7 tips to make your data analysis more robust

Increase the confidence in your results and build a stronger personal brand

Jordan Gomes
Towards Data Science

--

Photo by Ameer Basheer on Unsplash

Imagine: you have been working hard on this huge project. You just finalized the deck — at this point, this is no longer a presentation, this is art. The visualizations are sublime, the story is simple and easy to follow, and the recommendations are logical. The big day is finally here, and you start presenting to the key decisions makers. Your opening joke is a blast, and everyone is hooked.

But halfway through slide #2, someone raises their hand to let you know that they have seen wildly different numbers on a different presentation someone else did a couple of weeks ago. Horror! Your flow is broken, your palms are sweaty, knees weak, arms are heavy — you know how the song goes.

In some companies, this can be quite common. There might be multiple reasons for this: a poor knowledge management system, a lack of collaborative tools for data analysis, a lack of quality checks, etc. To avoid this happening to you, here are 7 tips:

#1: Non-technical subject matter experts are your best friends

  • Non technical SMEs can shed a different light on the phenomenon you are studying. By going through live examples with them, or by asking them to explain some data points, you can get a peak into their decision making process, what ‘feature(s)’ they value, or what they consider to be expected or unexpected.
  • They can also give you a lot of context around how the general market is evolving — and how general business decisions can be impacting the data you are seeing (e.g. market seasonality, revenue increase thanks to a new product launch, etc.). These elements can make your study even more comprehensive.

#2: Most likely, someone somewhere already worked on something similar

  • Doing an internal “literature review” allows you to understand what has already been studied, and who studied it. It can help you jumpstart your study as you piggyback on the work that has already been done and help you find a peer that already looked at the same phenomenon. Once you have identified this peer, you can share knowledge, and potentially use them as a brainstorming partner.
  • It can also allow you to better position your study. If you know what has already been studied in the past, you can make sure that you are not reinventing the wheel and you are actually answering something that has never been answered before — and for which there is more appetite.

#3: If it is too good to be true... it is most likely too good to be true

  • If you get a result that goes against a popular belief, assume you’re wrong and double/triple check your queries and data sources. If everything checks out, have a peer review it too.
  • Once you make your claim, your work should get a lot of visibility — unexpected results often do. There will be a lot of scrutiny, so you want to leave no stone unturned, as you don’t want to be that person who started a lot of turmoil from an incorrect result.

#4: If you want to go fast, go alone. If you want to go far, go together

Although it is “cool” to arrive in a room and blow everyone’s mind with a completely new & disruptive idea, in real life, this is rarely the case. It is always preferable to bring your partners into your study. By doing so:

  • You ensure that they have a good understanding of where you are coming from and how you got there.
  • They can help steer your study towards a place where it can generate greater value.
  • They can help shape the story you are telling to make sure it resonates with them and their management.
  • You can get their buy-in earlier in the process, and they can directly think about how to operationalize your findings.

#5: Every study has its holes — find them

Just like you wouldn’t go to a big sporting event without training first, you shouldn’t go to an important meeting without at least one rehearsal. Finding someone you trust and asking them to challenge your work can go a long way. Not only can it help you prepare your story, it can also help you understand which arguments/data points are weak. You can even go one step further and ask them to destroy your work and nit pick on everything they can find — this way nothing will surprise you on the day of the presentation.

#6: Don’t let your mouth produce words faster than your brain can approve them

  • If you end up presenting and someone starts questioning your results, it is better to clearly understand the issue first, do your own research, and then come back to the person offline.
  • Trying to answer number discrepancies live might lead you to make incorrect assumptions that you will have to rectify later — which can damage your credibility over time.

#7: It is never that simple

  • The scenario mentioned in the introduction above often comes true in what first appears as a simple and straightforward study.
  • When the ask is very straightforward, why bother doing a literature review or recruiting cross-functional partners or having your study reviewed?
  • Unfortunately this is where mistakes are made — and that will hurt your internal credibility.
  • It is always important to take a step back and understand the general context of the ask. If it is not clear how the data is going to be used, or who it is going to be presented to — it is better to err on the side of caution

Insure your personal brand today!

Following these tips will reduces the risk of you being wrong, and losing your organization’s trust — which is the most important thing for a data analyst.

Do you have any tips to share on how to reduce the risk of being wrong? Let me know in the comment section!

Thanks for reading!

If you had ‘fun’ see my other articles:

Next up

In the next article I will be talking about Bootstrapping — and why I love this methodology!

--

--

Head of LCS Analytics @Snap/ ex-YouTube. Analytics, Content, ML & everything in-between. Opinions are my own - https://analyticsexplained.substack.com