When should you avoid hiring a data science specialist?🧑🎓
Hello!
How are you taking in the frenzy of the US Election day (week, perhaps)? After hours on TV and restless browser refreshes, I finally called it a night! Hoping for saner and calmer weeks ahead.
This newsletter will take you about 5 minutes to read.
I. Spotlight: When should you avoid hiring a data science specialist?🧑🎓
You’re the CEO of a mid-sized US retailer. Your organization uses data for basic reporting. You are inspired by the promise of data science and want to transform your company into a data-driven enterprise.
What skills do you need in your initial data science team?
When posed with this question, there are 3 popular roles that leaders often consider. Which of these roles do you think should be hired first?
Data Scientists: Machine learning (ML) experts who can identify and build the best data science solutions, hands-on.
ML Engineers: Technical leads who can package their domain solutions by integrating them with ML APIs or on low-code platforms.
Data Architects: Experts who can drive an organization-wide study to discover data needs in order to build a robust data warehouse.
The right answer is None of the above! Let’s see why a specialist will make a bad choice.
(Picture by Andrew Toskin, CC BY-SA 2.0 via Wikimedia Commons)
Do you need a specialist or a generalist?
Your organization’s reporting capabilities give you access to simple data summaries that can answer questions such as:
‘What was my revenue? Which product was the most profitable? Which channel grew fastest?’
Most leaders assume that they must next get into predictive analytics in order to get strong value from data. They try answering questions such as, ‘What will be the sales for my top 5 products next month?’
There’s nothing wrong with this question. It’s just the timing.
The Pareto principle states that roughly 80% of outcomes come from 20% of the causes. On a similar note, early on in your data journey, most of your business value will be delivered by a small set of skills that you will build.
First, let’s see the two types of analytics that map to the most urgent and impactful business questions:
a) Deeper descriptive analytics: These are questions that find out what happened in specific scenarios: ‘Which competitor beat me in the mid-west, in my top product category?’
b) Diagnostic analytics: These are follow-on questions that help you understand why it happened: ‘Why did my online sales dip during Thanksgiving?’
What skills do you need to perform these analytics? The key to success is going broad, not deep.
What skills will make a “data jack-of-all-trades”?
Your team needs a strong business understanding. They must gain some familiarity with data handling and curation, exploratory analysis, basic statistical analysis, simple charting, and good communication skills.
How do you build such a team?
Scout for potential hires within your existing organization. Identify employees with strong business knowledge and an aptitude for learning - people who aren’t afraid to experiment. You don’t need to hire a Ph.D. in Artificial Intelligence, yet.
Find people who have the curiosity to ask probing business questions, rather than those who can build fancy applications.
Train people on simple tools such as Excel, Google Spreadsheets, R, or Python, rather than on Deep Learning techniques.
Put them through online courses and workshops to train them on the above skills, rather than trying to get them certified as Data Scientists.
What about the specialists?
Yes, you will eventually need the specialists. In my whiteboard video, I had covered the 5 skills every mature data science team needs, as organizations scale to the next level.
However, to get your journey started, and to be successful, you first need generalists.
What was your team’s skillset when you started data science? What lessons did you learn? Reply and let me know.
II. Industry Roundup:
1. DeepMind is working on AI that could make Nobel winning discoveries
44 min | Exponential View | Azeem Azhar
“We are getting to a point where powerful algorithms can accelerate scientific discovery by themselves”, says Demmis Hassabis, founder & CEO of DeepMind. He shares how the innovations behind AlphaGo are repurposed for consumer and enterprise products.
2. How Companies can build Analytics agility to withstand disruption
6 minutes | MIT Sloan Management Review | Lori Bieda
How can analytics help organizations respond to the pandemic? Increase your investments in data, plan a tighter integration of analytics and business in terms of skillsets, processes, and solutions. Check out the article for these recommendations.
3. Latest Industry Trends in Natural Language Processing (NLP)
by Ben Lorica and Paco Nathan
NLP is a hot area of application in data science. This industry survey found that investment in the space is growing. Check out the report to find the top NLP use cases, best libraries, popular cloud services, and industry trends.
III. From my Desk:
1. Video & Slides: Saving Lives by Applying AI to Satellite Imagery
21 minutes | Cities Rising Summit by Tom Tom Foundation
Dengue kills over a million people every year. Here’s how AI and geospatial analytics are helping control mosquito-borne diseases. I present a demo of our work done in partnership with World Mosquito Program and Microsoft.
-> Check out the Video and Slides
2. Interview: The unreasonable effectiveness of data
9 minutes | Diffbot DataBytes
In this chat, I talk about how we got started at Gramener, what excites me most about data, my biggest frustrations about data science, and why data solutions should become invisible in order to be successful.
3. 6 Cool Productivity Tools
5 minutes | Enterprisers Project | Carla Rudder
Do you ‘feel’ busy but get less work done? Wonder what happened to all your time? This article shares useful tools to track, save, and find additional time. Includes my recommendation.
Yours,
Ganes.
PS: Did someone forward this to you? You can subscribe here.
My Website | LinkedIn | Twitter | YouTube