A Simple Way To Better Decisions From Your Data


Your Keys to Better Decisions

Wouldn't it be great as a business owner, director or manager for you to make better decisions? In the face of all the change in the world today that's really hard. Wouldn't it be great if there were ways to make better decisions?

In an ideal world decisions would be taken on the basis of expert intuition. Trouble is everyone has biases. Someone has even catalogued over 100 biases!!! Who or rather what are you going to turn to? Many would naturally turn to the numbers or data to evidence decision making. So far so good.

Ah data; businesses and organisations everywhere are collecting it. However, are they making the best use of it to improve their activities? Usually they aren't because as Daniel Kahneman points out people operate with biases even on data.

A very basic problem is the wide misunderstanding of the difference between using data to count or to analyze. Proper analysis of data gives you the basis for effective decision making and action, ideally also as a basis for ongoing improvement.

Now I have blogged on that before and I think it is worth repeating. This time though I will take you through the simple steps so that you can extract meaning from your organisational data, which can be what some call Key Performance Indicators (KPIs).

The graph in the image at the top of this blog might give you a clue that the effective way to analyze your data is graphically. The graphical way to a better, more effective decision is shown in the following four stages:

  1. Collect and record your data in time-order sequence

  2. Put data in a time series - so you can tie each item of data to real world events

  3. Add the three special decision lines - these differentiate between signals and noise - more on that later

  4. Look for signals for action - data contains both signal and noise. The latter is filtered out using the decision lines

Stage 1 - Collect Your Data

We'll take some data or numbers that you might expect to see in any organisation, which for the purposes of this blog are the number of monthly visits to an e-commerce business's website.

The Managing Director (MD) makes it very clear that he is not happy with the last two months being way below budget. July was the worst month for a very long time. Although there was some recovery in the August figures.

The new Marketing Director, Matt, produces more data to better understand the situation looking back over the previous 2 years. Here they are in a table:


Monthly Website Traffic

At the sight of the this the MD starts to fume,

"We're going nowhere! Look at that July figure! This is completely unacceptable, it's all over the place. People had better shape up or ship out!"

Looking at this data in a table at first it isn't easy to see any trends. You can see the MD’s point. True February 2017 was a fabulous month with 148,000 visits, which Matt has heard might have been due to a social media campaign mounted by his predecessor after a disappointing December 2016.

Otherwise it is hard to see patterns. Recent performance has been relatively poor. As someone once remarked, “Some month’s are better than others.” Actually that is the point. With so many influences at play in the real world you must expect some variation between data points.

There is a lot more to collecting data, especially if you are using some sort of measuring equipment. We'll cover that in a future blog.

Stage 2 - Make Your Data Visible

In this blog we are going to make the data visible. Yes, I know you can see numbers in table! However, it is hard to see patterns and trends, also whether particular data points are truly exceptional. Exceptional values at known points in time can give us clues as to why the system behaves as it does.

The Quality Manager, Rebecca, intervenes at that point.

“Matt showed me this table of data earlier and I have taken it and firstly plotted it in a line graph against time as you can see in these copies.”


Line Graph of Two Years' Monthly Website Views

The MD explodes,

“What the flipping heck happened in February 2017? I’ve been telling you we could do better than last July. Heads will role if something doesn’t change and pretty smartly!”

At this point, highly diplomatically, Rebecca intervenes ,

“Yes boss, in the past the company has clearly shown on at least one occasion that the company can do better. Unfortunately it has not repeated since. Perhaps we can use the charted data to understand why.

Stage 3 - Add Three Decision Lines

At this point Rebecca hands round a second graph

“I have taken the line graph a stage further and added three decision lines to help us better understand what has been happening and how to improve in future. This is called a Process Behaviour Chart”


Process Behaviour Chart of Ecommerce Traffic

After handing a second sheet, Rebecca continues,

“It is certainly clear that February 2017 shows we can do better; however, this is a highly unusual event. Over time, mostly the rest of the data varies randomly around an average of 86,042 per month, as shown by the green line.”

The MD says,

“Yes, yes. I get the average, although I am less than happy about the amount that traffic is varying above and below. But what I want to know is what are theses two red dotted lines?”

Rebecca replies,

“They are calculated such that if a system or process is operating in a stable and predictable manner the likelihood is that 99-100% of the data will fall between these two red dotted lines. Data should also be randomly scattered on either side of the average.”

Stage 4 - Look for Action Signals

The aim should always be to make better decisions about what action to take. From the above there are two rules that are used to detect exceptional events, which are worth investigating further to understand the underlying or root causes as a basis for action:

  1. Any data points above the upper decision line or below the lower decision line

  2. Any runs of at least eight points on one side of the average, i.e. signalling that something highly significant has changed in the system or process signalling a change in the average

Rebecca says,

“February 2017 is outside the upper decision line and confirms that it is an exceptional event, as per Rule 1. We need to dig down and find out what happened, because we would hope to find something that enables us to increase traffic permanently to that level.”

“For the rest the point-to-point variation is much lower and between the upper and lower decision lines. The chart, though, does show something different happened in the eight months between August 2017 and March 2018, as per Rule 2.

“It is statistically highly unusual to have eight points in a row on one side of the average. This is even though the pattern of the data return to the wider variation afterwards.”

“We have two clear signals for action: February 2017 and the eight months from August 2017 to March 2018. Whilst there are methods for understanding these two situations, has anybody got any ideas what was different?

There was a pause.

The HR Director says,

“Well. If, as recall, the August to March period sounds like that might be after the Marketing Director, Alex, and Social Media Executive, Derek, left and we took on a part-time temporary employee. And in March 2018 is when you joined us, Matt.”

Matt says,

“Yes and the part-time temp left us at the end of June after being replaced by a new permanent full-time Social Media Executive.”


Says the MD, who goes on,

“So are we saying here that there is a link to the effectiveness of our Social Media that might account for the poor performance? But what happened in July?”

The HR Director says,

“Well Laura, our new permanent Social Media Executive, was being onboarded during July.”

“She seems very capable and is already suggesting changes to our whole digital marketing strategy and plans.”

Said Matt.

The MD then went on,

“OK. That all seems very logical. I am keen to see what Laura and yourself come up with for our digital marketing. But I want to go back to February 2017. What happened there?”

There was silence. Paul, the Finance Director cleared his throat.

“Erm… Wasn’t that about the time that Alex, our former Marketing Director was getting carried away with all that Chamber of Commerce Award PR and stuff?”

Silence. Everyone looks at Paul and then at the MD.

“Ah… Yes… “

Says the MD, who goes on,

“Um… That was just before there was that disagreement about the Marketing budget at the Board meeting, which led to Alex and Derek putting their notices in."

More silence. Rebecca, the Quality Manager, breaks the silence,

“So might we have found the the root causes that explain the two action signals shown in the Process Behaviour Chart? It is looking like Social Media and digital marketing in general plus PR are key to maintaining and improving performance.”

Let’s leave the meeting there as they start to get their heads around all that. It looks like the MD could benefit from getting his head around what digital transformation expert Warren Knight calls digital leadership.


We can conclude from this story

  1. That one piece of data taken out of context has no meaning, in fact it’s just counting

  2. Decisions taken on this basis are no better than guesses, especially when human bias is taken into account

  3. Effective decisions require analysis of data in the context of other data points

  4. A Process Behaviour chart, a time series chart with decision lines, enables you to separate exceptional events that are signals for action from among the real-world noise in the data

If you would like to have a go at creating a Process Behaviour Chart contact me for a FREE worksheet for the decision lines through the button below. Please include your email address, which we will only use to send it to you.

Thank you for reading this blog. Hope you have enjoyed it and found it useful. If you have any questions do please use the contact page. Do please share this blog with your friends, colleagues and connections. Looking forward to sharing my next blog with you.