Image by Author
#Introduction
I’ve worked in the data industry for over four years. During this time, I’ve witnessed a seismic shift in the role.
Previously, when screening candidates for data analyst positions, it was easier to tell who possessed the technical skill to do the job and who didn’t. Nowadays, almost everyone gets through the first few rounds of the data analyst interview.
With AI, candidates who have very little practice are building dashboards and writing SQL queries with precision — skills that once took years to learn. As a result, employer expectations have changed, with seniority and domain expertise becoming somewhat of a hard requirement.
Additionally, the lines between different tech roles are getting blurred, and seniority is becoming a prerequisite. Employees are expected to take on more projects, learn more skills, and produce more output in a shorter period of time. However, it isn’t all bad news.
From my experience working in the field, I believe that there is a specific type of data analyst who will not just survive, but thrive and withstand the AI revolution.
And that is a…
product data analyst
In this article, you will learn:
- What a product data analyst (PDA) is
- The difference between a PDA and a traditional data analyst
- The skills required to become a PDA
- My own experience working as a PDA at a large tech company
For a video version of this article, watch this:
#What is a Product Data Analyst?
To illustrate the difference between a PDA and a regular data analyst, let’s consider a “day in the life” of professionals in each role.
//Day in the Life of a Traditional Data Analyst
Brian is a traditional data analyst. He secured an entry-level position and has now been working for one year in this role.
Here is what Brian’s job consists of:
- Pulling last year’s sales figures in SQL and using it to build a dashboard
- Identifying a segment of customers who stopped transacting with the company in the past 3 months
- Figuring out why conversion rates dropped by 15% last week by combining data from the organization’s customer relationship management (CRM) systems and marketing channels
A role like this typically requires knowledge of SQL, Excel, building dashboards, and some programming skills. My first data analyst job required me to do tasks that were exactly like this. It certainly isn’t easy. But AI is lowering the barrier to entry to do these jobs.
Every task mentioned above can be performed much faster using AI tools like Cursor, Claude, and ChatGPT.
Due to Brian’s heavy usage of AI tools, he sometimes starts to feel like he’s becoming more of a prompt engineer than a data analyst. Brian’s employer realizes that Brian’s job can be done faster due to new AI tools. Due to this, they stop hiring other data analysts. Instead, they get Brian to complete all the data analytics projects. While Brian has a stable job and is a more efficient analyst thanks to AI, he sometimes feels like his responsibilities haven’t changed that much since last year. He isn’t climbing the corporate ladder or getting promoted.
More importantly, Brian wants to learn more skills and increase his depth of knowledge, rather than just using AI to do the same job faster. Brian is on a traditional data analyst path. This isn’t a bad thing, but by simply repositioning himself and learning some additional skills, he can climb the corporate ladder quicker and make more money. To do this, Brian must use AI as leverage rather than competition.
//Day in the Life of a Product Data Analyst
Sarah is a PDA at a social media company.
Here’s what her job looks like:
- Sarah works with the team that builds reels to understand why creators in specific locations are less inclined to use this feature. She then works with the design team to build new features to bridge that gap.
- She works on a new “creator boost” feature, to understand whether boosting new creators on the platform leads to better creator retention without impacting user engagement. To do this, she runs an A/B test. (Spoiler alert: This kind of analysis isn’t easy. Results are rarely straightforward and they are even more difficult to explain to stakeholders).
- Sarah also sits in product review meetings and challenges assumptions made by leadership: for instance, the VP assumes that users want longer videos, and Sarah needs to debunk this with actual behavioral data showing attention drops after ~35 seconds.
Do you notice the difference between Brian and Sarah’s jobs?
Sarah’s job isn’t necessarily more technically complex than Brian’s is. Both professionals share the same set of technical skills; they both know SQL, can write Excel formulas, and build dashboards.
The biggest difference between their jobs is that Sarah has a lot more influence on product decisions. If the new “creator boost” feature is rolled out and the company makes $1M from it, Sarah has directly contributed to over a million dollars in product revenue.
As a result, she is of high value to the company and gets promoted easily, with higher salary increases.
#How Can You Become a PDA?
I have worked in both traditional and PDA roles. For the first two years of my career, I worked as a traditional data analyst. And I now work as a PDA.
Here are the skills you need to become a traditional data analyst:
- Excel
- SQL
- Some programming skills (ideally Python)
- Data visualization
- Statistics
To become a PDA, here’s what you need to learn on top of the core data analytics skills:
//Skill 1: A/B Testing and Experimentation
You’ve probably heard of A/B testing before. If you have a website and want to know which would get you more clicks — a blue button or a green button — all you need to do is conduct an A/B test.
First, you select a sample of users who visit your website, and then randomly split them into two groups. One group will be shown the green button, the other will be shown the blue one, and whichever gets the higher number of clicks per user will be launched.
The above example is the simplest way to conduct an A/B test.
There is a lot more that goes into experimentation, like making sure you’re choosing groups with equal distributions, and ensuring that your A/B test has sufficient statistical power.
I was asked a ton of questions during my PDA interview, which I was able to answer thanks to Udacity’s free Intro to A/B testing course.
//Skill 2: Defining Product Metrics
One more thing that product analysts do that differs from traditional data analysts is defining success metrics.
To understand what this entails, let’s consider the new “creator boost” feature mentioned previously. When you boost new creators, it typically makes them want to post more on the platform, leading to higher retention. This improved retention rate is exactly what platforms like TikTok and YouTube want, as it keeps users on their platform for longer.
But… What is considered a new creator? Someone who has posted their first video? Posted 5 videos on the platform?
Additionally, after the creator gets their initial boost from the program, what if subsequent posts get far lower engagement? Could this lead to even higher churn in the future? Is this actually worse for long-term retention than not releasing the feature at all?
Also, as a social media platform, viewer engagement must also be taken into consideration. What if viewers use the platform less because they simply aren’t interested in being recommended new creators?
A PDA needs to take all these factors into consideration when creating success metrics. To measure the success of a new feature like this one, the product analyst can decide to create multiple success metrics, such as:
- Short-term creator retention rate
- Long-term creator retention rate
- Viewer engagement rate
In PDA interviews, you will typically be provided with a use case like the example I illustrated above. The interviewer will then ask you what success metrics you would define for this use case and why.
To learn the skill of metric definition, I recommend the following resources:
//Skill 3: Event Tracking
Let’s say you’ve defined a success metric. For this new “creator boost” feature, your success metric is creator retention rate.
Now, you need data to actually create this metric using app events such as creator uploads and clicks. You would typically build metrics with SQL. Sometimes, however, you might want to track a metric and realize that an event you need is currently not being captured.
For instance, if your platform currently doesn’t track “upload” events, you have no way to know how often creators upload content. Since you don’t have this event captured, you cannot build out your success metric (creator retention rate). You then need to work with engineering teams and explain to them which events must be captured so you can effectively track the success of the product.
To learn about more event tracking, I suggest reading this article.
//Skill 4: Applied Statistics
This is a skill that data analysts already have.
As a PDA, your focus must be on applying statistical concepts using programming tools.
The following skills are the most relevant to the role of a PDA:
- Hypothesis testing.
- Statistical significance: The difference between statistical and practical significance, effect sizes, and p-values.
- Causal inference basics (confounders, treatment effects).
- Simpson’s paradox and selection bias.
Khan Academy is a great place to learn these concepts; just type the name of the topic in and watch the video tutorial. I typically learn the theory behind a topic from a site like Khan Academy. Then, I would go to ChatGPT and have the AI platform teach me the practical application of the statistical concept on a real dataset.
#Where Can You Find PDA Jobs?
Facebook, Amazon, Apple, Netflix, and Google (FAANG) and other large tech companies hire a lot of PDAs because they launch new features and conduct A/B tests every day.
Meta usually has job titles like “Data Scientist, Product Analytics” or “Product Analyst.” These jobs pay between $249K and $382K, whereas traditional data analyst roles pay $180K to $282K. This is a pretty massive salary discrepancy, and tells you just how in-demand PDAs are.
Other than FAANG, you should also look out for:
- Fast-growing startups that build user-facing products.
- E-commerce companies.
- Fintech and healthcare tech companies.
Additionally, companies aren’t great at coming up with data-related job titles. A single job title can mean multiple things in the data space.
In fact, I’ve seen companies hire for PDAs under the following titles:
- PDA.
- Product analyst
- Product data scientist
- Data scientist, product analytics
- Analytics manager (product-focused)
- Growth analyst
Some companies will just post “data analyst” or “data scientist”, when in fact, it is a PDA role. I suggest reading the job description of a data role to understand whether it is a product-facing job.
Typically, the job description of a PDA role will mention keywords such as “collaboration with product managers,” “A/B testing,” “working with cross-functional teams,” and analyzing “product metrics.”
#Key Takeaways
We have covered a lot of ground in this article. Specifically, we learned:
- What a PDA role entails
- Why you should become a PDA
- How to learn the skills needed to become a PDA
- Where to find PDA jobs
Remember, AI is changing the way we work at a faster rate than you can possibly imagine. In this era, you must make yourself more marketable with skills that complement AI, not compete with it. This is exactly where PDAs come in.
Since these professionals add so much direct value to the company’s bottom line, their jobs tend to be safer, with higher salaries and quicker promotions than regular data analyst positions.
Natassha Selvaraj is a self-taught data scientist with a passion for writing. Natassha writes on everything data science-related, a true master of all data topics. You can connect with her on LinkedIn or check out her YouTube channel.