donderdag 4 december 2014

How to lie with your KPI

There are lies, damn lies and KPIs

Often KPIs are used to manage and prioritize activities within the organization. It is expected that employees act upon the KPI outcome. If in the end the KPI does not show any progress, it might have an effect on the performance assessment of the people involved. Especially in an autocratic lead company these consequences of KPIs turning RED might be harsh (see also my blog on management styles and KPIs).

A too rigorous usage of KPIs in relation to people management might lead to a culture of fear. Employees will do their best to avoid the RED status. Most of them will make sure that the KPI is moving up (or down) by just doing their best (hoping the KPI turns or stays GREEN). Others might go a little bit further to keep the KPI in status GREEN. 

Manipulating techniques are not that hard. Knowing a few of them might even come in handy. Not to use them yourself of course! No, just to recognize them when encountered. In the end: it takes a thief to catch one. I've listed six techniques here. They are subtle and are meant to create a smokescreen around the real results. In other words, they help presenting the results better than they really are.

1. Play with thresholds
This trick was already mentioned several times in my previous blogs. It is very easy to manipulate the thresholds above (or below) which you indicator-status turns amber or red. As long as you stretch the threshold long enough, your status will stay green. See here an example I found depicted in an article called "Why Red Amber and Green (RAG)?" at intrafocus.com.


It doesn't take much fantasy to see that one could easily make the green area larger by increasing the lower threshold.

2. Lie with your graph
Playing with your threshold is not the only technique. There are many tricks you can play with the way you present the results. Especially line graphs are easy to manipulate by altering the Y-axis. Consider your starting point, end point and scale of your axis. Use suggestive labels or add chart junk. Use two Y-axis to confuse your readers (if you can't convince them, confuse them). This short blog however isn't the place to discuss all these different techniques. There are some really nice books that I can recommend (1) (among which is my own book on the misuse of statistics ;-)

3. Work with percentages
The percentage is a very popular statistic. That's because almost everybody above 10 has a basic understanding of what it stands for.  The Dutch author J. Bakker once said: "percentages are like bikinis. They bring you to all sorts of ideas, but hide the essence". That's probably why percentages are used in commercials all the time (31% less wrinkles! 70% less fat!). Most often they are as hollow as the claims they support. This is because only the percentage does not say anything. It's the absolute figures behind them that really count. An increase of 200% sounds very impressive, but could mean everything or nothing (going from 1 to 2 is also an increase of 200%).

4. Choose your average wisely
Let's say you launched a new website and you want to see how successful it is. Your KPI is wisely chosen. Not the number of hits, but the average time people stay at your website is your performance indicator. The longer the better. The picture below shows three possible outcomes on how many people stayed a certain amount of minutes on your website.


Now have a look at the average-measurement most often used: the mean. Depending on the "skewness" of the results your mean could be lower or higher. So let's say that most people only stay a short time on your website (represented by the graph on the right). Using the "mean" as your type of measurement however gives you the impression that that the amount is higher. This is because the few "fans" that stay at your site a long time, push the mean upward.

5. Leave out certain data
KPIs don't like extremes or outliers. These incidents might influence your indicator and result in a (temporary) RED or AMBER status. So one of the most tricks used is to just name these extremes "incidents" or "a coincidence" and remove them from your graph.


6. Aggregate your KPIs
Remember the cartoon from the "Manager styles and KPIs" blog? The one with the birds? What does the bird at the top see? Not much really. If there is a KPI set for each of the departments below the bird on the top, most likely the overall status will turn GREEN every time. This is because aggregating three values (RAG) will do that. Look at the picture on the left. Even if there is are many AMBER and RED departments throughout the organization, the top level KPI is green. The chance of the top level KPI turning RED is very small. This is because working with the RAG structure makes you limited in the way you aggregate them upwards.




(1) recommended reads on manipulation with graphs: "Van tofu krijg je geheugenverlies" - Coen de Bruijn and "How to lie with charts" - G.E. Jones

Geen opmerkingen:

Een reactie posten