If you’re working in a business, they are very important. Have a look at the introduction at the beginning of the book.

Oh, just in case you hadn’t realized this, you can download the free kindle sample of the book and read the preface, introduction and first chapter or two without actually buying the entire book. This will probably give a better answer to your question than I could type in here.

Thank you very much, I will do as you say 🙏

Hi Toxicafunk, we need to consider the WHO, the WHAT and the HOW of the storytelling process. Making good charts is just part of the HOW. In my experience, many data scientists fail to understand their audience (WHO) and, surprisingly, also don’t have a firm grasp of their own message (WHAT). I talk about these in more detail in chapters 2, 6 and 7 of the book.

Great insight David Stephenson I think getting the WHAT right comes from experience, but the WHO is always a tricky.

Actually, getting the WHAT right is really difficult, even with experience. Most data science presentations don’t truly have a clear, focused, relevant WHAT.

I think the #1 skills is ‘empathy’ That is, to understand the background and perspective of the people you are working with. If you get that right, you’ll be able to identify the most important business questions, communicate effectively, and work through difficulties and roadblocks together with those stakeholders and colleagues. It’s the focus of chapter 2 in the book, where I try to illustrate how various horizontal and vertical segments of the company can be completely different. it’s important for us to understand those differences so that we can work effectively.

Hi Saulius Lukauskas, when I was finishing grad school, some classmates looked into consulting and went through that Mckinsey/BCG case study interview process. It’s interesting, and it fits their business model, but I think it’s a bit less relevant to data scientists. It really depends on what skills you expect for the role you’re hiring for, and whether you are expecting the person to identify novel solutions or rather focus on technical progress. This relates to the topic of whether to have a designated ‘Analytics Translator’. More relevant in our case, I think, would be case studies focused on identify data sources, scoping PoCs, and choosing the proper type of model to start with. Also, giving the applicant a scenario where a project was failing and asking them for suggestions as to why they think it failed (I have an exercise like this in one of my training modules).

Maja Thanks for the positive feedback on Big Data Demystified 🙂 Glad to hear you found it helpful, and impressed you read it so quickly!

You are welcome! Love the book!

Hi Pavitra, in my perspective the best thing that could be done is an assessment of Risk and Impact and share that with your C-Levels. What kind of disruptions would it cause if the data doesn’t have enough quality, if that finds to be a big issue that you can present the prof on your assessment that things could go side-ways and don’t simply expect for the best. From your description it seems to me that you required a team in order to provide a production ready product/feature and your C-level should support you on that or grant you the required skillets to accomplish alone what they ask.

Hi Pavitra Chakravarty It’s really difficult to be the only data scientist in a company that isn’t used to working with data scientists. It’s also very common. I’ve spoken with many people in similar situations.

I would say, no, you should not have insisted on first getting a team, especially in a startup, where resources are tight and everyone is expected to be flexible.
In situations like this, you need to be extremely agile, which, at it’s core, means working very closely with your stakeholders and getting input and feedback as often as possible.
They in turn should recognize that mistakes will be made.
It may be that you’d feel more comfortable in an established team at this point in your career. That’s also a legitimate career choice.

It’s also quite possible / very likely, that your CTO was simply very stressed and /or didn’t have reasonable expectations of you.

Guys thank you all for the great suggestions David Stephenson - so privileged to have your take on this. I didn’t know if your big data book but am buying it now. Looks really great

I think the hardest is learning to understand and relate well with colleagues, especially non-technical colleagues. I devote four chapters to this subject. We can stay in our technical bubbles all through school, but then suddenly need to relate to completely different types of people. Of course these aren’t typically the top-of-mind skills, such as storytelling, managing expectations, finding the best use cases, etc (ch 6-11), but not relating effectively in certain ways can easily shipwreck a data science effort.

Lalit Pagaria Also a great question, and very relevant. I know some data scientists who simply don’t want to work on business problems b/c they aren’t interesting enough!
I really encourage data scientists to work closely with the business stakeholders–to participate in scoping meetings, to present their work in regular demo’s, and to maintain regular contact with these business partners (eg. on Slack or MS Teams). I myself work hard to ensure that the data scientists recognize the importance of the business problem and check that their presentations to the business have relevant content (and are not too technical).
In my experience, not enough effort is put into developing the WHY of a project (I’ve called it the WHAT of storytelling above). When this is done right, it should provide a good basis to motivate the data scientists. Oh, and make sure you don’t hire people who truly want ONLY technically-interesting work and have no interest in delivering business value.

Thank you, great answer.

The interpersonal skills are a bit more challenging.
I’ve found it more difficult to read people and to forge new relationships when not working together in the same office. It’s also more difficult to follow up on people who don’t respond to emails.
Cultural differences can also be amplified (I think I give the example of cc’ing managers in chapter 3, which is a bigger deal when communication becomes more electronic).
I’m definitely looking forward to face2face work again!

I wouldn’t say there is an entire layer, or even role, that would be eliminated, but it would make everyone’s lives easier. Consider the product owner role. Can / could data scientists serve as their own product owners? I would say Yes, if they have the appropriate business skills, but that’s perhaps not the best use of their time. Regarding middle management, again No, as its function is broader than simply providing business skills.

Can you share the link? I will also add it to the book’s page on our website

One question David Stephenson is the book only available on the kindle store ? I have a kobo reader trying to understand if i will bump into issues

Ah, that’s indeed a good question. Let me check into that. BTW, you know you can read kindle books on a phone or computer app, as well as online, right?

Yes, i know. But i’m trying to reduce the amount of hours at the computer and the ebook readers are a good option for me. I’m finishing some other books, and yours on the radar, thanks for sharing with community by the way.

Rui Ramos Can you check the kobo store again now. It should be there either now or within a few hours. Not sure if kobo displays color, but you’ll want color for chapter 7.

Thanks David Stephenson just bought it, was checking the cap 7, i have a gray scale on the charts so i think i will manage. Many thanks 🙌

Hi Tim Becker, congrats on the new job!

• 
  Here are a few tips
  
  1. Identify what’s currently important to your organization.
     
  2. Actively grow your internal network and identify the colleagues most open to working with you.
     
  3. Work to produce value quickly, even in small ways.
     (I talk about this more in chapter 1)
     
• 
  Managing expectations is tricky, but really important. It’s really important to have both a stakeholder scoping meeting and a project kickoff at the start, and during those times you need to not only understand requirements but also manage expectations, which includes making clear both the estimated effort and the amount of inherent uncertainly involved. I talk about these topics in the second half of the book.
  
• 
  Regarding career, it depends on where you want to go and what’s important to you. I’ve devoted an entire chapter to this one. If you can ask something slightly more specific, maybe I can answer that here 🙂
  
• .
  AI solutions are often not 0-1 replacements, but are tools to make tasks easier. Either way, you’re typically going to be working through a senior manager, and they will manage this tension. From your side, you want to communicate well with the subject matter experts (who may feel threatened) so that you understand their pain points and can work in the initial stages to develop a solution that makes their life better.

David Stephenson thank you for your advice and answers! Yesterday, I read the chapter that is available in the kindle store and in my opinion it is a really fun and interesting.

Concerning the career, at some point I would like to take more responsibility and lead and guide other data scientists. In addition, help the management to decide which project are particularly promising and to develop concept on how to best approach these.

Tim Becker, glad to hear you enjoyed the chapter 🙂
Regarding your career goals, I would say focus on three things:

1. Get to know stakeholders outside of your team and understand what their work looks like. This will help you deliver meaningful projects.
   
2. Demonstrate a high degree of responsibility, taking initiative and ownership whenever possible. This will help people trust you with additional responsibility.
   
3. Develop good communication skills and take the opportunity to present outside of your team when possible. Good communication is critical for leadership and influence.

that was fast, thanks again 🙂

Hi Luis, this is a very common challenge. Fortunately it’s getting somewhat easier these days, as traditional companies become eager to not miss the boat in this area (and supervisory boards even place pressure on CEOs).
What I sometimes do as an external consultant is to set up data strategy workshops to quick-start projects, but that assumes a certain level of buy-in being there already.
As an internal, I would recommend that you

1. Identify people in the company who are interested in trying out new solutions. Sometimes this is top-down, from executives who latch on to buzz words, and sometimes it’s bottom-up, when practitioners are eager to bring in new approaches.
   
2. Identify top-of-mind business goals and build a strong case for how data & analytics can meet those goals. This could be in solving a pain point or in making a substantial improvement and might generally fall into the business goals of increased profit, increased market share, decreased cost, or decreased risk. First you’ll need to understand the currents flowing in your company, else you’ll be selling a solution that is swimming upstream.
   I might be able to give you more specific advice if you have a few minutes to describe your situation. If you send me a DM, we can continue from there.

Thanks, Laia, for the positive feedback 🙂
How success was measured…. that is indeed a challenging one to answer. The ‘obvious’ answer is through primary and secondary KPIs. Depending on the project, these may have been return on advertising spend, forecasting accuracy, increased revenue, higher CTR, lower churn rate, etc.
In reality, I’ve seen many projects where KPIs were not clearly defined (OKRs are even worse!) One thing I really encourage is to push stakeholders to put careful thought into KPIs right at the start, otherwise you find yourself half way through the project but with no clear north star.
But to be blatantly honest, success is typically measured by how happy your stakeholder is at the end of your project, and that doesn’t always depend on how you’ve moved the KPIs.

Hi Alex, that’s a fair question. TBH, I expect 80+% of the book is material you didn’t cover in business school, and the other 20% is presented with the additional perspective of an analytic role. The book is very heavily based on experiences of myself and others I’ve known, put in the light of research and industry literature and livened up a bit with personal anecdotes.
That being said, the book is definitely targeted at analytic roles, which typically are filled from STEM backgrounds, but I’ve heard from several non-technical people that the content is very valuable to them regardless. For example, the chapters on scoping projects and managing stakeholders are pretty broadly applicable, as are the chapters on working with colleagues.

Thanks a lot for the reply, David! Will give the book a shot 😄

Ok, this is a bit of a dangerous one to answer!
First, analysts tend to process quantitative information faster than their colleagues, which makes it challenging for them to communicate effectively.
Second, the mental precision that enables success in the analytics field can be detractors in the business realm. For example, we find it difficult to cope with ambiguous situations.
Third, analytic work by nature involves less social interaction, which is a big minus when it comes to understanding your colleagues. This hits especially hard in a business environment, where people in marketing, finance and strategy are so very different in their priorities and ways of working (see chapter 2)
What would help in education is to give students more exposure to experienced professionals, ideally including non-analysts but realistically at least analysts with strong business experience.

Hi Quynh Le, it’s really good you’re asking this question. A common complaint from senior managers is when analysts only provide charts and data but don’t act as ‘thought partners’.
First, it’s important to understand the ‘question behind the question’ When someone asks for data or a chart, try to understand the business motivation.
As you prepare your presentation, try to imagine what follow-up questions your audience might have, and prepare for those questions.
Also, be prepared with recommendations. It’s possible that your audience will not ask your opinion (or even want your opinion!), but it’s good if you have thought through in advance what you think the best course of action would be, based on your analysis.
In addition to doing this for your own projects, take the initiative to chat with colleagues in other teams to hear what their challenges are. This will not only help you identify new applications but will also give you a more wholistic view of the company, which will go a long way in helping you propose solutions that are most relevant for your company,

David Stephenson thank you for sharing your opinions! I find myself often caught up with the analysis and forget about what else next. From your answer, I think that an analyst really needs to think deeply about what data could means to business strategy and to anticipate what stakeholders really ask more from the analysis. I also find your point about chatting with other colleagues really helpful. Thanks for your sharing!

Hi Michael Kachala, that’s an extremely difficult question. DS is often an enabler, rather than a core part of strategy. Not sure if I would succeed in giving a good, concise answer in this thread.

Thanks for the reply! I meant simple in formulating, but answer indeed can be puzzling. I was thinking on this topic myself and don’t have a definitive answer.

Hi Tony Gunawan, I’m afraid I can’t answer your question about kindle availability. I see it available now, but I’m not in Indonesia 🙂 Curious if anyone else on this channel can help?
Indeed, someone asked a similar question yesterday about end-to-end delivery in a startup. It’s generally not an efficient use of resources within a company, but in a startup everyone needs to stretch themselves way past their comfort zones. It’s just the nature of the job choice. If you want to be more focused, it’s probably best to find a company with a developed program and larger teams.

So it seems the issue is not in the colleagues per se but on the nature of the startup company itself, isn’t it? It’s encouraging enough (somehow) to know that it is a common problem within the startup culture and not in overall companies. Thank you so much for your reply. Have a great day !

I don’t really write about that topic in this book. My general principle for companies is to align their interests with those of their customers and to not do anything which they would be embarrassed about if it were made public.

Hi Kyle Shannon, not specific feedback, but there is a lot of overlap between the target audiences. That is, there are two types of corporate trainings I generally give. The first covers business skills for data scientists. The second covers skills for Analytics Translators, who are, roughly speaking, the non-techies who work closely with data scientists. Some of the chapters in the book are used in both trainings (specifically, chapters 5-11). You might check out the chart at the bottom of this page
What’s different in the Analytics Translator training is that I also talk about the basics of Statistics / Machine learning, to help business colleagues put everything into perspective and to help them manage some of the pitfalls inherent in DS work.
A third target group would be business executives, for whom I host workshops rather than give trainings. In this case, I focus on the big picture of applied analytics, how to bring value to a company, how to choose technologies, hire teams, form a roadmap, etc. That is really more the subject of my previous book (“Big Data Demystified, How to … “)

yes

Hi Eimhear Rainey, Most of the skills in the book will still be relevant, as they are foundational, but there will also undoubtedly be unique aspects of change management that you’ll have to deal with. It’s hard for me to speak to those nuances, as I have little experience working in the public sector. What the book should still do, however, is to give you the basis for discovering the best way of working in your own surroundings.
If your organization is using a highly structured project management framework, such as Prince2, you may have difficulty using some of the techniques I write about in chapters 8 and 9 (scoping projects and planning kickoff meetings), as well as in ch 11 (agile, kanban, etc,) If that’s the case, the principles in those chapters will still be valuable, but you’d need to adjust how they are used.

Hi Duverger PETGA, the book covers much more than problem definition. I would encourage you to download the free kindle sample and read through the introduction. It will explain the six different topics covered.

Hi Sepp, first, congratulations on the promotion! Sounds like some great feedback.
I’m seeing several questions in this paragraph. For the first, I have a method I typically use in initial scoping meetings. It takes 1-2 hours to run through, but it’s worth it. I describe it in detail in chapter 8, but if you don’t have a copy of the book, here is a link to at least show your the canvas for it: http://dsianalytics.com/4
Your second question is a bit more difficult to answer, as it depends on whether you already have a place at the strategic table or not. If not, see chapter 1 in the book (available in the free kindle sample). If so, it’s more in chapter 10.
It’s difficult to really answer this w/o more context. Send me a DM and maybe we can set up a short call.

Maja, It’s largely a mindset, from which you develop the skills. In chapter two of the book, I walk through examples of how colleagues differ with respect to Goals, Values and Working Environments, to help get you started with that mindset.
A lot of companies are sending people to trainings in something called Design Thinking (DT), a customer-focused technique for designing relevant products. This might be helpful. I use principles from DT in my strategy workshops, but I’ve never attended one of these specific DT trainings.
Of course I’m biased, but I’d suggest the best way to quick-start this skill for data scientists is by reading my book or attending one of my trainings. I have a few trainings scheduled for this fall. Which reminds me, I should probably mention that in the main channel.

Thank you so much for everything David Stephenson! I’ll tell them about your trainings. Have a great rest of the week!

Hi Krzysztof Ograbek,
I would say that the business skills make data scientists much more valuable. Without them, you’re dependent on your manager or colleagues for communication, prioritization and stakeholder management, which really puts a roadblock on your career.
A few surprises in store:

• Your non-technical audience probably understands about 5% of your presentations
  
• Many of your colleagues have completely different definitions of ‘work well done’
  
• Your network is more important than your knowledge when it comes to advancing your career
  In terms of biggest misconception, I would say that it’s very difficult for non-techies to understand the effort required and the risk entailed by DS projects. This is why stakeholder management skills are so critical.

This is why I said “the who” was the hardest. Many times have I found myself thinking I’ve successfully presented something available for non-techies only to find out I grossly overestimated the minimum tech knowledge required to understand my presentation… 😅

Hi Asmita, yes, DEFINITELY build domain knowledge. Try to use the company’s product, sign up to your own mailing lists, buy your own products online, etc. When you work on a project with stakeholders, ask them questions about the project background, default solutions, how competitors do things differently, etc. You’ll build the domain knowledge while also earning the trust of your stakeholders, as they notice that you are listening. It will also help a lot in designing your solutions.
On that subject, your first ML solutions should be designed in light of current (manual) best practices. They should probably use similar intuition and produce comparable results. I’ve seen analysts deliver complicated models that couldn’t compare with manual methods, and they lose credibility quickly this way.