A redditor recently asked us how to break into causal inference when their current company doesn’t offer opportunities to work on it. This is a common roadblock for data professionals trying to build expertise in a new area, so we’re sharing our response publicly for anyone facing the same challenge.
Question 1: Do you think it’s feasible to implement these solutions (mostly experimentation and causal inference) without a senior specialist on-site? What would a realistic roadmap look like for someone starting from scratch in terms of bringing this type of resource to the company?
Yes, definitely feasible! What matters is knowing which assumptions need to hold for you to draw causal conclusions.
Here’s the thing: every piece of information you extract from data relies on assumptions. Even counting registered users in a spreadsheet assumes the data is complete and accurate. Being explicit about assumptions is what makes causal modeling anti-fragile.
You can apply causal inference ideas without really doing deep statistical work at all! Simply thinking about causal relationships, counterfactuals, and assumptions is a great place to start.
A realistic roadmap
Start with statistics 101 coupled with data science. A solid foundation for that is Practical Statistics for Data Scientists: 50+ Essential Concepts Using R and Python (see here).
For experimentation specifically, learn from applied theory. Trustworthy Online Controlled Experiments (see here) will get you up to speed. Ron Kohavi also offers courses from basic to advanced levels on Maven, which I recommend (see here and here). Not an advertisement, just genuinely useful.
But causal inference is much more than experimentation. You’ll want to learn tools like difference-in-differences, synthetic control method, and others. Some intuition-first books for that: Causal Inference for the Brave and True (see here), Causal Inference in Python (see here). More classic but accessible: The Effect: An Introduction to Research Design and Causality (here) and Causal Inference: The Mixtape (here).
If you still want to read about causal inference in your spare time after all that, two books that changed how I see things: Experimentation Works: The Surprising Power of Business Experiments (here) and The Book of Why: The New Science of Cause and Effect (here).
Question 2: How would you recommend building ‘marketable’ expertise if I can’t practice these at my current company yet’
I understand that struggle. I’ll point out two ways and let you build the right mix for your reality.
1st way – Internal evangelism: Propose applying causal inference, the causal way of thinking, and causal inference tools in your work. Evangelize some people about it. Create a Slack channel about the topic, share the materials I recommended, and spur discussion.
At worst, you can say during a future job interview that you created a causal community at your current company. And at best, people will become increasingly interested and you’ll have buy-in to start applying these in your projects.
2nd way – external portfolio building: There’s plenty of data on the internet you can use to build cases for your portfolio. For instance, the book Everyday Causal Inference (here) comes with realistic data you can use to build report-like documents with your own causal analyses. Make analyses using these data and upload the reports in your Github. Matteo Courthoud’s awesome causal inference (here) repository also brings a lot of use cases.
Worst case scenario, you can ask Claude, Gemini, and similar tools to create business cases for you so you learn to apply these tools and build your expertise. Just remember to get your analyses out there, so use Github and eventually consider sharing your learnings on LinkedIn, Medium, Substack and other social platforms, so people get to know you as part of the field.
Question 3: What would you expect from a candidate who has the theoretical knowledge and interest but lacks professional experience in Causal Inference or MMM? What tasks should they be able to
Real-world expertise is valuable because with real problems come not only the possibility of applying the concepts of causal inference, but also many violations that textbooks don’t include. So professionals end up learning how to deal with messy reality.
But for non-senior positions, it’s okay not to have that expertise. Being able to understand and apply the concepts under common situations is enough.
What I’d expect from a candidate: Based on many selection processes I’ve been through and learned from, I’d expect the candidate to understand:
- How to ask relevant causal questions and translate them into relevant business questions
- The assumptions that need to be met so the question can be answered as intended
- The limitations that these assumptions impose on the methods
- How wrong we’ll get if these assumptions are not met
- How different tools try to solve the same problem (e.g., Meta’s Robyn vs Google’s Meridian)
In other words: knowing what question you’re actually answering, what needs to be true for that answer to be valid, and what breaks if those conditions don’t hold. That’s the core.
