How to properly present a Data Mining project?

Building models and getting insights are job half done for the data scientist, presenting them to the audience is an art itself. See, how to approach the presentation after wrapping up the data science project.

By Peter Zvirinsky (Algolytics).

Being a data scientist or data miner (as it was called in good old times) is a challenging job, isn’t it?

One needs to know about data imports and formats because there is the crucial data preparation stage. Then there is important data selection stage, followed by the next essential algorithm selection part. Moving on to yet another critical phase: model interpretation. And finishing with the not-less-important model implementation stage.


presenting_projectA bad news is that passing successfully through all these important steps doesn’t guarantee the project success. You still need to present the results to your boss or client.

Here are some hints how not to fail this last crucial step. How to make a presentation of a Data Mining project to your client or supervisor?

1. Start with big picture


First, spend a few minutes to introduce a big picture of the project. The idea is to establish a common ground with your listeners. Summarize what was the problem that needed to be solved. What was the business task the project should address? What was the context, constraints and what resources were available (in this case, what input data you had available, what information you gathered before the project, etc.).

2. Overview of process


Explain what stages your project had. But remember to be brief on that; no need to explain all details of all tasks you had performed. Just make listeners aware the process was not a simple one (if it really was).

You can show what stage took you longer, if you need to iterate, for example because of new data came in. You can show a simple flowchart diagram to show the process.

3. Show the main outcome


Focus on this part of the presentation. Present real data example, if possible. For example how the predictive model calculates a score for a specific client.

If you are talking to non-technical people, avoid abstract and too technical terms (like: classification model, logistic regression … ). Better present real data ( customer no 1 scored 90% as potential churner; 20% of new customers had incomplete address data, 15% had the country of residence different than IP…). If you are showing a lot of data charts and tables, it’s good to underline what was the main point.

Ideally if you can translate the predictive model performance in cash. How much money was saved? How operational business costs has changed? What will improve in the business thanks to your model?

Consider twice if you need to explain technical details.

If you are a data mining expert you can be tempted to dive into topics like what algorithm you used and why, what parameters you tested, what data subsets you tried. However, consider if this information is important for your auditorium. You can move that for Q&A session, if someone asks and if there’s time left.

Next time you will be presenting a Data Mining project, remember those essentials:

  1. Start with big picture
  2. Provide process summary
  3. Show the main outcome (if you are presenting to business people, translate the outcome into $$)

Good luck!