Как исследователю данных выжить и пробиться в корпоративной политике

1. Get Executive Ownership

One of the key contributing factors for any project is getting executive buy in. It is your job as a data science program manager, project manager or data scientist, to get your executives to believe in your project. Without their ownership, and funding, your project will not go on. When executives see that your data science project will help drive their strategy. They will be all in.

2. Be bold, tell your manager they are wrong…with data

Managers and executives are humans, they can only process so much information at once and see so many correlations. You, as a data scientist have access to the know how and tools to process 100x that amount of data and accurately look at thousands of correlations. Use that knowledge and boldly show your manager they are wrong. No, we don’t mean be a dick about it. We mean, support your manager. Help him or her go into meetings with the right conclusions. They will thank you for it. However, this requires you tell them they are wrong.

3. Gain the trust of your peers

As we travel around helping different teams. We realize that many managers don’t even trust their data. Yet, they want new dashboards, data science teams, the whole nine yards. What is the point though? If you can’t even trust your data. Our favorite quote from Sherlock Holmes talks about how data is the foundation for the building blocks of thinking. If that is true, and you don’t trust the house you have built. It will fall on top of you. Get your managers to trust you and your data!

4. Successfully implement a simple project first

Look we get it, everyone wants to develop the next Google or Facebook algorithm. Why not, they are cool, super powerful and rack up billions of dollars a year. However, if your team is just starting out and you want them to succeed start small. Don’t worry, even something simple, if done well can provide your executives with insurmountable value. Once you get that first win under your belt. Executives will be begging you to help them on everything. Then you will have to work on making sure your projects are bombarded by requests all the time, or at least, only the right projects are being worked on.

5. Communicate the value of your project

One way to get buy in from executives, be a salesmen. How, tell them why they need the project, create the need. Data science is still new and many executives don’t know why and where to use it. Show them! That is your job! Show them how to take data science and turn it into saved money, resources, etc.

6. Standardize your data science processes

Data science has a lot of cool technologies and tools that allow for great insight. However, like software engineering, even with all the cool things you can do. Without processes, you can fall behind projects, make poor products and fail to maintain finish projects. This means you need to document your processes. It seems like a waste of time, until you start having internal breakdowns of projects. So make sure you have amazing data science processes early on!

7. Don’t Box In your Data Science Teams

Don’t limit your data science team to what you know. Challenge them, ask them what they think, ask for their input and don’t hold them back. Let them know you support them and trust in their abilities. Data scientists are smart people, they just need to know you believe in them(like tinker bell).

8. Plan, Plan, Plan..but don’t plan too much

If there is anything we have learned from all of our different experiences and projects. You need to plan. This ensures you don’t go out of scope, this makes sure you get a good handle of all the data sources, and requirements and it ensures you succeed. However, things change quickly in business today. So you can’t take 1 year to plan a project. When a project gets in the pipeline, you need to jump on getting requirements. Just don’t spend all your time planning and never developing.

9. Play well with other departments

Every business is a team sport. You have accounting, finance, operations, sales and all the other departments that your team needs to work with. They all usually have their own data warehouse and you need that data! If you are lucky, there is one central team that manages all the databases. Even if that is true. I Still need to get the expertise from multiple teams. In addition, all those teams will probably want to have some requirements on your projects. So make sure to play nice.

10. Learn from as many SMEs as possible

Like we mentioned in the last point. You will want to gain as much expertise from all the different departments as possible. Data scientists aren’t pharmacists, or doctors, they aren’t accountants or financial managers. We need to gain some insight about the business or subject from those who know it best. When starting a new project, make a list of the topics and data you need and seek out those SMEs.

11. Remove company bias

However, even though you should get SMEs opinions. Don’t allow their bias to block new insights. It happens all the time. Executives, managers and other team members believe that business drives have always been “XYZ”. Then, your team comes in with a totally new result, but instead of bringing it to managers. The team buries the new insight because it would go against status quo. That is not your job as a data scientist. Your job is to challenge status quo! But be right!

12. Challenge Status Quo! Be a Rebel

Look, as a data scientist. You have data on your side. That means, when you are right, you are right. We don’t mean be a jerk. We just mean, don’t be afraid. Don’t let your managers or executives back you out of your opinion. In all honesty, they want your opinion. They want you to give them information that they can go to their bosses with and stand confidently. At the end of the day, your boss has a boss. Guess what, they feel the exact same way you do when they talk to them. So tactfully challenge your bosses opinion using data!

13. Use data to drive initiatives

We believe you are seeing a theme now. Use data! There is so much power there. It isn’t even a new idea. People have been using data forever to prove things. Science relies on the methodology of repeatedly proving theories with data. Even ones that we consider true today. Do the same with every initiative. Why are you doing it? What is driving it? It better be data.

14. Build a prototype first for early buy in

How do you get early buy in? Build a prototype(sure, in python)! Show your team, and your manager what it can do. People want action, not just theories and words. Set up a prototype, if you can, get real data. If you can’t then pump it with some data but make sure the functionality is there. Make it tangible, interactive, and actionable!

15. Design for robustness and maintainability

We can’t stress this enough. Make sure whatever dashboard you build, process you set in place, or algorithm you develop is maintainable. If you leave the company tomorrow. Will the project still work, or will people curse your name. Seriously! People will if you left behind no documentation, and never shared your code.

16. Work to automate yourself out of boring work

Stop doing boring data munging, and QA manually. Just stop. When you first design your system, make sure as much of the boring, basic work is automated. Don’t worry, your company will have plenty more data science projects when you are done. You will be much better off putting the basic stuff at the will of a computer than you having to spend 2 hours a week to upload data.

17. Get a Data Science Guide

There are a lot of data science consulting companies that will develop a data science guide of good business practices for your team. This will require they assess your team’s current status and work with them to realize where they could be more effective. Often times, this is skipped by most teams, so it is helpful to bring in outside help.

18. Write your own data science guide

Maybe you have an amazing data scientist on your team that can do both his or her work and develop a good hand book for your team. That means onboarding, coding practices, system documentation etc. If so, get them to build it. Trust us, nothing is more helpful than walking into a company and getting documentation. Then you can assess early on what is going on, develop a new solution and walk out fast. This saves your company money, and makes data science consultants happy.

19. Learn to creatively gather and classify data before you start

You know what is a bad idea? Having 20 analysts classify 50,000 data points for 1 month. This is a terrible waste of resources and finances. When you develop an initiative, try to develop a method of data classification that doesn’t require analysts. Try to crowd source public opinion, offer a service, design a new product, whatever it has to be. But try to avoid getting your teams involved in tedious work.

20. Build your systems to properly gather data first

Sometimes, you are lucky enough to work on a project that is the first of its kind. The system itself will be gathering data for your analysis in 6 months. Build it with the end in mind. Think about how you want to use the data, the other systems you want it to interact with, etc. Don’t merely build a functional system, and then add in the data gathering component as an afterthought.

20. Collect as much clean data as possible

Data comes from all different sources. You can get it from internal warehouses, external APIs and just about everywhere. Gather as much of it as you can, and make sure it is managed and clean.

21. Be a great story teller (Hans Rosling was and really still is one of the Best!)

When it comes down to it. We all have bosses that we have to convince we are right. As a data scientist, you have to do this all the time. Why, because you have a data backed opinion that could get really boring. You could start to tell your executives about the different percentages, and standard deviations and watch their eyes glaze over. Instead, you should develop an infographic, a presentation, anything other than a bunch of numbers. You want to elicit passion and emotion for your cause. To gain more of point one. More executive buy in, means more funding and likelihood of success.

22. Communicate the internal and external values of data science to management

Data science has the ability to affect customers and employees. Tell your managers how it will work and its value. This point is somewhat repeated through this list. However, the more angles you can communicate to management that your project is going to save money, the better. Data scientists love to talk about how their algorithms are calculating the probability a person is scratching their nose when they are scrolling through Facebook. However, the business teams only care how much money they can make from that knowledge. Otherwise…why?

23. Learn management’s processes

Your team needs to know how management works. How many committees does your project have to go through, how often do they meet, what do they like to see? The better you know how upper management works, the better you can help drive their processes in the right direction. Also, the easier it will be to get your funding and projects through oversight.

24. Understand executives strategies and their whys

Executives have their own politics going on, their own processes, and strategies. Many employees don’t even know what is going on behind closed doors. One of the things we push for is more open discussion between data scientists and their executives. Not just, but a cohort of executives. Once you have an idea what is going on at the higher level, it is much easier to start developing projects and programs that more closely align with that strategy.

25. Be able to explain to your managers and leadership your failures

Failures happen all the time in the world of data science. Especially in the data science world. Make sure you can tell your manager why, and ask for help when you need it. Don’t let a project sit half finished because you are stuck on some small problem that requires outside intervention. That makes everyone lose.

26. Seek outside intervention when necessary

Sometimes seeking outside intervention is necessary. This may mean hiring a consulting team, or hiring new employees. There are times projects grow, or there is a temporary influx the number of projects. Getting some temporary employees to meet timelines is not a bad thing. Spending a little on a project that could save millions makes sense.

27. Read just enough outside news to be inspired

Too much outside news can bog down the mind. It may cause a fear of falling behind, as your projects may not be where these other titans of industry are at. Don’t worry, just read enough to continue to inspire you to move forward, but not so much that you think you can never compete.

28. Question every project you do

If a manager comes to you with a project, even if it has support from other senior VPs or executives. Question it. Why are you doing it, who will it affect, how much will it save? You might be the one finding these answers too, but make sure you know. Otherwise, you might be working on a dead project.

29. Be Positive!

Corny, but true. It is really easy, especially if you are a data science project manager, to lose hope. Maybe the insights your team has come back with are not very valuable or maybe they haven’t found anything at all. Guess what, that is much more common than you think. Not every project will lead to instant success. Be patient and be positive. If your data is clean, and your data science practices are solid. Something will eventually shake out.

30. Make a decision, give an actual opinion

As a data scientist, you have power. You have data, that means you can make conclusions with confidence. Don’t forget that. Say things like:

  • The best decision would be to …
  • I propose we ..
  • I know that …
  • Let’s try solution x because ….


Data Scientist # 1

Машинное обучение, большие данные, наука о данных, анализ данных, цифровой маркетинг, искусственный интеллект, нейронные сети, глубокое обучение, data science, data scientist, machine learning, artificial intelligence, big data, deep learning

Данные — новый актив!

Эффективно управлять можно только тем, что можно измерить.
Copyright © 2020 Data Scientist. Все права защищены.