My evolving feelings about data

In God we trust. Everyone else must bring data.

I stared through the thick windowpane at the central quad. It was a spring morning of my senior year of college; one of the last sessions of my final class.

The chalky blackboard showed algebra equations. Some say that a math degree is a signal to employers. I've also heard that it "teaches you how to think."

I don't know about that.

I do know that I've never directly used my undergraduate math degree.

What I did get from my math degree is the certainty that Some Things are True. Mathematicians start from basic axioms and write proofs of theorems. Generally, if you believe the axioms, you must believe the theorems.

"Out in the world everything is messy. I'll never know things for certain after this," I thought glumly.

I stared at the branches gently swaying on the quad.

***

I started working with data more regularly later that summer.

I worked as a research assistant for an economics professor. His office had a sign with a quote that roughly read: "Don't expect understanding to be linear. If today you discover that your methodology is incorrect and you have to start over, that's still a day of progress."

I cleaned and analyzed a dataset about household income. I took so much pleasure in manipulating the data.

It was basic stuff, and it was thrilling. I'd find that New York zip codes had higher median income than rural Tennessee. It's not that I doubted that it was true, I was just shocked that data reflected reality.

***

After college I spent several years at an analytical consultancy.

My confidence in manipulating data grew. I became proficient in Stata, SAS, SQL, and Python. I learned techniques to clean and analyze data, to identify insights, and to tell a visual story.

The work was particularly valuable when we brought a data-driven lens to a decision that was historically made without data.

I felt like with enough of the right data I could do anything. I didn't just enjoy data. I trusted it.

My confidence grew also because data felt less personal: it's not me, it's the data.

***

Over the next few years, I began spending more time presenting to clients.

Somehow, when I was the one "pressing send," I felt less sure. I checked the model again; reread the charts.

I also began to come up against some of the limits from using data:

• If the most urgent question isn't one we had data for, then we had little to offer
• Many ideas look good in the data but have other practical limits that can't be seen there
• Once there is momentum around a specific direction, organizations generally cherry-pick the analytical results that support it

I developed a more nuanced approach. Data is a collection of observations about the world. The better the data, the more accurate model you can build of the world.

But it's not magic and it's not comprehensive. There are other ways of knowing other than segmenting and aggregating observations.

***

At 麦肯锡 , data was used heavily in building recommendations for clients. However, at times I sensed a slight insecurity about heavily analytical projects. Organizationally, it was viewed as "an input." Technical analysts were treated as service-providers for the consultants who "owned" the client relationship and had the full context and business judgement.

For many of the old guard, "data" (or, better yet, "big data") was a marketing buzzword and a way to build client-impressing charts.

For me, I became disillusioned. On the one hand, I had spent enough time in the trenches and I understood that every analysis has messiness and shortcomings. On the other, I wasn't generally deep in the data myself so I couldn't build comfort with results. And, finally, I saw how data was often abused to support the popular narrative.

I used data for my needs, but didn't trust it any longer like I used to.

***

At McKinsey, a trendy attitude was: "you can get data to say anything."

I didn't think about it that much, but in retrospect I chalk it up to many consultants' inflated belief in the credulity of clients.

Personally, I think you can lie with data, but it's generally harder than lying without data. A savvy audience can easily rip apart misleading or sloppy analysis.

***

Today, data is an important part of my job, driving product strategy for Business Messaging at Meta .

With that, my feelings about data continue to evolve.

When deciding what to build, it's important to look at existing behavior. This is especially true at companies with scale like Meta - if the behavior is valuable, someone, somewhere is doing it.

However, data is not the only way to make decisions.

Data is about observing.

Wisdom is about experiencing directly.

Intuition, fantasy, and art are about envisioning something new; something, perhaps, that generates new data that has never existed.

All those years ago when I thought that math would be the last thing I know for sure - I was wrong. The biggest Truths that we learn in our life are learned based on experience not data - that time is finite; that we love what we love; that hugs feel good.

And data is extremely important, but it is not The Only Thing.

To quote Elliot Eisner on the value of art: "despite the cultural bias that assigns literal language and number a virtual monopoly on how understanding is advanced, the arts make vivid the fact that neither words in their literal form nor number exhaust what we can know"