Stop Worrying About Your Tools — Start Thinking About Enabling!

Photo by Christina @ wocintechchat.com on Unsplash

Today, have you checked your data science and analytics newsletters, social media groups, and medium blog posts? Have you dug into your data science bubble and liked and commented on some posts about the newest rocket science technology? How often did you scroll over yet the next deep learning, image processing, Python, or dashboard building tutorial?

Technical solutions have always fascinated, so it is not surprising that data science blogs and publications have a high proportion of design patterns, introductions to frameworks and application examples for a variety of technologies. From the reader’s point of view, it seems promising: One quickly reads into a new technology and can use it to solve concrete problems at work or in a hobby project. Maybe you think something like this: “The old MySQL database is very inflexible and doesn’t scale anyway, with Snowflake we would solve so many problems and make our projects more successful.” But is this even true?

In this article, I want to share some thought-provoking ideas to stop thinking about technology only and start minding how to enable your stakeholders. I will start with discovering the root cause of failed projects and introduce the concept of empowerment. With this information, I invite you to think again about your “toolbox” and present some ideas on better preparing your work for teammates without a data science/analytics background. Mind that this is just a thought-starter and doesn’t go too deep in any of these concepts. Also this writing will have a focus on the analytics workload.

Why projects fail

Let’s use your daily habit of reading through technical essays and articles and claiming that you are doing so because you want to become a better analyst, data scientist, or developer. You want to increase the quality of your work and be a well-oiled cog in your company’s bigger wheel, which leads to successful projects. But have you ever asked yourself why projects keep failing?

In their famous book Peopleware — Productive Projects and Teams [1], DeMarco and Lister wrote that 15% of small and 25% of big projects fail. Horner quotes Gartner with a failure rate of up to 85% (sic!) [2]. When asked why they failed, project members answer “politics” in most cases. DeMarco and Lister translate this to the project’s sociology. They even found that not a single technical issue explained the failure of most of their studied projects.

Now ask yourself, if many projects fail because of ‘politics’, then why are you reading about Python frameworks and deep learning algorithms? The answer is: It’s much easier to broaden your technical skills than it is to fix your unspoken conflict with Nicole from marketing or have some honest discussion with your manager Dave, about his impulsive behavior.

What is empowerment?

While a lot can go wrong and even more can help, I found it very helpful to shift my mindset from a technical to an interpersonal domain. If you stop thinking about the tools, just for a minute, ask yourself: “How can I help Nicole with her daily struggles in A/B testing?” or “How can I help Dave find the balanced level of communication to work effectively with him?”

Those thoughts can enable you and others to work together more effectively, thus raising your projects success rate.

But before we talk about enabling, I want to introduce another concept called empowerment. Jaffe and Scott [3] describe empowerment in an organizational context with three core factors:

- Employees who feel responsible and act as active problem solvers.

- Teams that work together to improve their performance and productivity.

- Organizations that are structured to help people achieve the results they want.

Furthermore, “Employee empowerment involves the delegation of decision-making authority to lower levels in the organisational hierarchy, with employees provided with the autonomy to make day-to-day decisions about job-related activities” [4].

And suddenly, there are a lot of decision-making and action-taking everywhere in the company. That said, you’re not just building a dashboard or a statistical model for your manager or a decision maker higher up in your companies hierarchy. It’s you developing a system for Nicole’s A/B testing so she can decide on her ad claims even faster and without much knowledge of statistical hypothesis testing.

That’s enabling!

However, this does not happen by itself. As mentioned, as an employee you have to feel the responsibility and be an active problem solver. That means you need to understand a lot more problems from many stakeholders in your company, so let’s break that down a little bit.

The toolbox revisited

When you think of your skills and tools, you probably think about your ability to deploy a Kafka pipeline or to create an advanced Tableau dashboard, but take a step back and revisit the big picture.

Suppose you are not thinking from the perspective of your stakeholders. You are likely to fall into that tempting trap where you try to find the proper nail for your hammer. In a situation like this, you think from your tooling and skills (hammer) and are very tempted to solve only problems (nails) which fit them. You are then not as helpful as you could and do not have a mindset that enables others to do their job at their best.

In Figure 1, you see four types of analytical capabilities. While there is a lot more around it, especially for data scientists and engineers, this will focus on the analytical part of the ‘data job.’

Figure 1: Four Types of Analytical Capability [5]

One of the first considerations is to reflect on your actual skills from your stakeholders’ perspective. They care less about you using threading in Python or your deep insights into Bayesian statistics. They want to decide or take action, and they want your help. If you can map a question like “What happened to my A/B test?” to an analytical solution approach like descriptive analysis, that’s already half the work done. With this mapping, you can even go up to Nicole and tell her that she doesn’t need to ask herself what she should do with your A/B testing dashboard because with a little prescriptive analytics you can automate the whole process, and she only needs to push a new version of her claims and images into the pipeline now and then.

Mind! You are not thinking about how you do that and what tools you need, but only what she needs. After that is clear, you can go on and find the right (technical) tool in the box.

Clarifying mechanisms and limits

As said, halfway done with understanding the questions, but now we need to prepare the answers, and here it is essential to help your stakeholders understand your findings. Seiter [6] describes four phases of his analytical process: Framing, Allocation, Analytics, and Preparation (see Figure 2). While this process has a lot to offer, it’s the last phase, Preparation, which plays a vital role as the interface with stakeholders. It’s divided into three parts: clarifying the mechanisms, determining the range of validity (limits), and visualizations.

Figure 2: Business Analytics Process Model. Translated from [6].

Clarifying the mechanisms of your analysis is essential for you and the domain expert, aka the stakeholder you work with. You should explain your model and approach to persons without a data-science background, and let them ask critical questions - maybe even invite them to do so. It should be apparent where the results came from for them (and you). If you enable a domain expert to understand your model, she has the chance to tell you that this model will not represent reality. You could, for example, have used linear regression for some correlation, but the expert tells you that in real life it’s exponential. And this is where the limits of validity come into play. Your model maybe only working in a specific range, and the correlation in this range shows almost linear behavior, but if you had widened the scope, you would have been able to see the exponential nature of this correlation. Your expert teammate just helped you become a better data scientist. Not only did you enable her to understand your model, but she also did so with you.

Two easy ways for simpler visualizations

One of the main ways of communication is a proper visualization, and here enabling means helping your stakeholders understand your points without too much obstruction, and in a visual language, they are familiar with. There are books about it, so I’m only pointing out two key issues. If you are interested in a deep dive, I recommend “Storytelling with data” [7].

One of the easiest strategies is to remove everything from your visualization that didn’t support your story. This can be an axis, numbers, texts, and too many colors. Often a simple slope graph explaining the story of growing or decreasing behavior is enough.

Another point is that not every professional understands every form of visual representation. The box plot is an excellent example. It visualizes key statistical indicators like the quartiles, the interquartile range, and the median. These are not accessible to everyone without explanation. Either this is provided or, another visualization should be used.

Here too, think from the viewpoint of your stakeholder, not from your data and analytics. For example, using a logarithmic scale or a box plot might be tempting and technically the right choice, but that is worth nothing if the audience can’t understand it. If your visualization can’t transport the story, you should think about splitting it up and using descriptive and more accessible language.

Conclusion

We now took a shallow dip into empowerment and enabling in a not too theoretical and philosophical way — at least, I hope so. We also discovered a slightly different view on how to think about our skills and tools and then took an extra round on preparing our work.
While this is just a little reading on those topics, I hope your mindset has shifted slightly away from a technical focus to more people-centric thoughts. Projects keep failing because of the human factor. So if you want to increase your success rate and that of your organization, keep reading about these topics. If you want to share your thoughts or widen my perspective, please leave a comment. I also have to admit, that this is one of my first writings. Should you have some tips on improving my output, please don’t be shy — just contact me.

Bibliography

[1] DeMarco, T., & Lister, T. (2013). Peopleware: productive projects and teams (Third Edition Ausg.). Upper Saddle River, NJ: Addison-Wesley.

[2] Horner, P. (12 2019). Why analytics projects fail and other success stories. OR-MS Today, 46(6). (I. f. Sciences, Hrsg.) Linthicum. doi:http://dx.doi.org/10.1287/orms.2019.06.13

[3] Jaffe, D., & Scott, C. (1991). Empowerment : A practice guide for success. Menlo Park: Course Technology Cris.

[4] Baird, K., Su, S., & Munir, R. (2018). The relationship between the enabling use of controls, employee empowerment, and performance. Personnel Review.

[5] Gartner. (21. 10 2014). Gartner Says Advanced Analytics Is a Top Business Priority. Von https://www.gartner.com/en/newsroom/press-releases/2014-10-21-gartner-says-advanced-analytics-is-a-top-business-priority abgerufen

[6] Seiter, M. (2019). Business Analytics: Wie Sie Daten für die Steuerung von Unternehmen nutzen. München: Franz Vahlen.

[7] Knaflic, C. (2015). Storytelling with data: A data visualization guide for business professionals. Hoboken, New Jersey: John Wiley & Sons.