So far, we have seen how to align our objectives with the larger organizations and how to develop a strategy that plays to our strengths. On a day-to-day level, however, we can’t assume that people will adopt our ideas or implement our projects—even if they asked for them! They also need to sometimes be reminded that we are there, and that data science is there also, ready to solve their problems.
Selling Data Science Projects and Ideas
The art of selling appears to be often overlooked both by data science curricula and by the myriads of bloggers and authors that write enticing articles for emerging data scientists. Often, when reading about data science, you can get the impression that building a model with a great accuracy score and then constructing a data pipeline to service it are the start and finish of the job.
The reality is that to be allowed to build and then implement a model, you will have most likely needed to sell yourself, your idea for a model, and then your actual model to people at different stage gates along the way, and frequently there were different groups of people at each of those stage gates.
Often the task of getting past those gatekeepers is reduced by data science writers to a matter of storytelling, and it is true that effective storytelling can be key to success. However, there is a lot more to convincing stakeholders that you are equipped to discover a solution to their problem, and then convincing that your proposal is that solution and should be implemented than telling the basic story of your data analysis.
In this chapter you will learn how to tailor your sales techniques to your customers, how to sell your model, how to sell your team, and how to sell data science as a versatile solution to your organization’s problems. I will also look at how your pitches should vary at different stages of your project, and how to make storytelling work to sell your project, instead of just explaining your data.
Beyond that, we will discuss the need to market the art of data science as well as yourself to ensure that your solutions are seen as the key to solving your client or your organization’s problems.
A key thread running through this entire chapter will be the need to go beyond your purely rational mathematical arguments into considering the emotional needs of the people you are trying to convince. Although this may be challenging, it will enable you to take your arguments a long way further, and therefore make the best use of the carefully crafted rational arguments that you may be more used to.
Mastering the emotional side of persuasion, in conjunction with the rational side, will give you a strong advantage when it comes to getting your models adopted.
Selling Your Data Science Project
No matter how great your data science project is, it’s all for nothing if no one out there wants it. If we can’t convince other people that our work will improve their lives, or at least has that potential, it will just sit on a server somewhere until bit rot stops it functioning.
But what’s the right way to go about telling the customer why our work is going to change their lives for the better? Creating sales documents or presentations that list out all the shiny new things that our data science application can do is very tempting. We worked hard on those features and everyone will appreciate them, right?
Well, not really. For one, it’s very likely your target audience doesn’t have the technical ability to understand the point of what you’re selling. After all, if they had your technical skills, they wouldn’t be thinking of hiring a data science, they’d just be doing it themselves. When you’re creating a sales document, the first thing you have to is reduce the references to the latest tools to the barest minimum that lets them know you know what you’re on about.
The next problem is that you can’t trust that the customer realizes how your solution helps them out of their present predicament. Moreover, it’s disrespectful to get them to do your job for you. Hence, you need to make sure your pitch joins the dots between what you intend to do for the customer and how it’s going to make their life easier.
I will now focus on what kind of model to present to a customer at a pre-implementation meeting where you’re looking for the customer to say yes to implementation and therefore commit more time and money.
This stage could turn out to be the most difficult hurdle to get over because this is where the time and money commitment increases from a small amount to potentially a much larger commitment. In an internal sales situation, this could mean the executive is deciding whether to move the project from within a small data science to a larger team who will need to implement the project and allow it to be used widely through the organization.
Obviously in this scenario, it is vital to be able to link implementation of your model to solving a real-world problem your customer is experiencing, and that the real-world problem should be important to the customer. Where possible, you should be able to link solving the particular problem with saving a certain amount of money.
You will be more likely to be able to find a reasonable dollar value when you are presenting a model internally. At the same time, if you have been successful with your earlier discovery meetings, you will have some sense of how important the problem you are working on is to your customer.
The key to buying model credibility is that the results make sense to the customer. That means that not only will your model need to show great results but also your customer will need to understand how you evaluated your model, relate your evaluation to her business, and believe in the results.
There are many available evaluation methods, and their use can provoke controversy within the statistical and data mining communities. Frequently used evaluation methods such as the receiver operating characteristic have been criticized on statistical grounds, with more robust but nuanced (and therefore more difficult to understand) alternatives proposed, and yet even these simple but flawed methods can be hard for a business audience to understand.
There are also methods such as lift or gain that are tied tightly to the business problem that the customer is trying to solve. In the context of a presentation to sell your results, this type of evaluation is ideal, where the data and problem are suitable. For example, lift is explicitly tied to the marketing goal of increasing sales.
If these metrics are not suitable, developing your own metric that fits the problem may be a way forward. In either case, it is still best practice to also perform the evaluation using a statistically robust method to ensure a correct assessment of the model’s performance without necessarily using that method to communicate the results.
Ultimately, you need to make sure that you don’t let the perfect be the enemy of the good, and when you are communicating results to customers and clients choose the evaluation method that will be understood and accepted by your audience.
Now, although for a few shining minutes it was okay to produce a model that was very accurate without knowing how it came to its conclusions. Certainly, you can still win Kaggle with a model of this kind. Unfortunately, though, it can be very difficult to convince a user to trust a model with an accuracy score alone. On the one hand, many metrics used to evaluate models are inaccessible to people, who themselves are not machine learning users. On the other, for many people an accuracy score isn’t convincing on its own behalf — such people wonder, “Will it work with next tranche of data?” and “Is it too good to be true?”
Self-Promotion
Many people in the data science community are better at doing data science than they are at marketing themselves. This is only natural—if marketing and promotion came naturally, they’d probably do that job instead.
However, as we have seen, within an organization, it is usually not enough to do good work. It is important that your good work is seen and recognized. Beyond what you have already achieved, you also need to ensure that your organization can see what you can achieve, or what you could achieve, if only you were given challenging enough projects to work on.
You need to do this both as an individual and on behalf of your team, whether you are the team manager or not. In both cases, a big chunk of what you need to do is essentially managing upwards. This is obvious in the case of the team member but is true for the manager as well because it is often the case that (excepting specialist data science businesses) the data science function is a relatively small department feeding into something larger.
Consequently, the team leader will need to devote a substantial amount of time to explaining the value of their department to the business at large, including both other team leaders and their teams and more senior managers.
You can find a lot of advice on the best way to manage up with only a cursory search in your library’s management books section or on the net. Obviously, some of it is more applicable to the situation we are looking at than the other. For reasons of space, we will just look at a few of the most easily applied advice.
The first piece of advice is to communicate. Similarly, to the individual team members we will discuss below, you need to keep your senior managers up-to-date with your achievements. Even more than you don’t want to hear about every problem your team encounters, senior managers are even less likely to be interested in the difficulties you had along the way to what you’ve achieved. In fact, beyond a sketch, they won’t be interested in the how at all. What they’re mostly interested in is the business problem your innovation will remove, and how much money they will need to invest.
A lot of the persuasion you are going to need to do to get people’s trust who are either above you or sideways at the same level will need to be indirect persuasion. People in those positions who don’t report to you, and who sometimes may feel threatened if you are successful, won’t always respond to your direct message, even if what you are trying to explain is logically in everyone’s best interests.
If you are the team manager, individual successes by team members will have a habit of being seen as your individual successes. You may feel that you deserve this credit, at least inasmuch as you probably assigned the task to someone initially, gave them the tools to do, and prepared a smooth path for them to achieve their results.
At the same time, you need to take enough of the credit for the team’s successes to demonstrate that you are an effective leader. However, in the long term, this will cost you as the team members who are achieving on your behalf will feel cheated out of praise that more rightfully belongs to them.
As an individual, you have a little more freedom to take credit for your work, given it will often be credited back to the team as a whole. However, there are two common mistakes that I often see. One is presenting work too early, and the other is not presenting it, or holding it back for too long.
If you are a data science team member, you need to keep your team leader up-to-date with your progress. However, she doesn’t necessarily need to see every twist in the path. It’s reasonable to update her when you’ve either reached a point she will recognize as a milestone or when you’ve hit a genuine block.
In both cases, you need to put yourself in your team leader’s shoes to a degree to know whether you’re really there. What would someone on the outside of your task understand as a step forward? What would your team leader expect you to have tried first before asking for her help? At a minimum, she would probably expect you have looked for help in a few of the “usual places”—your other team members, the documentation, StackOverflow, etc.
It’s also important that you consider the best time to present your idea. Buttonholing your manager just before close on a Friday won’t be such a good idea. Ideally, give her some warning that you want to update her on your progress—even if you have an open plan office and she sits right next to you. That way, you’ll have the best chance that she will be ready to talk and think about your work at the same time that you are.
Ensuring that some credit for success flows back to the team is important because it drives the rest of the business to come to you when they have a problem that needs solving—and it drives them to come to you early, before premature “solutioning” means that you need to attempt an inappropriate solution before being allowed to solve the problem properly.
In exactly the same way that it is important for team members to consider how to communicate their achievements upwards to their team leader, it is just as important for the team leader to create opportunities for her team members to showcase their work.
Tailor Your Message
Even when you are talking to people from the same organization, different people performing different job roles have different concerns. Sometimes, when people change jobs, you can even witness the change in how they view different problems.
At the same time, it will be common to present to groups of people, often with conflicting concerns. It can be very damaging to give a presentation that addresses none of a particular stakeholder’s concerns, especially if that person is especially influential in the organization.
- 1.
The charismatics
- 2.
The thinkers
- 3.
The skeptics
- 4.
The followers
- 5.
The controllers
Each of these requires a tailored approach and differing preparation. The overall message here is that you need to understand who you are talking to, and the best way to persuade that person. Sometimes you will even need to consider more than one different person in a presentation audience, and ensure that your presentation speaks to each of them.
However, knowing these aspects of your audience won’t be the whole of your task, and awareness of the existence of these audience types isn’t enough by itself to ensure success. For that you will need an overall guide to persuading others.
Conger,2 also in the Harvard Business Review, breaks down the elements of making your case persuade people when your position doesn’t give you authority over them.
- 1.
Establish credibility. This should be the easy one for a data scientist, and it is true that the job title “data scientist” gives you instant street cred as the local smart person. However, what is often missing is credibility that you understand the business, and, more importantly, that you understand what others in the business are going through. The secret of success here will be leaving aside your data science expertise, and focusing on your knowledge of the business and the immediate problem.
- 2.
Frame goals on common ground. Similar to establishing credibility, this depends on applying your knowledge of the business and your audience to ensure that the solution you propose will meet both the business’ needs and their needs.
- 3.
Emotional connection. In some ways the most difficult, if you have developed the habit of using facts and reason to persuade. This is a two-way connection, where you can express your own emotional commitment to an idea, but also where you are sensitive to your audience’s emotions. The latter is the more useful of the two—learning to read your room’s emotional state will enable you to correctly calibrate your presentation.
- 1.
Trying to make your point with an upfront hard sell
- 2.
Resisting compromise
- 3.
Believing the art of persuasion consists of making a great argument (or arguments)
- 4.
Assuming that persuasion is a one-shot process
Of these, the last two are my picks for the ones most likely to be a problem for data scientists. In the first case, a data scientist is likely to be a person who puts a great store in the content of their argument, rather than the way it is packaged, in some ways as befits someone who has a pretense to apply a scientific approach. Sometimes it can almost seem that spending time on how an idea is packaged is a little shameful—making an idea look better than it really is could be seen as a little like confidence trickery.
Unfortunately, how an idea is received is only rarely related to its intrinsic quality. This can be easily seen from some of the ideas corporations of all sizes take to market—organizations adopt many bad ideas, so it is almost certain that they reject many very good ideas.
Rather than being won on the soundness of their logic, it is far more common for ideas to win the day when they are able to make a strong emotional appeal to their prospective audience. While this goes against the grain for many data scientists, what you should consider is that the idea that you are sure is the superior logical idea will lose out to another path that is logically inferior unless the right emotional appeals are put to the decision makers. For this reason, you need to take the need to frame your idea in the correct emotional terms seriously.
One way of looking at the emotional side of the argument is that it could be seen as the last mile of your persuasion effort. Of course, the rational side of the argument must be sound, but attention to the emotional side of the argument is needed to ensure that your audience or client doesn’t refuse to engage with your argument rationally. We will see more of this later in the chapter when we talk about the role of emotions in winning trust.
Given that the emotional aspects of persuasion are there to ensure the success of rational argument, your appeal to emotions needs to be done without undermining your reason-based case. From that perspective, a little can go a very long way, just like salt or spices in cooking.
You can add a touch of emotion to your argument by doing things like using a rhetorical device, in the manner of a political speaker. Also try to break down barriers between the audience and yourself by personalizing your speech. Opening a presentation with a personal story that turns out to be relevant to your overall message is a good way of narrowing the gap.
Overall, you need to aim to create as much connection between the audience and yourself, in order to build the idea that you are all in it together. That way, the audience will stop thinking of you as someone who is trying to sell to them, and start to think of you and them as part of the same team, who are trying to achieve the same goals.
Benef it Sales
In sales jargon “selling the benefits” is making it clear to the potential customer how buying your product will improve their lives, and has been encapsulated in the phrase “nobody wants to buy a bed—they want a good night’s sleep.” The rub is that in most data science scenarios the problem that corresponds to the potential benefit is a business problem—such as reduced inventory or decreased cost of sales—rather than a human problem, such as a getting a good night’s sleep.
Therefore, being able to complete the journey from feature to benefit requires some knowledge of your customer’s business (whereas everyone knows the benefits of a good night’s sleep—and the horrors of not getting one—far fewer under the fine points of mattress springing and bed construction) and the ability to explain the links. This last is crucial, as the benefits of your work are too important to allow your customer an opportunity to miss them.
What all this means in the end is that the approach of inspecting data sets in the hope of finding “insights” will often fail, and may border on being dangerous. Instead, you need to start with what your customer is trying to achieve, what problems they are facing before seeing which problems correspond with data that can be used to build tools that can overcome the problem. In this area the old adage that “you were born with two ears and one mouth so listen twice as much as you speak” comes into play.
If you became a data scientist in part because your temperament was better suited to quietly analyzing data rather than glad-handing customers, the whole notion of selling may sound daunting. It shouldn’t sound too daunting, though. This advice works equally well for internal sales—selling to other areas of your company—as it does for external sales. If you’re trying to sell internally, there should be many opportunities to find out what causes other people in your company pain. People love talking about themselves, and many people in the world can’t pass up an opportunity to complain.
If external sales are involved there’s a good chance a team sales approach is being used or could be. In this case you will be there as the technical support to a lead sales person (or people). Let your lead sales person do what they do best—cultivate the relationship, get the lead, introduce your business, and get the customer excited.
Use the time to learn as much as possible, flush out where the customer’s life isn’t as easy as it could be, and match your analytic solutions to making it easy. If a customer’s life is painful enough, the remotest prospect of someone fixing it will have them reaching for their checkbook in no time.
For management consultants, an interview guide is an invaluable tool to get the best out of time spent with clients. However, it could be a little unsubtle and strange to use one with an internal customer, when you are trying to maintain an ongoing relationship with them.
That doesn’t mean that going through the process of preparing an interview guide won’t help you. Just the act of writing it out will mean that most of your questions are at your fingertips, as the process of developing the guide will force you to think carefully about the questions you are going to ask in the interview.
Closing
There is a lot of advice for salespersons on the art of closing the deal. A lot of it is too high pressured for data science sales where the targets are people who want to be sold to. Tactics favored by door-to-door salesmen are likely to have the wrong effect; in any case, if final persuasion is actually required with an external customer, this will likely be handled by a sales professional from your company.
More important from the point of view of a data scientist is that the details of what is to be done are as nailed as possible. Often there will have been a long lead time, where possibilities were canvassed. During the initial phases, this is often necessary, to convince a potential client that all problems can be solved. At the point when they sign on, however, it’s now important to narrow the focus considerably, to one solvable problem, which is also defined narrowly enough that a solution can be provided that is recognizably what was promised in the contract.
This has a double benefit, in that the smaller-sized project will be less costly in both time and money. Moreover, if you are doing work for another company, it’s likely that the people you are trying to convince will need to convince their own management, and a small, targeted effort will be more likely to succeed.
Promoting Data Science
Data science as a discipline has been enjoying a privileged position as the “in thing” for a few years now. This is a blessing and a curse. It is a blessing that people with the expectation that their problems can be solved or alleviated with data science are more likely to put their trust in data science solutions. It is a curse that people expect data science to solve their problems at the touch of a button regardless of whether there is reason to think that there is a good fit.
There is also an ironic aspect to this that there will be some people for whom the very fact that there are such glowing testimonies will cause them to be all the more skeptical. Maybe they are right—in recent times, data science has been near the peak of its hype cycle, and its intrinsic to a hype cycle that the activity or phenomenon being hyped receives more praise than its due.
The problem is that the hype becomes noise that obscures the most useful “real” part of whatever being hyped’s benefits. So many fanciful claims get made that they obscure the sensible and practical claims. Therefore, there is as much a need for careful promotion when something is already overly hyped as there is when it isn’t.
Careful promotion, however, is something that is clearly very different from hype. Rather than talking up pie in the sky claims for the way that data science and artificial intelligence (AI) could change the world, careful promotion means putting the case that data science is a useful tool when applied to the right problem. This means being realistic about the occasions when data science is not the right tool as well as extolling it when it is.
Anyone you meet by now will have heard about the way that the machine learning and AI is going to change the world, and many of them will have made up their mind whether they think what they have heard is either complete twaddle or a prophecy that is as important to humankind as the second coming.
The Double-Edged Sword of the Hype Cycle
Everyone has heard quotes about data science. “Statistics is the sexiest job of the 21st century,” and all that. The hype cycle for data science has been surprisingly long-lived, and in 2018, Gartner3 wrote “the hype around data science and machine learning continues to defy gravity.”
Dealing with hype can be a problem in many walks of life. An article in the LA Times4 describes the difficulties that young athletes face when they receive too much praise early in their career. In fact, this was part of the difficulty faced by Billy Beane of Moneyball fame that has been linked to his not living up to his potential when he was a player.5
The hype in data science doesn’t occur at an individual level in the same way. When people in data science are singled out as particular rock stars, it tends to be after they have established track records. The problem for data scientists is more that the reputation of an entire industry precedes them.
There are two contradictory symptoms when your audience has been exposed to too much data science hype. The first is a jaded response to anything you say; the second is an audience member with inflated expectations.
The remedy is the same in each case—reframe the discussion somewhere new to force them to forget what they’ve already heard and reconsider the topic from scratch.
Hence, when discussing a business problem that might be solved using data science, the key is to look at the problem from a different angle than the angle most often used in the hyped approaches to data science. A particular angle to avoid is “the power of big data” or the idea that effectively a big enough data set renders other considerations meaningless.
Instead, draw attention away from the kind of big sky thinking that’s stereotypical of machine learning hype from the concrete gains that can be made within your organization. Avoid trying to find parallels with the large tech companies, such as Google or Uber.
Instead focus on moderate and realistically reproducible gains made by smaller-sized companies. Find use cases from within your audience’s industry, and stick with those examples. Ensure that the use cases you pick are relevant to the business problems faced by your audience and have gains measurable in dollars, increased sales, reduced effort, or a similar metric that is relevant and understandable for a business-oriented audience.
If you are stuck for finding such examples, a book such as Eric Siegel’s Predictive Analytics6 which promotes data science is a useful resource. It’s the kind of book whose only mission is to tell the world that data science is the best thing since the transistor. As a data scientist already, you might, therefore, not need this kind of book. However, the real-life examples are a great resource when planning a talk or pitch, especially when talking to an audience in a less familiar industry.
While there are numerous entries in the “data science is awesome” category, Siegel’s offering is notable for its case studies over a wide range of industries—a special section in the first edition boasts 147 examples of predictive analytics applications.
By sticking to this plan, you will be sure, at a minimum, to give the audience what they are really looking for. However, if you are successful at providing a relatable example of data science in action, you will be able to make the audience member who thinks about data science in an unhelpful way reconsider how they think about the benefits of data science in their situation.
The best way to approach people who are tired of the hype is to bring things back to basics. You can demonstrate that you can be trusted—that you aren’t one of the snake oil salespeople—by talking about data science in measured terms to explain its real benefits.
Branding—Personal and Shared
Arguably, data science has evolved from being a descriptor or classifier for a particular occupation, and become a brand. Of course, any label for an occupation becomes a brand at some point. Lawyer and medical doctor are terms that in addition to describing the kind of work their bearer does also evoke various stereotypes about what kind of person would do that kind of work.
Data scientist, as a label, has been around long enough to do the same thing. However, “data science” doesn’t have the advantage of a professional body such as lawyers, accountant, or engineers have. Apart from setting professional standards and facilitating networking, the professional bodies for each of these professions have also taken on the task of marketing their professions.
As the “sexy” profession of the 21st century, data science has an advantage over other professions. However, these advantages are more obvious when it comes to attracting new entrants to the profession than they are when it comes to persuading people to listen to data scientists’ advice. In recent years, we have seen professions such as accounting and actuaries build marketing campaigns to expand their professions’ reach in relation to the areas that they can provide value.
In the latter context, the care that the professional bodies of professions such as actuary are putting into building their brand has the potential to reclaim some ground lost to data science, both in the context of recruitment and the context of users taking actuarial advice.
With no central body, these kind of campaigns don’t occur for data scientists, who then have little control over their profession’s image, which is often conflated with AI and big tech, meaning people with data science who don’t fit that mold may find they struggle to meet people’s “expected” notion of what a data scientist is.
To counter this, individual data scientists need to both be ambassadors for the data science brand, and to pay careful attention to their own personal branding. In each case, you need to have a strong concept of the image you are trying to project.
The oft-quoted idea that a data scientist is “someone who codes better than most statisticians, while knowing more statistics than most programmers” is unhelpful for this purpose, as it doesn’t relate to how a data scientist can be useful.
The occupational branding used by the accountants and actuaries reflects their value proposition a lot better. The equivalent for data scientists might be that a data scientist will reveal what your data is telling you, or a data scientist is someone enables you to get the best value from their data.
Based on these labels for what a data scientist is, there is a grain of truth in the “programmer/statistician hybrid” conception of a data scientist. A data scientist is someone who straddles the divide between two worlds, however, the programmer/statistician divide isn’t the important divide. Instead, the important divide is the divide between business and technical, and a data scientist is one of the groups of people who straddle that divide.
Once you have established your own definition of what a data scientist is and worked out how to market that definition within your organization and your wider network, you can extend and deepen that definition into your personal brand.
You are both an example of a data scientist, yet also other things besides being a data scientist, so tighten your data science definition to establish what kind of data scientist you are while at the same time defining a brand for yourself that includes your non-data science attributes.
This means that the relative lack of a professional body pushing a standard concept of what a data scientist is, and why they are useful is a double-edged sword. Professionals who have a professional body determining the brand are less able to decide their professional branding to suit their own strengths and weaknesses in the same way.
Moreover, older professions, such as accountants and actuaries have found that half the branding battle is to fight against a preconceived idea that doesn’t suit them—in fact, both of those professions have had to fight against a stereotype of being boring.
Having decided on a definition of data science that plays to your strengths, you can move to building your personal brand around it. While a lot of data scientists might be the kind of people who would like the quality of their work to speak for them,7 the reality is that you can’t leave others impression of you to chance.
Don’t Shoot Yourself in the Foot
Before we plunge too deeply into this topic, I’d like to mention that if you do a Google search on “personal branding,” a lot of what you will read will be advice around your personal digital marketing strategy. That is, they talk about how to improve your LinkedIn profile, or how to use Twitter effectively to get a better job. These are important issues from one angle, but the kind of personal branding I will discuss revolves more closely around the way you are seen by your co-workers and clients and therefore has more to do with how they are able to see you act.
As the previous paragraph makes clear, your personal brand is established whenever others encounter you directly or indirectly. In contemporary life, that often means via social media, but it wasn’t all that long ago that individuals who weren’t celebrities mostly encountered each other either directly or at one or two removes via someone else who had encountered the individual directly. Hence, you built your personal brand mostly by the image you projected to people who were in the same room as you.
This is still true within your own organization, where people see you in meetings, in the tea room, and at your desk with great regularity. What does what people see in these interactions say about you? Does the way that you arrange your desk project the image you are trying to cultivate? Co-workers, your boss, and your boss’s boss all see the way that you keep your desk—is the impression you make, the impression you want to make.
On the other hand, perhaps your boss has access to an executive kitchen, so she never encounters the dirty dishes you leave in the sink. Even so, there are still plenty of opportunities for you to damage your brand in your interactions with her.
This stuff can seem trivial, and in a sense it is, but it’s the prerequisites for building your personal brand. If you don’t get those parts right, all people will remember is that you’re a slob. Having set down those foundations, you can move on to establishing yourself as a trusted advisor within your organization. All in all, it’s hard to get people to take you seriously, so don’t lose the battle before you begin by failing something obvious.
Opt Out of the Boxing Game
A natural tendency for people meeting for the first time is to define themselves by what they do for a living. We often answer the question “What do you do?” by stating their job title or occupation name. However, for most possible audiences, a bald statement of “I’m a lawyer” will do a poor job of explaining what it is the speaker does to earn a crust.
On the one hand, most people are aware of a difference between a defense attorney and a corporate litigator, and many are probably aware there are lawyers who spend a lot of time in court and others who’ve never set foot in one. At the same time, most people don’t have a strong idea of what a lawyer does all day.
Worse than that, when you answer “I’m a lawyer,” the person asking puts you in a box, and when you ask the same question and hear “I’m an accountant,” you do exactly the same thing. This is a boxing game.
In the case of data scientists, your audience will lean toward the “no idea what you do all day” end. Answering the question “what do you for a living?” with “I’m a data scientist” could result in a blank look. Worse, you’ve lost control of the impression you are making—they’ll put you in a box according to their stereotype of a data scientist.
This could be okay if they got the memo that data science is the sexiest job of the 21st century, but it could also mean they assume you meet all their worst stereotypes of a software geek or statistician and then some. Crucially, you won’t know which one they landed on.
A better question to answer is the Steve Jobs question “What did you do for [our company] today?” Answering this question rather than the literal question posed by the person you ran into allows you to develop your own narrative of how you help your company.
You aren’t even bound by the strict wording of Jobs’ question (which was designed specifically to discover the most recent thing a hapless employee had done for Apple)—you are free to answer it based on the most impressive thing you have done over for your company over any timeframe you choose. By reframing the question in your head before answering you’ve taken control of the opportunity to promote your brand.
To be effective in this arena, you will need to have the answer on the tip of your tongue—to some extent, this means developing an elevator pitch for what you do for your organization.
Although the idea of an elevator pitch is frequently associated with entrepreneurs, most often in pursuit of either sales or finance, the basic concept can easily fit to the task of explaining how you contribute to other workers in your business or potential clients.8
The key to a successful elevator pitch is preparation. You need to have a clear idea of what you are trying to achieve, so write down your goal before writing out the first draft of your pitch. Then practice saying it loud to get an idea long it goes for, and to get it to flow nicely. You’ve got to get in and out within around 30 seconds to succeed, and you’ve got to have it on the tip of your tongue to work properly.
Remember to record yourself a couple of times and play it back to make sure you’ve got a decent intonation. Try to avoid speaking in a monotone or speaking too fast, or other problems that mean that you sound like someone reciting a shopping list rather than explaining themselves in a natural voice.9
The idea of the elevator pitch is a simple one, and its application doesn’t require particular talent, only some effort. However, used in the right place, it can be a highly effective method for explaining how you fit into the big picture, and how what you do benefits the greater good (while giving the freedom to define the greater good to your own advantage).
At the very least, you will have a better answer when you are asked, “What do you do for a living?” at a party than “I’m a data scientist.” More than that, by making you focus on how you make a contribution instead of trying to label yourself, by preparing an elevator pitch you will have a firm foundation for use with any other kind of personal branding exercise you want to consider.
Earning Trust
The highest state you can aspire to in a relationship with the people you are trying to assist is to be trusted. Trust is what ensures that your organization comes to you early with problems that need your help to solve, ensures that they give your solutions a fair hearing and implement them without trying to prove that they won’t work first.
Earning the trust of your customers and clients takes the idea of a personal brand further. Although trust is earned repeatedly within each new relationship, it is a deeper level of engagement than having a personal brand and provides a firmer foundation for future work.
Just as for the previous sections, this needs to be done with an eye on promoting people to trust your team and to trust data science as a profession that will aid their company. Trust in either yourself as an individual or trust in the idea of data science by themselves won’t be sufficient to ensure they trust your solutions to their problems.
In a later chapter we will see that ensuring that users trust your models is of paramount importance for them to be used. If your users are able to trust you on a personal level, or the model is presented to them by someone they respect who trusts you on a personal level, you will have crossed an important barrier for achieving your user’s trust in your model.
At the same time, it is more difficult to get your clients to trust you than it is to get them to trust your model—it’s just that if you are successful in getting them to trust you the pay-off is that you need to re-earn that trust every time you want to propose a new model.
Building that trust is something that will take place over a longer period of time, with a different approach. The point here is not that you provide models that do what they say on the tin by attaining a good accuracy measure or even that the models you provide develop a reputation for significantly improving the organization’s smooth running.
The famous book on building trust for consultants, The Trusted Advisor,10 defines four ingredients for personal trust—credibility, reliability, intimacy, and self-orientation. A lot of this book has to do with the first two, as these elements can apply to statistical models as well as to humans. The last two really only apply to people.
More generally, the authors of The Trusted Advisor are keen to point out that although credibility and reliability have a black and white, technical dimension (which is what I have covered in relation to models in other chapters of this book) they also have an important emotional dimension, which will not map so easily onto models, but is important for your personal credibility when trying to win over your clients or potential users.
Intimacy can be seen as more difficult to achieve than either credibility or reliability. To be credible, you just need to know your stuff. To be seen as reliable, you just need to deliver what you say you will, when you said you would—on the face of it, it’s entirely in your control. To establish some degree of personal intimacy with others requires accepting the risk that you will be rejected.
The final ingredient is an inverse relationship to trust—the authors of The Trusted Advisor call it “self-orientation,” and the less oriented toward yourself your clients perceive you to be, the more successful you will be at gaining your trust.
At least, if you appear to only take your interests to heart, you will find it difficult to win their trust. The Trusted Advisor authors could have called this factor “altruism” or “awareness of others” to avoid the inverse relationship.
Although some ways you can be overly self-oriented can be obvious, such as only caring about the paycheck or the kudos of solving the important problem others have already failed at, others are more subtle and pernicious.
Many of the latter could trip up a data scientist—“a desire to be seen to be right” or “a need to look clever.” This book was written before smartphones became ubiquitous, so the easiest and quickest way to be seen as paying insufficient attention to your client in 2019 is missing—“phubbing,” or looking at your smartphone while talking to others.
- 1.
Engage: Show the client they have your attention.
- 2.
Listen: Show the client you understand their problem.
- 3.
Frame: The root issue is identified.
- 4.
Envision: A vision of an alternate reality is sketched out.
- 5.
Commit: Steps are agreed upon.
Some of these steps could be seen as overlapping or complementing the design thinking process—envision could be seen as another way of saying Ideate. In some ways, Chapter 1 of this book could be seen as being a guide to engaging and listening.
As mentioned earlier, the element that the treatment in The Trusted Advisor emphasizes that is often lost, and likely to be the more difficult for data scientists is the emotional component over the rational component. Some paraphrasing of their advice on listening can illustrate the idea.
The authors identify a number of types of listening, and they also identify the message that the way that you listen sends your client as at least—possibly more—important than the information you get from the client by the process. Listening, then, is an opportunity—yes to learn—but more than that to show by your demeanor and body language that you are on your client’s side and that you care that she succeeds.
As mentioned previously, listening can come in a number of types, and the types listed in the book related more to how the listening process affects the person who is being listened to than they affect the listener. For example, “supportive listening,” as would be expected, is listening that makes the listened to person feel supportive.
However, essential to all the types of listening that are presented is the need to avoid interrupting the speaker’s flow and allowing them to present their story as they seem fit. This may often feel difficult to do, as so often we think that we have heard the story, or one like it, before, and rush to jump in. It takes practice to avoid doing that.
The honest truth of this is that there is a limit to how much of this can be explained in a book. The best way to learn how to do it is to practice and accept that you will fail but also to accept that when you fail it’s not the end of the world.
Pushing yourself to push forward with the emotional parts of your arguments, even if it’s scary, will enable you to solve more problems with data science, as you will be able to win over more people. By starting small and ensuring your sales pitches target their real needs, you can grow in your clients’ trust and become one of their highest valued partners.
Summary
Selling yourself, your model, and data science itself are all fundamental to being a data scientist. Unfortunately, it is not simple to persuade people to use your models, to persuade them that you are to be trusted, or that data science is a valid approach to solving their problems.
Part of the difficulty is that there are often multiple sets of people you need to convince to be allowed to both work on a solution and later implement it. Their needs are likely to be different and what you will need to do to convince them is likely to be different.
The first part of the journey is convincing your clients that your model meets their needs. Building trust in a model is a little less onerous than building trust in a person but some of the elements are the same—a model needs to be credible and reliable. To be credible, the model needs to be understandable, and you need to ensure that your client understands why your model is the solution to their problem.
Selling individual projects is one part of the picture, but marketing yourself and your profession as a data scientist is also important. Data science, for a few reasons, including its relative novelty, the hype surrounding it, and the lack of a specific body representing the profession, is arguably a blank slate from a marketing perspective.
This can be a good thing or a bad thing, as it means that both the responsibility and the opportunity for explaining data science to potential customers and clients is in your hands. Either way, a data scientist needs to re-orient new clients or customers to their own understanding of what a data scientist is, establishing moderate and achievable benefits of data science with their audience.
Your personal brand is another crucial element required for success with clients. This is true both for consultants working with external clients and data scientists working on problems within an organization. There are myriads of techniques suggested for improving personal branding, although some are more applicable to data scientist’s situation than others.
An example of a technique with wide applicability is the elevator pitch, and the exercise of focusing on what your personal brand is and distilling to its most important elements to develop an elevator pitch is a great platform for other personal branding activities. For the data scientist, it also helps you ensure that whoever you’re talking knows what you mean by data scientist when that phrase means many things to different people.
Being able to achieve a relationship with your client to the point where they trust you will ensure that they not only implement your solutions but also seek your advice early. However, this requires a big commitment, and the determination to not let feeling uncomfortable or the possibility of embarrassment interfere.
The Trust Equation suggested by Maister, Green, and Galford says that human consultants need to be credible, reliable, able to achieve (sufficient) intimacy with their clients, and able to convince their clients they have their best interests at heart. Although the concepts of intimacy and selfishness can’t be applied to a model, credibility and reliability can be.
We will explore more specific ways of ensuring that your model is both believable and understandable in order to ensure credibility in the next chapter. In the chapter after that, we will examine ways to ensure your model is reliable by keeping it performing at as close to the way it first performed during testing and validation as possible.
Sales Checklist
Does your presentation of the model create a link between your model and the solution of the customers’ problems?
Have you considered the emotional aspect of your presentation?
Do you have an “elevator pitch” answer to the question “What do you do?”
Have you thought about your own definition of data science, and how it solves your customers’ and clients’ problems?
How are you building trust with your clients? What would your clients say if someone asked what kind of person you were?