How to create more impactful analyses

❝

This article is for🔒️ paid subscribers only 🔒️, but all subscribers can continue reading a free preview of the article.

The most successful B2Cs didn’t grow virally. As Lenny Rachitsky shared in an article, virality is a myth. What they did was grow consistently with quality.

Looks linear to me.

The same can be said about careers. A few have temporary virality, but on average careers are linear. If you’re a data analyst or scientist that means to have a long lasting and growing career you need to produce quality consistently.

Quality without consistency = unreliable

Consistency without quality = unpromotable

How can data analyst or scientist produce consistent and quality analyses that drive your career?

You have to P.A.C.E. yourself.

The P.A.C.E. framework

I’ve produced 100+ analyses and reflecting on what went right vs wrong, there are 4 major milestones in an analysis. Many data analyst or scientist focus on just the analysis, but the battle is actually won before you start. After you deliver an analysis, you have to continue to follow through or your analysis won’t be remembered.

Let’s dig in.

Prioritize

One of the biggest challenges data analyst or scientist have is prioritizing work. We look at analyses as problems to be solved instead of tools to improve the business.

“Why is revenue down Y%?”… ooooh that sounds like a fun problem to dive into when our first instinct should be “OH F*CK, revenue is down y%. That doesn’t sound good!”

Our education systems don’t really prepare data analyst or scientist for the reality of professional work. Nowhere in high school or college are we ever told “Here are 5 problems. Pick the 3 most important ones and you’ll be graded based on which problems you pick and your answers for them”. However, that’s exactly how it works on the job. You’re graded on which problems you solve and how well you solve them.

That’s why the first step of the framework is prioritize.

Here is how you prioritize:

Get in the room with your stakeholder.
If your only interaction with a stakeholder is through a Jira ticketing system, then good luck. You’re just a data monkey to them. Instead, set up a weekly or biweekly check in. Understand their top priorities and what they’re seeing in the business.
Once you have a few ideas of what’s important, do a quick impact analysis. Do back of the envelope calculations to see the estimated business impact of what you can work on. “If I work on X, this could lead to $Y in revenue”. It’s okay if you’re wrong at first because over time you’ll get better at this. The exercise is a forcing function to maximizing your impact per analysis hour.
Estimate the effort you think it’ll take for you to deliver the analysis. I prefer the t-shirt sizing methodology (Small, Medium, Large).
Small = a few days
Medium = a week or 2
Large = a month +.
In almost all cases, throwout the ones that take a month or break them down into smaller analyses. The reason is because the business doesn’t have the patience to wait that long. If you told a stakeholder it’s going to take a month, they’ll bug you the next week and say “okay can I get it next week”.
Pick the analysis with the highest impact to effort ratio and add a buffer of a couple extra days to the estimated effort. Now start to analyze

Analyze

There are 5 key reasons an analysis flops:

the data was wrong *
it’s not actionable
it took too long
it says too much
it says too little

this one is just a doozy. Hard to avoid until it’s too late.

I’ve fallen into all of these traps myself and coached data analyst and scientist through how to avoid them. An analysis has to be long enough to be interesting but short enough to be engaging. It has to be actionable yet timely while the business problem is still a priority.

How do you do this?