a speaker from out of town makes people less likely to engage you for actual work (even if they love your presentation).

The goal of each event like that is to send the right message to at least 1 person in the room, regardless of how many attend. You only need to impact the right person to build a lasting contact!

Our consulting office being built

Lifestyle

What does your day-to-day look like? What kind of life can people expect from having a consultancy as opposed to a job in a company?

It's quite varied, and it changes from project to project. The location is constantly changing every 9-10 weeks on average, however, personally I generally schedule 1-2 days a week working from home (dev days I like to call them) so it works out quite well.

Effectively all of the projects I have been involved in and run are done so with agile methodology, so morning scrums, retros and planning weekly or biweekly. This affects the way you work as well, as its completely independent, and up to you to stay on your deliverables and that of the project.

Generally, most projects contain ‘stages’, with workshopping and design sessions the first 1-2 weeks which is useful to define the project and gather inputs from stakeholders around the business to what the expected outcome is as well as get a handle on what's available (data, platforms etc).

Then there are multiple weeks of development and iteration time, inside of this generally I take on a few different hats. These include understanding and documenting the business problem (managing stakeholders), architecting and designing the solution, presenting and validating the solution with stakeholders, consistent updates and checks with project sponsors, and finally actually building it as well! It relies heavily on understanding the business process (and people) in order to effectively get through each stage.

Projects and Pricing

What sort of projects do DS consultants do?

It really does vary, but oftentimes it's inside of other larger projects (software projects etc) where data science is a component of that project.

For example, setting up a recommendation system for products is generally a step on the end of setting an MDM (master data management) system or a data warehouse. It pays to be aware of how these more traditional types of data projects influence the data science space as often they go hand in hand.

There are of course a lot of POC (proof of concept) projects which are smaller scale, 4-6 week projects that focus on validating a concept at a small scale before moving ahead with a larger project. This may involve demoing capability of a certain tech stack, or building a quick model to show that there is potential. For example, you may be given a use case to test like ‘build a model that shows the effect of x on y’, in that 4-6 week period you would build a project that assesses the validation and business value of that proposition and if it suits being made into a full project.

There is also workshopping, strategy work, and teaching internal client teams.

How does a consultant differ from an employee? Why do companies hire you?

It's the process and interaction surrounding the actual work that differs the most. Consultants are also required to wear a lot of different hats from time to time, from architecting implementations, to setting them up, building, testing, iterating, documenting, to teaching client resources for a handover, and even sometimes building models!. Then on top of that, consistent stakeholder feedback, understand what the client actually wants out the project, explaining concepts, pre-sales pitching of ideas and concepts.

The main differences are around the more general understanding of business fit & application, and the ability to fit your ideas and projects in a way that suits the client and what they understand.

In consulting there is rarely a chain of command telling you exactly what to do, giving you 20 weeks to do it, and then reporting your progress when you're done. It’s much more involved and consultative within the process itself. This essentially means that hiring a consultant you are not just getting a predefined project built, you are paying to get the experience, business knowledge/understanding, to define and build a project that suits your business.

There is also an interesting phenomenon that tends to happen where after some initial pushback, internal teams work well with consultants and consultants can generally get results internal to a business quicker than people within that business can as your outside of the normal business framework it allows a lot more flexibility.

What can people expect to earn and how they should charge for their services?

It’s hard to pick a specific rate because it really depends on what your doing and the perceived value of the projects you are undertaking. Are you providing something no-one else (in the vicinity) can provide? Then your project/rate should reflect that, never undersell a skill that's hard to find.

What's an example of a project you worked on?

A good example of this is a project I was brought into to determine the roadmap (design the strategy, what tech to use, machine learning pipeline, etc.) it was a 2.5 week project which required looking and understanding their data by meeting with SME’s (subject matter experts) and determining exactly what needed to be done to the data to achieve their goal (which was to identify & predict customer pain points with billing data).

From this I designed the solution and the roadmap and wrote out how their internal team would design it. I also was brought in a few more times on the project to validate and check their work, as it was not only used to build a good project but train up their internal resources for the work which was a great way to approach it.

On this project I was on a rate of $2000 AUD a day, for the 2.5 week period which was inclusive of further recommendations later on (which totalled no more than 4-5 hours). For a top-shelf data science consultant with the ability to explain to and understand stakeholders, as well as architect and design large-scale data science projects with a considerable background can fetch AUD $2500+ per day, with more niche skills clocking in higher than that still. With most rates coming around $1600 - $1900 per day for contract work.

However, the more effective way of charging is to charge based on the project as this accounts for all costs and its inclusive of the wage, and oftentimes the margins will be higher on these types of projects.

Do you outsource any of the work?

It depends on the nature of the work being done, the team I work within has scope for a majority of the facets of data science (and sometimes near enough that we can figure it out) but when there are obvious gaps we generally try to contract trusted people and failing that we use outsourcing. When using both contracts and outsourcing, its a good idea to do the architecture and overall integration design yourself ahead of time, not only to provide clarity but to make sure that what you're after in context can be delivered by the resource. Outsourcing can be good if the resource is trustworthy, but it also has to be factored into the larger project at hand if the development style and knowledge of the person undertaking the work fits into the overall picture. For example, if you serve models in docker images, yet a resource doesn't understand that method, does it still work in your architecture?

Advice for Newcomers

If you could go back to the beginning with your current knowledge, what would you do differently?

I’ve been fairly fortunate in the way I’ve come through the ‘ladder’ so to speak, however personally I would have focused on key skills earlier on, instead of trying to be overly generalist, which is a difficult role in data science as near impossible to have an adequate handle on all of it at once. Being realistic with your skills and capabilities is quite key, however, don't be afraid to jump into new tech and things just because of lack of experience as its the best way to get that experience.

What were the biggest mistakes you made?

One of the things I did early on in my career is I didn’t quite recognise how much certain skills are worth and undersold myself quite a lot. Understanding exactly what you are good at and your capabilities (a healthy dose of introspection) as well as what you are not good at can save you a lot of issues and headache.

One data scientist can't do everything, and its best to be realistic about that and embrace what you can do. The other thing would be to never undersell your skillset, it takes years to build some of the niche skills in this space and underselling them can hurt your prospects moving forward.

What are some tips for newcomers that want to start a consulting business?

Consulting is built on trust, and trust is developed a number of different ways. Becoming a trusted advisor to a client is a key step, that means walking them through the good and bad, and making sure honesty is present in the relationship.

From there you also need to have experience (or perceived experience, more on that further on), as experience will help you become that trusted advisor.

There’s a saying in consulting, “as long as your 5 minutes ahead of the client” which I find to be interesting thing to consider when your starting off, and it speaks to the nature of consulting, using your experience, skills, and flexibility to understand where the needs of the client are before they articulate them. This is developed more and more over time, but keeping this in mind can help you become an effective advisor from day 1.

Having mastery over technology concepts helps a hell of a lot but it's only half the battle, the other half is knowing how to explain them to business people and then relating that concept to a business problem they are having.

For example, how would you explain neural networks to a business user and justify using an RNN (recurrent neural network) for a language understanding problem over other approaches? If your first thought is ‘well because it's the best way to approach the problem’ take a step back and put yourself in the shoes of the business person, they are trying to solve a problem at hand the way they solve it doesn't really matter as long as it works within budget and timeframe. Your response has to include the expert opinion on the approach, as well as how and why it affects their goals in order to be an effective advisor.

Actionable Steps

"Making AI Approachable," presented by Jeremiah Mannings at the Victorian Digital Innovation Festival

What are the first steps someone can take to get started consulting?

Focus on building yourself and your status with individual businesses and people first. This can be things as simple as running workshops to get business owners and people in to understand the fundamentals of data science, you can start to help people realise where it can actually fit within their business. Because you were the one to show them that, you start to become the trusted person in their mind. This means that they will approach you first then they find use cases within their businesses.

Consulting and being a developer for hire are very different things and it pays to keep that in mind when trying to finding long lasting clients. You not only want to build cool projects but generate understanding around those projects and helping to develop people's knowledge around them. This is what will get you repeat and valuable projects.

The other approach to this is the niche route, if you have a skill that not many people focus on, for example, natural language processing or a specific type of neural networks of which you have established yourself as an expert, you may have a platform to solicit specific types of projects.

A few tips to get started today on your journey to consulting:

  1. Start building a network and become a trusted advisor - This can be done many ways, but some of the best ways to build trusted advisor status is offer free workshops, offer free evaluations, speaking events, articles, writing, etc.
  2. Get a portfolio - one of the hardest things to show for a DS consultant is models in production, but it’s also one of the key validation points for your background. This could even be hosting an API that does something interesting with a simple website, its out of pocket expense but it pays off
  3. Attend industry events - both paid/meetups/presentations etc. and make sure to connect with people there. Oftentimes people in the audience are not experts and they are there looking for answers
  4. Contact some of the salespeople at larger consultancies for referrals. Larger tiered consultants often don't take work under $100k (they call them micro projects) and if you get a useful contact in one of these places they may hand you off to smaller work at the start
  5. Look into partnerships with data firms - there are a lot of data firms (warehousing/cloud/etc.) that want to move into data science and have clients interested by can't justify hiring a full time resource. If you can show them you are a trusted recognised data scientist you may be able to come into their projects in a consultant's role
  6. Market research - Learn what your peers and other consultants are charging in the space, perhaps undercut them slightly (not too much, remember not to devalue your work) it may help you be competitive.
  7. Start reaching out and building your own network of consultants in different spaces. You’d be surprised how often you need advice from cloud experts and data engineers during the process of a data science project, and this works back and forth.
  8. Patience! The hardest part to all of this is building a network and trust out of pocket until it gets to a point where it can pay you back. Don’t give up when you don’t get things out of doing things, as the most valuable thing a consultant can have is the the view of an advisor in the client's eyes, which may take time to pay off, but when it does it will be worth it.

I hope those tips weren’t too overwhelming! The main thing to think about is consistent effort toward your goal. Becoming a trusted consultant takes a lot of out of pocket time and effort. You may not see much at the beginning, but keep moving with it until you get to a point where your network starts giving back.


Meet the Authors

Jeremiah Mannings

Senior data science consultant at Capgemini, co-founder at Evolved Projects, & data science speaker. Intent on removing the abstract from the wide world of data science!

Linkedin - Website

Get updates in your inbox

Join over 7,500 data science learners.