Data Science Explained by Notitia Principal, Ariel Pilcer
When we asked Canberra-based Notitia Principal, Ariel Pilcer, to explain how data science is applied to extract value, we didn’t expect him to send us a link to what he made for dinner. Read on to find out about Ariel's kitchen analogy, or if you’re just here for the lamb shank recipe - it’s at the end.
Notitia Principal, Ariel Pilcer is an experienced solution architect and analytics delivery lead, with a background in mathematics and data science. Ariel has driven solutions for clients in government, defence, aviation, banking and retail. He thrives on using data to transform the way we live and seeing the impact it has on organisational decision-making.
Get in touch: ariel.pilcer@notitia.consulting or book a chat here.
Hungry for insights
Before we get too hungry, let’s quickly define data science.
Data science is the analysis of both structured and unstructured data, through different methodologies, algorithms, systems, and tools.
The goal is to extract valuable insights.
These are then used in data-driven decision making across diverse sectors such as government, health, business and various industries.
Some of the outcomes include informing policy and refining public systems, enhancing profitability, innovation in products and services and optimising infrastructure - the list goes on.
The “oven analogy”- Training cooks or oven builders?
Consider the following analogy:
Oven = Data science methods and models
Cooks = Data scientist
Ingredients = Raw data
Meals = Insights
To make the best meals, of course, it helps to have fresh ingredients and a good oven.
But what makes a good cook?
- Knowing which meals the eaters want, and
- Knowing what ingredients are needed.
What does a cook need to know about an oven?
- They must know what it does (it gets hot)
- They must know how to operate it (where to put the food, what dials to turn), and
- How to identify the right settings for the desired outcome (temperature, function, time).
There is a lot that a cook does not need to know about ovens.
A cook knows how to use an oven, but not how to design, fabricate, calibrate or install one.
Our language easily distinguishes oven users (cooks, chefs) from workers involved in oven production (designers, fabricators, technicians).
Data science has not been around for quite as long as cooking.
We tend to refer to people who use data science models AND people who build models as data scientists. Heck, we’d probably call people who research models and derive new techniques as data scientists as well.
Overcooked data scientists?
In her well-known talk Cassie Kozyrkov, Google’s now-ex Chief Decision Scientist, argues that data ingredients are plentiful and fresh, data science methods and computational power ovens are accessible, but our current approaches for embedding data science acumen in organisations tend to “overcook” our data scientists.
We emphasise the statistical and algorithmic first principles upon which teaching models and methods rely, rather than how they should be used.
As a result, we are producing data scientists who know the mathematics and processes that underpin data models, and can perhaps can construct them from first principles - but are not equipped to identify and apply the most appropriate method for a given problem.
An output-focused, hands-on approach, is far more valuable to produce the right type of data scientists, than a rigorous academic approach to learning. Jacques Pepin and Gordon Ramsay apprenticed in kitchens, not oven factories!
The view from the dinner table
That was the view from the kitchen. Now, Ariel tells the story of his dinner last week from the perspective of his dinner table:
There are five rumbling bellies.
Right now, no one gives a damn about ingredients, cooks or ovens.
They care about lamb shanks and potatoes, and very little else.
One belly is mine. I made the dinner. I know the recipe and I know I’ll like it.
My wife is eating too. If I hadn’t cooked, she’d have been able to cook this dinner for herself.
My 5-year-old is glad it’s ready, but doesn’t want to hear about the process.
My 2-year-old wants me to take out the chickpeas.
My 9-month-old needs it to be pureed because he’s only got a few teeth, and I’ll have to choose between spoon feeding him or wiping mash off the walls.
Similarly, data scientists. and analysts must understand that the work does not end the moment the data lands in a table. Our job is to also ensure that our product can be readily consumed by end users - who may have very different data appetites and palate maturity.
The role of "data" in the kitchen-to-dinner table analogy Ariel's data science recipes
In Ariel's kitchen-to-dinner table analogy, the food follows a one-way, linear flow.
The food is first bought. Then cooked. Then presented. Then eaten.
The process is not iterative or bi-directional (perhaps unless you have toddlers).
Data, however, is consumed differently.
The act of seeing and analysing data may prompt new questions, requiring additional processing, investigation and analysis.
Users may wish to drill down further, or run alternative scenarios. Sometimes the insights they need are context dependent.
This often means creating new flows - the equivalent of allowing users to send their food back to the kitchen for a tweak from the chef, or adjust it themselves from the table, or even send it through a whole other kitchen.
Ariel's data science recipes
Aside from cooking a mean lamb shank and creating a quick-witted meme, Ariel has an extensive background in mathematics + data science, with experience across government (defence, employment, finance, agriculture, transport), aviation, healthcare, retail, banking + publishing.
A Canberra local, Ariel most recently consulted for Australian Government Department of Defence, Australian Government Department of Employment and Workplace Relations + global tech giants Qlik and Teradata.
Ariel has also developed some interesting integration recipes in his time:
The Teradata Sandwich: R on Teradata on R
The side dish: Qlik with R Server-Side Extension
The jelly snake: Control-M workflow with Teradata-Python
The nut cluster: Qlik R Websocket Arduino
If you’d like to hear more about these integration recipes or about the work that Ariel does as Notitia Principal, he is always happy to share!
And if all this talk has made you hungry, here’s a recipe for lamb shanks.
Get in touch: ariel.pilcer@notitia.consulting or book a chat here.
