So, the idea of having a central message isn't just about text, it also applies to data graphics. Remember, we've always asked this question, why should your audience care when thinking about finding a central message? You have the same question when you're trying to present data, why should your audience care about these data? What should they remember about these data? There are many ways to represent the same data, many different kinds of formats and designs that can represent the same data points. Choosing among these formats, ought to be done based upon the exact same factors that we've been talking about for any message. Who's your audience? Who are you trying to reach? What is the context? Where are they going to be interacting with this data graphic? What is your goal? What do you want them to get from looking at this graphic? What's the takeaway? What's the big point? So, here's an example. This is one simple data point. At 18 months, whatever this is that I'm representing here was something like 36 or 37. But is that good or bad? I don't know. Because I have no other information other than the fact that this measurement was 36. But, we often don't just give people one data point, we give them more. So, let's say there was measurements at 12 months and at 18 months. Now, what I see is that at the 18th months, measurement was a lot higher than the 12th month measurement was. So, I've already gotten one message. It's higher. I don't necessarily know whether it's good or bad. And so, without any more context, I might not actually remember this graphic. I might not be able to use this graphic. Notice what happens if I put one more piece of information however into this graphic. Let's say there's some maximum level. Well now, I start to recognize that the 18th month measurement is approaching that maximum level. The 12th month wasn't. I might have concern about that. I'm noticing a trend and notice the words that I'm using, approaching. I'm interpreting from the trend between that 12th and 18th month data, the idea that the values are going up. I still may not know whether this is good or bad but I'm getting a lot more information consciously from the context, the comparison between those two data points and that reference point. Now, I've been talking about this like there's a trend. The values are going up, except maybe they aren't. So, here's the same two points, the 12th month and 18th months point, put into a trend of data that goes back to the zero point. Which you see here in this graphic on the right is that in fact over time, the values have been going down and only ticked backup in that last little bit. Put into that larger context, you might or might not think the same thing about the difference between that 12th month point and that 18 month point. Again, I'm not saying one is right or wrong here. I'm just saying, notice that just simply by adding a little bit more contextual information or historical data, I'm actually changing the way you interpret that 18th month observation. So, let me make this concrete in a context that I've actually been studying in my own research. So, when we're communicating to patients laboratory test results, we have the same problems. So, here's a graphic that shows a variety of hemoglobin A1c test results. Now hemoglobin A1c is a test that's often done on people with diabetes to measure how well they're managing their blood glucose, their blood sugars. Which is a measure that helps you to know how much their diabetes might be affecting their future health risks. So, what do you see in this graphic? Well, what I see is a lot of up and down, a lot of variants. And so, the central message of this graphic is that the data vary, it goes up and down. Now, let me compare this graphic to a few other ones. And I want to be clear here. I'm not going to change the data, I'm just going to change the format, the way it's represented. So, one thing I'm going to do here is to show you this one. Is this the same data? It actually is the same data but boy it doesn't look the same. Here's another one. Same? It is the same data. But again, it doesn't look the same. Why? Well, let's look at the central messages here. The central message of the graphic on the left is variability. It goes up and down. The central message on the one at the top-right is that it's stable. It's in the middle of the graphic and it doesn't move much compared to the way in which we've laid out the graphic. The one on the bottom, I think the central message here is that it's high. The scale gives me an impression that the values are high. Again, I haven't changed the data here. I'm just changing the structure of the graphic, and I'm taking away different central messages because of that. What did I do? Well, the one on the left one is scaled to the variations of this particular patient's data. The one on the top-right is scaled to the variations that are commonly observed in the population. We see people regularly with values up around 9 or 10. We see people down around four or five, and so the data are represented compared to the population distribution. The one on the bottom is just scaled down to zero. Now, it happens to be in this context. You don't tend to see, people don't have zero A1c. So, that may or may not be an appropriate way to scaling it. But for many measurements, scaling down to zero is commonly the way we think about how a graphic ought to be scaled. Notice what I'm doing here, I'm making a choice about which reference scale to use, that changes the central message that I take away from the data. So, is these values good or bad? If this is all I give you, you don't really know because you don't have a clear reference point as to what is good. I can do one other thing though, which is what's commonly done in test results communications. Which is to give you a reference standard, a standard range. Just adding that single reference point, instantly makes me think like oh, man my values are too high. I haven't changed the graphic here. I'm just giving you a reference point. Notice that the choice of the reference point though is really important. In this case, I gave you the standard range, which is what applies to people who do not have diabetes. But, let's say you've already been diagnosed. You know that that standard range doesn't apply to you, you've been diagnosed with diabetes. So, now let me change it. Here is let's say a target range for patients with type 2 diabetes. What's the central message of this graphic? I don't know about you but what I get is, I'm doing great, all my values are right in the target range. Notice how differently that feels compared to the graphic we saw just before. The point here is simple, visual context matters. Many different things determine the central message that we see in a data graphic. Yes, the data points also are critical point. They are the actual information that are being presented in that data graphic. But the central message is also determined by comparison data, that historical data or not, the axis scale that we choose to use and there's not one right answer. There are many different axis scales that might be considered appropriate, and reference standards. What reference points do we want people comparing the data that are shown to in order to get the message that we want to give them? There's no single answer for what is best here. It's going to have to depend on the situation. And yes, I know some of you are thinking, ''You're manipulating the data here.'' Yeah, in one sense I am. There's an old book, How to Lie with Statistics from 1953 that talks about exactly what I'm talking about here, it still applies today. But the honest truth is; there is no such thing as a neutral data graphic. It doesn't exist. It can't exist. It can't exist because we always have to make design choices about reference standards, or axis scaling, or comparison points. And these choices determine the central takeaway message that our audience is going to interpret from the data graphics that we give them. The best thing you can do is simply to be intentional. Know before you start, what is that one central message that you want your audience to get from your graphic? Not just about presenting the data. You're giving people a message and you better know what that message is, and then make those choices. Design the graphic to highlight that message. Maximize the chance that the audience gets what you intended when you designed that graphic. That's when you can use the data graphic process, the communication, to achieve your communication goals.
