data interpretation
January 6, 2023

How much focus do you give to Data Interpretation skills?

By Martina Pugliese

With so much focus on the technical skills to access & manipulate data, how much do you work on data interpretation? Is this crucial element being overlooked by a focus on data & modelling?

My conversation with a recent podcast guest reminded me how much the term ‘insight‘ from data is widely misused & misunderstood. I’m also aware of the demand for training that many analytics teams are keen to improve their insight generation or data interpretation skills. Not just the storytelling aspect, but also spotting what is most important & meaningful.

So, I was delighted to find an article written by two experienced data professionals, Selby Cary (from Testcard) and Martina Pugliese (from Shopify). They kindly agreed that I republish their wisdom here in a short series. So, here is part one of a two-part series on data interpretation, why it matters and 4 lessons to unlock it (first 2 in this post)…

Data interpretation – the missing element (first 2 lessons)

Over the last 10 years, the importance and influence of data in our lives has become unmistakable. You may have heard phrases like “data is the new oil” or buzzwords such as “algorithms” or “machine learning’’ — and, that’s the simple ones! Whether you like it or not, data is part of our daily routines — from social media to online banking — powering everything from electric vehicles to healthcare systems.

There is no doubt that the increasing availability and usage of data will continue to change our lives. New products appear on the market nearly every day, making it easier to consume goods or services, effectively spawning new industries or fundamentally altering existing ones. Streaming services (such as Netflix or Spotify) have shaken up the entertainment industry by selling subscriptions to a near-unlimited supply of personalised content, all fueled by algorithms and viewership data.

To transform raw data into an asset, and maintain a healthy or valuable workflow, it must be relevant and relatable to the task at hand. As such, interpretation is essential — no matter the amount of data or analysis method used. For example, recommending the same movies to every Netflix user would inevitably result in poor retention rates. Tailoring a product, based on user input or usage history, allows companies to provide a more enjoyable experience.

In this series, you will learn to challenge your critical thinking and look past the raw numbers by questioning and contextualising the information presented to you. This will help you build more valuable, effective and ‘sticky’ products’ that scale.

Get to the Point — What are the lessons?

  1. Data is more than ‘Stats’ (see below)
  2. Interpretation and Context are Everything (see below)
  3. Empower People through Data Literacy (next post)
  4. Exercise your Critical Thinking skills (next post)

Lesson #1 — Data is more than ‘Stats’

Working with data isn’t really a new thing. Its rise to fame arguably started with the cultish HBR article entitled “Data Scientist: The Sexiest Job of the 21st Century”. You could say it’s the avocado of tech due to its recent popularity and historical foundations!

About 25 years ago, data science was the new kid in town and a very shiny one. Early adopters, such as supermarkets, were still trying to understand their customer’s behaviours and began experimenting with purchasing data. The success of the Tesco Clubcard (~28% growth in one year) sparked the introduction of the ‘loyalty card’ by the likes of American Express and British Airways — revolutionising everything from marketing to product development.

Today, we are so immersed in services that scour our data, so much so, that most of us take it for granted. Our data is used to build new products, inform choices, give recommendations, measure sentiment, and a whole lot more. This information boom has been accelerated by the affordability of cloud-based data processing services (such as AWS) —allowing us to store and transmit more data each year than ever before.

Exhibit A = just the statistics

It’s estimated that by the end of 2022, the world will have created 94 zettabytes of data. For context, the global data volume was only 2 zettabytes in 2010 (according to the IDC). That is mind-bogglingly fast (~38% CAGR)!

Chart by Our World in Data

Statistics like the ones above are great, but they only tell part of the story. Using numbers alone to make decisions can be dangerous and misleading, especially in the startup world — where mistakes don’t come cheap.

Be more scientific

For instance, focusing solely on one metric, such as new App Downloads, whilst ignoring others could be costly in the long run. If your user acquisition is a direct function of advertising spending but your churn is unsettlingly high — you have a serious problem on your hands. Only by asking the right questions and understanding the root cause (e.g. the 5 whys method) can you truly grow sustainably.

“The core word in data science is not data, it’s science” — Jeff Leek

As the loyalty card example illustrated, understanding your customers is a much better strategy than blindly chasing numbers. Unfortunately, data science is not a magical recipe for increasing revenue, retaining customers or resolving society’s many problems. To unlock the potential of data, we must think like a scientist and look beyond vanity metrics.

Lesson #2 — Interpretation and Context are Everything

No matter how great your data is, what actually it represents can be misunderstood without the right interpretation and contextualisation. Put simply, it’s the data scientists’ responsibility to build solid data narratives, void of logical missteps, and it’s important for everyone to question any information provided in isolation.

The limits of our perception become painfully obvious every time we watch the news or scroll on social media. We are bombarded with big headlines and catchy promotional videos, focusing our attention on the object of the story. However, as the ‘Invisible Gorilla’ study showed, when we concentrate intensely on one thing, we cannot see what is right in front of us (even if it’s a dancing gorilla).

As you’re likely beginning to notice, data is not necessarily neutral. It often comes with a whole load of baggage — constraints, limits and biases. Sometimes data biases are unintentional, resulting from unsound logical foundations or incorrect assumptions. More concerningly, data biases can be deliberate, choosing methods of application for an intended purpose.

The danger of coding biases

The world has already seen the harmful effects of data bias in (some) machine learning/AI systems in our societies. To cite just a few examples; the use of facial recognition by the police without considering human rights, algorithms treating gender and skin types unequally or even issues dealing with language.

As the Netflix documentary “Coded Bias” highlighted, certain machine learning systems have been unfairly trained on one demographic data set. To simplify an incredibly complex issue, some algorithms detect lighter skin tones better than darker complexions. This unfair representation of human diversity can alienate certain groups and have nefarious long-term consequences if left unchecked.

The real problem arises when we attempt to use raw data to power insights without reflecting on it first. What does the data contain? Is it appropriate for the task at hand? And, most importantly, what are the implications? Simply having access to data does not imply that we can build something useful. We need to understand, analyse and close any gaps in our assumptions before we can be sure of the outcome. Even the fastest sports car without the right tyres can lose a race.

Which lesson will you apply & where can you learn more?

Many thanks to Selby & Martina for that post. I hope it helped all our readers think about the importance of focussing on this area. In their next post, they turn to how data literacy & critical thinking skills can close the gap.

While you wait for that next post, I should also say that the authors have been kind enough to recommend some further reading on this topic. They suggest you read “Scary Smart” by Mo Gawdt. You can find out more about that book here:

Scary Smart | Mo Gawdat

The Future of Artificial Intelligence and How You Can Save Our World eBook and Print Versions Are Now Available Technology is putting our humanity at risk to an unprecedented degree. This book is not for engineers who write the code or the policy makers who claim they can regulate it.

Enjoy the love of learning this year & protect some time to hone your data interpretation skills.

(Originally published on Selby’s Scaleup Lessons blog on Medium.)