Done, your profile is created.Finish your profile by filling in the following fields
Forgot Password Earn Money,Free Notes
Password sent to your Email Id, Please Check your Mail
Updating Cart........ Please Wait........
How to Give Good presentation
How to Give Good presentation 23
Welcome to this little introduction on how to give technical presentations. I hope you will find it useful.
The presentation is meant for relative beginners even though even seasoned presenters may find
something new in it. Naturally, we don't touch all topics and some only superficially. But once you are
aware of some basic ideas you can start a spiral of self improvement.
Note that none of the basic ideas in these slides are novel or invented by me. Rather they are extracted
from a set of excellent books listed at the very end.
http://www.ece.cmu.edu/pueschelHere on the title page we already see two principles in action.
Alignmentis particularly important; it gives structure and order to a slide (or any graphical design).
Interestingly, center aligment usually looks weak and unsophisticated. As a basic rule I suggest that you
align left if in doubt (also applies to tables BTW). Try it Beyond that try to align most elements on a slide to
Contrastmeans that if two elements are meant to look different, make them really different. For example,
look at the title and my name. The fonts differ in both weight and size which makes it look visually
appealing. Try the same with only size or only weight changed; it will look worse. As a general rule,
contrast in fonts is achieved by changing at least two attributes among weight, size, color, and font type.
http://www.ece.cmu.edu/pueschelI like this proverb and it fits well. Applied to the topic of this presentation I want to motivate you to strive
The photo is from istockphoto.com. The color for the text is extracted from the brighter parts of this photo
using Photoshop's color picker (useful trick).
The slide also goes along with one of my major rules: start your talk with an interesting slide and
certainly avoid a text slide.
Unfortunately, all slides in the remaining talk will look worse than this one.
http://www.ece.cmu.edu/pueschelHere is a couple of bullets expressing the motivation for this talk. The text seems reasonable enough but
there are some visual shortcomings. Can you tell
http://www.ece.cmu.edu/pueschelHere, the shortcomings are listed. Note how we violated to some extent the principles from the beginning:
alignment and contrast. Let's fix these.
http://www.ece.cmu.edu/pueschelMuch better, no The alignment and the contrast between bullets and sub‐bullets creates structure. As a
text slide it looks good, but do we really want or need a text slide (The answer is no as you may imagine.)
http://www.ece.cmu.edu/pueschelThis is a more visual slide designed to convey the same content.
The disadvantage is that now you have to remember what to say. However, if you don't, you are badly
The big advantage is that people will have more time listening since they are not busy reading (there is
a biological reason for this, explained later). Equally important, the visuals make the slide much more
interesting. So people are less likely to return to their laptops. This is an important goal, in particular on
the first few slides.
On the content: Presentations are a unique chance to connect your work with your person. If you do a
good job, people may remember you, which is particularly important in any type of job (academia: at one
point you need reference letters; industry: at one point a job up the ladder will open up and they will only
consider the people they remember). Unfortunately, most presentations fall far short of this goal and are
rather a waste of time for most of the audience. This is your chance: learn how to do it well and you have an
The bottom part conveys that content and visual quality are equally important. And they are. Note that
slides that communicate well are also beautiful. This implies (ever had a logics class) that ugly slides
will not communicate well.
http://www.ece.cmu.edu/pueschelNote that I had added an acknowledgment, since the plot was suggested to me by a colleague in my
department. Acknowledging is good style as much as not acknowledging is bad style. Imagine you
create a fabulous data graphic, give it to somebody who uses it and she does not mention you.
http://www.ece.cmu.edu/pueschelHere you see the natural enemy of the presenter: the addictive communication device of our times.
When you design your talk keep this in mind If you are boring for 2 or 3 minutes, people will start
checking email and all your work was in vain.
http://www.ece.cmu.edu/pueschelHere is the overview of this presentation.
http://www.ece.cmu.edu/pueschelNaturally, we start with the first point.
http://www.ece.cmu.edu/pueschelLike many other things, quality requires sufficient time for preparation. We are all busy, but preparation
is a matter of priority. And, until your are famous, presenting should be a priority for you. Once you are
famous you can hire graphics designer to do them for you.
It is hard to give a good rule for the number of slides. For technical presentations, I found that 2/3 times
minutes is a good rule of thumb. And this includes all slides. Yes, all. Don’t convince yourself that some
are just overview slides and some sort of empty. Does not matter. 2/3, not more. Leave room for you to talk
about your slides. In particular about your results (often towards the end).
Boot up your laptop in the morning of your presentation, You do want the latest OS to install itself one
minute before your talk.
http://www.ece.cmu.edu/pueschelKnow your laptop How often have you seen people trying to figure out how to get the image onto the
screen. You lose valuable minutes
Absolutely get a remote mouse. This allows you to roam around freely, stand where you want, and
naturally blend the advancing of your slides and animations with your speech.
Sneakers and shabby shorts at a major conference Are you famous yet
Superimportant: acknowledge your co‐authors. Not doing this is a crime. Not kidding. Usually they
worked as hard as you (or tried their best to advise you). So their names have to be verbally mentioned.
Imagine what you would do if half of the talk is created from your blood and sweat and you watch your
colleague talking about it and never mentioning you.
http://www.ece.cmu.edu/pueschelNow let's talk about the content.
http://www.ece.cmu.edu/pueschelYou have to start your talk with the motivation. What are you doing, why are you doing it and why is it
important (The last two questions are often connected.) Make sure you really answer these questions.
Don't start with a text‐only slide (remember the natural enemy) but with an interesting visual. The first
slide is your best chance to draw the audience in.
If at all possible give a precise problem statement. Ideally visualize it. If people don't know what are you
doing, why would they care about how you are doing it
Show other cool things if appropriate. Hint of the solution Cool example result Cool conclusion of your
work Everything that catches attention and makes people want to listen to you.
In long presentations (let's say 25 minutes), the overview slide helps to reinitialize people. You should
help with that by saying something like "I just explained ….., next I will show.” Maybe some people will
come back from playing with your natural enemy. Short talks don't need repeated overview slides.
http://www.ece.cmu.edu/pueschelHere is a very standard organization for a technical presentation. Use it if you cannot think of anything
Put slide numbers if one of your audience is not in the room (but connected over phone etc.). This way, you
can name a particular slide if things get messed up.
http://www.ece.cmu.edu/pueschelHere comes a big one. I just learned this recently. Text and speech is processed in the brain by the same
"channels." This means you cannot read and (really) listen at the same time. In contrast, images and speech
can be processed in parallel. The consequence is easy, right
http://www.ece.cmu.edu/pueschelThe previous slide tells us to minimize text. Ideally the visuals are on the slide and the text is produced by
you. Again, this requires preparation but it's worth it.
Honestly, when did you remember the last time a text only slide
Further, for text use bullets and be short (no need for complete sentences). Remember, every word on the
slide competes with you for the audience's attention.
If you need to have many bullets consider letting them appear as you speak (here, the remote mouse is a
must unless you stand right in front of your laptop).
Define every acronym and use them sparingly.
Use the following rule to force yourself to get rid of text. Go to "View Slide Sorter" and make sure no two
consecutive slides are text only. Once you mastered that on a regular basis, take it to the next step and
avoid text‐only slides (except maybe overview slides) altogether.
Since it is so important, I put the slide in red. As an aside, the nice shading effect is standard in the ppt 2007
menu (Shape fill → Gradient).
http://www.ece.cmu.edu/pueschelOf course, on of the main goals of your talk is to get the technical content across. This includes the
motivation, the exact problem statement, the main idea behind your solution, and your main result.
Do not try to get every detail of your work across. It is simply too much for most of the audience. This
does not mean you should trivialize but rather focus. This may include going deep for a few slides but not
too long, otherwise everybody will be lost.
Did you ever give a talk, presenting all the details of your nice work, and at the end you get a question that
shows that not even the problem you are solving came across Certainly happened to me many times. Find
the right balance for your audience (and this may differ by audience).
If many people remember anything from your talk it was a success. Seriously.
http://www.ece.cmu.edu/pueschelIn particular if it gets complicated.
Or equation heavy.
This may include properlyused animations (don’t fancify).
Note that this goes along with minimizing text but is somewhat different. Here I want to emphasize that
complicated technical content has to be visualized to at least some extent to give the audience the right
mental picture and at least a basic idea of what you are talking about. This requires practice. Let's look at a
http://www.ece.cmu.edu/pueschelI see this situation very often. A student gives a technical presentation with rather deep technical content.
Soon the slides are filled with equations and the students talks vividly about all kinds of details. The
audience is long lost since most don't see these exact equations all the time so they are struggling to create
a mental picture.
The interesting thing is that the student most likely does not have the material stored in her brain inthe
form of equations but probably does have some sort of structure or visualization in her head. So obviously
there is a much better way of doing it.
http://www.ece.cmu.edu/pueschelNamely like this. And it communicates immediately. Of course, not everything is a circle but any visual
helps. Remember: it is extremely important that people have the right mental picture of what you are doing
and how you are doing it. Words mean different things to different people and equations can be quickly
understood only by the expert in exactly the same area.
How do you visualize Draw from your own mental picture of the material.
http://www.ece.cmu.edu/pueschelHere are a few simple examples of visualizations. Block diagrams are particularly useful to identify
components, how they interact, input and output etc.
http://www.ece.cmu.edu/pueschelA simple figure as on the left can be used to give a problem specification (which you have to include:
mentioned in earlier slides). In this example, the authors built the red block called Spiral.
On the right side is a visualization of Quicksort. In the actual talk I let it appear in steps corresponding to
the steps in the algorithm.
http://www.ece.cmu.edu/pueschelComplex processes always require a visualization. Here is an example from biology. Note that when it gets
this complex, you need time to explain it if you want to.
http://www.ece.cmu.edu/pueschelHere some more things to remember. The red ones are common mistakes that you have to learn to
http://www.ece.cmu.edu/pueschelToo many text slides already In the talk I let these bullets appear one after the other. The two key points
here are the use of overview slides to get people back (from the natural enemy; see earlier slide) to
listening to you. For example, you just finished explaining your algorithm in all its technical glory and now
you come to the results. Emphasize this while you talk. Many people are interested in the results even if
they did not care too much how you did it.
http://www.ece.cmu.edu/pueschelLet's talk about the actual design of your slides.
http://www.ece.cmu.edu/pueschelWe argued earlier that the visual quality is as important as the technical content. This is not about adding
beauty to good technical content. This is about making slides communicate efficiently. Interestingly, slides
that achieve that are beautiful.
As basic layout I prefer black on white since it can also be printed if needed.
A word on logos: do not, I repeat not, stuff something into every corner of your slides. Keep the basic layout
(template) empty so you have a lot of real estate. If you want logos, put all into one corner and make them
small or avoid them except for the first and last slide (where you can make them larger).
Be consistentthroughout the talk (same fonts, same colors for the same things, etc.). As you get better you
may want to break consistency occasionally if it helps with communicating (an example are the two red
slides in this talk).
http://www.ece.cmu.edu/pueschelThere is a lot that can be said about fonts but we spend only one slide.
Serif fonts (like this one or Times New Roman) have little extensions in every character. This eases the
reading of large text block like books.
Sans‐serif fonts don't have these extensions and tend to be advantageous in small blocks of text (like
headers). Hence use a sans‐serif font. My suggestion: use Calibri. It comes with Office 2007 but can also be
downloaded for use with Office 2003. Since it is free it can probably used with other software as well.
For code (e.g., a C program) use Courier New bold.
http://www.ece.cmu.edu/pueschelThe math type setting abilities of Office are very limited. First, only very basic symbols are available.
Second, it does not look good. Compare the top right (Office) to the top left (texpoint, discussed below).
If you use a lot of math (and do not use latex to make your slides), get texpoint, a latex plugin for Office.
This means you can do all the math latex can do and the appearance is equally great. If you have already
written a paper, chances are you used latex, so you can just do copy‐paste.
http://www.ece.cmu.edu/pueschelColors are great and you should use colors. But how
Often I see the above ones in presentations. They are garish and, well, horrendous. Occasionally there may
be a good reason to use such a color but usually not. To look at these one actually needs ….
So if not those then which
http://www.ece.cmu.edu/pueschelPowerpoint offers these colors at the first level in the color menu. Nothing wrong with the shades of gray.
But most of the colors are not really useful, in particular if you need to fill a box or any area.
The solution is in the custom tab on the top. Let's click it.
http://www.ece.cmu.edu/pueschelWe end up with the menu on the right. First we need to understand color models. By default, Office shows
RGB (red, green, blue), meaning that every color is specified by 3 values between 0 and 255. However, it is
hard to relate these numbers to color.
On the right we show a different model: HSL (hue, saturation, lumniosity). H gives the rainbow
spectrum, the x‐axis above. S gives the sturation, the y‐axis above. The fully saturated colors are all on top.
All colors from the previous slide are fully saturated. A fully desaturatedcolor is gray. L is the luminosity or
brightness, the z‐axis so to say and shown by the slider on the right.
If you want a color choose it on top (or on the previous screen), say red, switch to HSL mode, and then
desaturateit. Then adjust luminosity and fine tune.
See how the desaturatedcolors look nice. They don’t command as much attention. More on that later.
http://www.ece.cmu.edu/pueschelHere are a few boxes that you can consider. Usually, for larger areas, choose a pastel color = a bright,
desaturated color. A box in such a color will serve its purpose but not distract attention. Sometimes dark
boxes can be useful. Consider putting white text into it.
http://www.ece.cmu.edu/pueschelThe biological process shown uses only pastel colors as does practically any map. In particular maps are a
good example of what is called layering. A way of overlaying different levels of information inside one
visual. These layers can be accessed by refocusing. For example, on the right, you can focus on the ocean
and rivers, on the streets, on the cities etc. This is possible due to the choice of colors (subdued = pastel). If
every component screamed for attention at the same time one could not look at it.
For example, on slide 23, the bottom block diagram uses some simple layering: the red boxes and the gray
area with gray text. Later we will see view graphs that use layering. Below on the page I use layering to de‐
emphasize the text and the slide numbers.
BTW, do you know where Aveiro is Did you know CMU has a joined Master's and PhD program with this
http://www.ece.cmu.edu/pueschelIf you have any quantitative result, you should show a viewgraph. Here are some basic rules. Most often I
see the font size rule and the contrast rule violated.
Make sure the plot clearly shows exactly the message that you want to get across (if any).
http://www.ece.cmu.edu/pueschelHere is a view graph. Now it is important that you present it so it is understandable. Don't forget that you
have seen it a million times but the audience has not. So they need guidance. Proceed in the steps shown
above. I'll do it for the plot shown.
•In this plot we compare the performance of single and double precision code for the discrete Fourier
transform, namely Spiral‐generated versus the IPP library.
• The x‐axis shows the base‐2 log of the DFT input size, so 5 means 32.
•The y‐axis show the performance in Mflop/s (I know the audience knows what this is), so higher is
•Now I point to one data point and explain what this means. E.g., for k=4, the triangle shows that Spiral's
double precision code runs at 1300 Mflop/s, which is more than twice as fast as IPPs code (500
Now you can proceed by talking about everthing you want. Then you canconclude. Here: Overall, Spiral‐
generated code is about twice as fast as IPP.
The plot is rather ugly. And see how the legend is placed somewhat randomly.
http://www.ece.cmu.edu/pueschelHere is an attractive plot. Study it for a bit and try to verbalize why. Also, try to find design decisions that
you maybe have never thought about trying.
Most of the design decisions I stole from The Economist.
http://www.ece.cmu.edu/pueschelHere is an explanation of the main features. Interestingly, most are based on principles we already learned.
Most surprising to me when I learned it was the following:
•There is no reason to have the y‐axis label vertical. And if you think for a while it does not make sense at
all. One simply has a hard time reading a vertical label. As done above, it is perfectly clear that the y‐axis
is the number of people.
•You do not need a y‐axis, at least not here.
Besides that, note how contrast is use to create a layering effect. The data is at the fore (as it should), and
the grid lines and everything else are in the back.
As Edward Tufte says: "Above else show the data" (see references on the last slide).
http://www.ece.cmu.edu/pueschelHere is a black white plot since it was used for print publishing (in a presentation, color would be much
better). We decided to invert the background since it separates the plot clearer from its surroundings in a
The one possibly surprising feature is that there is no legend. In fact, if possible avoid a legend and put the
description right into the plot. Think about how long it takes to parse the connection between lines and
legend. Eyes go back and forth and back and forth …..
As done here it is immediate.
http://www.ece.cmu.edu/pueschelThere is much to be said about tables; and as with plots and presentations, it is not straightforward to
make good tables. But again, some of the same principles we have already learned apply. See the above
guide for a short introduction.
http://www.ece.cmu.edu/pueschelA brief note about Powerpoint.
http://www.ece.cmu.edu/pueschelI much prefer Powerpoint over any other software for making presentations. 2007 is in my opinion a major
improvement and makes it much easier to create attractive presentations. As a mathematician, however,
you may want to consider latex. A very nice style for slides is available in latex (my only major complaint is
that the template uses too much of the real estate). But back to Powerpoint.
If you frequently make presentations, learn how to really use Powerpoint (or whichever tool you use). It is
worth it. This includes the above tips.
http://www.ece.cmu.edu/pueschelAs said above.
http://www.ece.cmu.edu/pueschelI introduced a couple of basic principles in this presentation. The first four apply to graphic design in
general. The fifth is crucial in presentations, since it makes it possible to communicate complex technical
content and it minimizes text, which is one of the key goals in making presentations. The last one is
generally good practice. Let me breiflyrestate the main points:
Alignment: Don't randomly place elements. Most should be aligned to something. If in doubt, try first left
Contrast:If you want that elements look different make them sufficiently different. For fonts change at
least two attributes. With colors play around. E.g., warm against cold, dark against bright gives good
Layeringmeans that in every visual you decide which elemntsshould be in the forgroundand design
accordingly. Lesser important elements you put in the background, e.g., by making them for example gray
or bright or both.
Consistencymeans that similar elements should be designed similarly. For example, use the same font
throughout the talk. You may break the principle for emphasis.
Visualization:Try to display problem statement, your approach, your algorithm, complicated technical
background, theorems, etc. Everything that works is allowed. Start from the way you think about things.
Minimize text more and more until you can create talks without text‐only slides.
Acknowledgeyour co‐authors (say their names), material borrowed from other sources, and related
http://www.ece.cmu.edu/pueschelI repeat the most important points. In your summary, consider repeating a key visual from your talk.
Take every opportunity to practice presentations and always give your best.
Watch other presentations and try to see what works and what not. The best exercise is to try to
verbalize why something looks good and why it does not. Once you can verbalize it, it is under your
control. Also, design is everywhere, so look at it. Look at viewgraphs in newspapers and magazines. Look at
any form of design. E.g., posters in museums alywaslook great. Spend a minute looking at the design
decisions. Do you recognize some of the principles we talked about
You cannot become a great presenter over night. Like everything else, it takes practice and the desire to
get better, i.e., it does not happen automatically. I hope that with the few tips from this presentation you
can get on your way to becoming a presenter that people remember. Good luck
I am happy to hear about any comments. Email me (my website is below).
http://www.ece.cmu.edu/pueschelAs I said in the beginning, I did not invent anything in this talk. Everything is collected from books. Here are
a few of them that I consider particularly helpful. Check them out.