Practicing What I Preach: Graphs on Slides (Part 2)

But I HAVE to show that graph!

In workshops I run to help science and engineering professionals and students think more consciously about their technical presentations, I often get pushback on the recommendations of my last post.

Essentially researchers are accustomed to having the data stand for itself, and their job is to show why the data says what it does. To them this requires showing the data: the most central part of the talk. The rest of the slides are backgrounds on previous data (often from other people), then predictions on where this current data might go.

The pushback is that without showing this data there is no talk. There is no purpose for this talk. So suggestions to remove graphs are antithetical to a researcher’s entire purpose of giving a talk.

Let me be clear – that is an absolutist version of what I am saying.

Firstly, audiences first. If your talk is not to other technical experts, even if they are other scientists or engineers but in a different field of study, they may not internalize the graphs the same way you do. (Curse of knowledge). Therefore think about whether you absolutely need to show every graph.

Read that carefully: I am not saying you should remove all your graphs. Or simplify all your graphs. I am saying you need to think very consciously as to whether a given graph is distracting or aiding the viewers. This is not as easy as it sounds, because you need to pierce through your curse of knowledge and think as though you do not know what the graph is saying.

If the graph is essential, focus on the best design for it. If the data can be said more simply in non-graph way (e.g. the result was 15x larger than previous), just say that and don’t sap mental energy from your audience with a complex graph.

If the data is a sidebar, DELETE THAT SLIDE.

This is why showing a draft slide deck to other people can make or break a talk. If they like and understand it, good going, move into practicing for delivery. If they are confused and frustrated by it, edit and revise the slides more. Repeat.

However, if you are presenting to direct or immediate peers that is a very different audience. Your closest peers are very used to the sorts of graphs you show. They may have even co-created them with you. They want you to cut to the chase and only talk about the central (often newest) data.

For technical talks, just show the most essential graphs and nothing else.

Here is the scenario. You have new data. You need to talk it out with other people before you proceed. They know exactly the sort of research you are doing. It is best for everyone involved if you just cut to the chase.

So here is a contrivance using the deck I’ve been revising in this series. Let us imagine that the purpose of the talk is to show the newest data I have collected. I need to chat with my direct peers about what I think this data means and how the next experiment should be set up.

That is, I am presenting at “group meeting” – the direct colleagues I work with in a single (laboratory) research group. In an academic setting, this is usually a single professor and their particular students and postdocs. Probably the most technical audience a given scientist can encounter because everyone speaks the same jargon dialect day in and day out and works to the same overarching research context.

The “full” version of this talk looks like the below, version 3 for those following along:

But let us say that slides 14 and 15 are the newest data: slide 13 essentially says a new experiment was done after slides 11 and 12 and here is the data. Let’s forget about slide 16.

Now, let’s use audiences first to decide what slides to show:

Yup, that’s it.

Here is why:

Slide 1:
Your immediate peers know who you are (probably) and what you’re doing (probably). Remind them of both. Then use this slide to just say a few things to bring them up to speed: you’ve done a new experiment and you would like to show them some new data. This will let their minds wake up to your talk.

Delete slides 2-13:
In a group meeting where you want to focus on data, focus on the data. Do not waste time giving background, past data, set up etc. Audiences first. The point here is to show the data and talk about the data. So just show the data.

Slides 14 and 15:
This is the data. Show it. Importantly, keep the graphs on screen for a long time so everyone knows what they are showing. As I said previously, graphs are a heavy mental lift. Give your audience time to see what you are seeing. Ask them questions. Discuss. Do not proceed until the majority of the room suggests it is okay to proceed.

Slides 16-23:
Delete. The purpose of this talk is slides 14 and 15. You can end on slide 15 because at this point you should probably take notes on what your group members suggest.

Think of yourself as a tour guide for your data. You show your peers what you see and ask what they see. Discuss.

Sometimes the best slide deck has just one slide. In this particular talk, slides 14 and 15 are essentially parts A and B, so both should be shown. But perhaps the group meeting comes up before you conceive of part B – in that case you could just show slide 14 and discuss.

This is how the majority of research gets done: you try something, discuss it with peers in some form of communication. Then you try again or try things newly thought of because you had that communication. Repeat.

The key is to let the audience determine the content, style, duration, and technical level of the communication. Audiences first.

Now, there are still some tweaks that can be made to these graphs to make them more conducive to a presentation – remember talks are not papers:

Like I noted previously, remove the graph titles – you say those in your oral remarks and it lets you show a larger graph. Delete design elements of the graphic software that are unnecessary. There is no need for the graph plots to be a square, so just go back to the essential axes. Make the axis labels bold and as big as what fits. If you have labels on the data itself, make the important ones easy to read and the secondary ones fainter so they don’t compete for immediate attention.

Remember too that text tipped on its side, as many vertical axis labels do, can be really hard to read. How many times have you seen an audience tilt their heads to their left trying to figure out what the axis says? So I have redesigned slide 15 to help the audience a little:

• in the next graph I am going to show you the result of predicted minus true (concentration) – that is, how much predicted deviates from known
• that data will be shown on the vertical axis against temperature
• here is the data

Since this sort of presentation is a discussion driven talk, not a presenter driven talk, there is no need to tell the audience when to clap: I have deleted the “Thank you” slide.

Now check yourself: do you think of this revised slide deck as “shortened”? If we are using slide count alone to determine length, yes, this is shortened. But talks are measured in time, not page (or word) count.

Slide count should not be used to infer length of talk.

Let’s say the original slide deck was delivered as a 20 minute talk. This 5 slide deck could also be a 20 min talk. You could spend a minute or so on the title slide, then 8 minutes on the next slide, and 10 minutes walking into then hanging out on slide 5. If you have a complex graph to show, only show that complex graph, and strike up a discussion with the audience to explain what you see and ask what they see.

Design your slides to give the audience proper time to digest your slides. Put audiences first.

If you have to show that graph, only show that graph.

This is a series on showing a little of how I approach a slide deck from a client. Specifically in this case, a technical talk that I gave in 2002 as I was preparing to finish my Ph.D. in analytical chemistry. If you’d like to see the other posts in the series, check out my blog.