Deadline: Thursday, July 8, 2016 6:00PM Pacific Time.
Submission:
- Please submit your presentation via a private message on Slack to Bob and Ivan.
- Present materials in class.
This is it! It's presentation time.
Whether during an interview or as part of a job, you will frequently have to present your findings to business partners and other interested parties - many of whom won't know anything about data science! That's why it's important to practice communicating clearly and effectively so that anyone can understand.
That's why your goal here is to create a 10 minute presentation that guides your viewers through the problem, original data and hypothesis, findings, and results. This is what you've been building toward over the past few weeks! You should already have the analytical work complete, so now it's time to clean up and clarify your findings.
Come up with a detailed 10-20 slide deck or interactive demo that explains your data, visualizes your model, describes your approach, articulates strengths and weaknesses, and presents specific recommendations. Plan for a 10 minute presentation with 1-2 minutes of QA; be prepared to explain and defend your model to an inquisitive audience!
Goal: A detailed 10-20 slide presentation deck that relates your data, model, and findings to a non-technical audience.
Requirements:
- Show off your work to what would be a less technical, more business oriented audience.
- Summarize the work you've completed from earlier deliverables into a clean presentation, including:
- Your project's background, problem and hypothesis.
- Descriptions of the datasets you used.
- Data exploration with summary and charts.
- An explanation of your model (for non-technical audiences).
- Recommendations based on your findings.
- An appendix that includes all of your work and technical terminology.
- Review next steps with your audience; what could you do beyond the scope of this course?
Detailed Breakdown:
A 10 to 20 slide deck consisting of:
- 1 Outline Slide
- What is your project about?
- What is its history?
- What relevant information is required for a colleague to jump in to understand your project?
- 2-3 Summary Slides (including data and problem statement)
- What were you trying to accomplish?
- What steps did your project take?
- Where did the data come from? What does a sample look like? Was there data you considered but decided to remove?
- 3-4 Modeling Insight Slides
- What is the visualization explaining?
- What do the x and y axes mean?
- How does the visualization help either prove or disprove your work?
- What caveats have to be explained to best understand it?
- 2-3 Modeling Approach Slides
- What was your model trying to optimize for? Why was it the right metric for optimization?
- What algorithm did you try? How does it work?
- 2-3 Results Slides
- What worked? What didn't? Why?
- 1-2 Conclusion Slides
- What had the most impact on your work?
- What can you confirm? What can you suggest? What is still to be determined?
- 1-2 Next Steps Slides
- What should this project do moving forward?
- What would be the next two or three things you want to try? What impact might they have?
- What might your conclusions enable others to do?
Bonus:
- An Acknowledgements Slide is always a good idea. :)
- You might also include a Citations Slide, if necessary.
- Refer to the presentation template as a blueprint for how to organize your work.
- A quick outline (e.g. "what do I need" and "where can I find it") can help you prepare.
- Practice your presentation with a friend or family member! Outside feedback can help you identify gaps in your material.
- Limit the amount of visuals and text on your slides for maximum clarity.
- For instance, try not to use more than 2 visuals or 3-5 bullets per slide.
- Clean & informative presentations > Fancy Presentations!
- Keep your charts simple, and make sure they are clearly labeled.
- You can find previous General Assembly Presentations and Notebooks at the GA Gallery
- Presentations from PyData
- Presentations from DataGotham, a shortly-ran data conference in NYC.
- Seaborn has a handy easy way to set figures into a "talk" context, which blows up the text and makes it easier to read.
The rubric is available here.