Analysing the heck out of decisions

Decision QuestionSo finally – after deferring it off for several posts – we have made it back to decision analysis and the use of decision tables. We had started to explore this in a post ages ago – on the 10th of June 2013 to be precise – but today the focus is on how to actually do decision analysis and create decision tables. I think we will be covering a lot of old ground first though… last time we talked about decision tables it was very much about what they are with a limited focus on how it fit into the big picture. Now I would really like to get a better handle on what decision tables – and decision analysis generally – are all about.

I would like first up to acknowledge the professionals in this field from whom I’ll be taking a lot of ideas from – Ronald G. Ross and Gladys S. W. Lam. I connected to a brilliant online training course with them at the end of last year, with the only downside for me being that the course was run at 2am my time. I must say I am not functioning at my best around then! The good side was it was DEFINITELY worth it. Check out if you are interested.

I think we’ll use a question and answer type of format to this post. There are two benefits to this. It makes it easy for me to track what I need to cover and it (hopefully) makes it easier for you to follow my logic. Feel free to let me know if there is a better way of handling these relatively complex subjects.

So, first up, what is decision analysis?

Decision analysis is the analysis of how operational business decisions are made. This is not the analysis of policy decisions. Nor is it the analysis of how individual people make personal decisions about whether they comply with an established procedure or policy. It is specifically about exploring the logic behind what professional decisions are made in the course of following an established procedure.

Why would you undertake decision analysis?

To know how a particular operational decision is made within your organisation.

Why would you want to know?

For several reasons. Firstly, you might want to retain your corporate knowledge so that this decision-making ability does not rely on the availability of just one or two people. Secondly, you might want to support the decision-makers by making the required information more easily, quickly or reliably available. Thirdly, you might want to make sure that the decision is based on appropriate logic and the correct information. And fourthly, you might want to be ready to change or improve (eg include more detail in) the decision-making logic based on the outcome of previous decisions. This fourth reason is at the heart of adaptive management.

So how do you actually do decision analysis?

Here is the nuts and bolts of it. A possible high-level procedure for how you actually analyse a decision.

  1. Identify what decision-making step you want to analyse. This may be through generally questioning someone in the business who is involved in the making of a decision, or might be through the documentation of a business process. We will use the example of a fruit chew ordering process, where the decision-making step is ‘determine fruit chew order’.
  2. Identify the question that is answered through the decision-making step. For our example, the question we are trying to answer through this step is ‘How many of a certain type of fruit chew should be ordered?
  3. Identify the possible answers to the question (ie the possible outcomes of the decision). Using the example from step 2, if the answer is derived purely from a mathematical algorithm, going into any detailed analysis will be pretty pointless. You might as well just write down the algorithm and be done with it: (Store capacity of fruit chews) – (Current number of fruit chews in stock) + 1000 = (how many fruit chews should be ordered). However, if it’s conditional on a number of considerations then further analysis will be possible. In our example, it is probably sufficient to say that the answer will be a certain number of fruit chews.
  4. Start digging down into the next layer of detail to get to the crux of the logic behind the decision by identifying the relevant considerations to explore in answering the question. So yes, this is the heart of decision analysis – time to get your brain into gear. For example:

Is this type of fruit chew usually in stock?

Has this type of fruit chew ever been stocked before?

Is this type of fruit chew going to be promoted?

Has it been decided that this type of fruit chew will never be stocked again?

  1. Look at the different considerations and work out if any of them are decisions (really sub-decisions) in their own right that should be explored separately. In the event that a consideration is a decision in its own right, spin if off into its own little analysis project and start back at step 1, although you will need to keep addressing this considerations over the rest of these steps at a high-level. When you get to Step 9 you will somehow want to reflect the relationship and dependencies between the different decision you are dealing with.

A good example of a separate decision is the fourth consideration above. Has it been decided that this type of fruit chew will never be stocked again? Really this is a decision in its own right, and although it definitely is a consideration, it would be worth exploring the question ‘Should this type of fruit chew be ever stocked again?’ It is worth noting that this decision dependency is completely unrelated to the dependencies between process steps – decision dependencies are based on logic, not sequence.

  1. Next, breakdown these considerations into all the different possible options and combinations. For example:

Is this type of fruit chew usually in stock? – Yes / No

Has this type of fruit check ever been stocked before? – Yes in last 10 years / Yes in last 2 years / Yes in last 6 months / No

Is promotion of this type of fruit chew going to begin next month?  Yes / No

Has it been decided that this type of fruit chew will never be stocked again?’ Yes / No

  1. Identify any exceptions. These are instances where the normal decision logic does not apply. For example:

Rainbow chews are the store’s signature chews and are used liberally and on a complementary basis within the store and outside the store for marketing purposes.

And… Where a fruit chew is being promoted, additional fruit chews will need to be ordered.

  1. Now develop the business rules. Make sure you are being very clear in your wording otherwise ambiguity may get in the way of what you are trying to clarify. I’ve bolded some of the words to show where I’ve got rid of some of the previous fuzziness. Especially where there end up being big long lists I would suggest that – especially if you are planning on developing a decision table – there is no point in writing the business rules ad nauseum. However I think getting somewhere down the path of documenting business rules is useful in clarifying your thinking.

For example:

 ‘For every fruit chew type on the Standard Stocklist, the same number fruit chews that were purchased in the month prior to the order must be ordered.’

‘For every fruit chew type on the Suppliers’ inventories that has never been stocked, 250 fruit chews must be ordered.’

‘For every fruit chew type on the Suppliers’ inventories that has been stocked in the last 2 years, but not in the last 6 months, 100 fruit chews must be ordered.’

‘For every fruit chew type on the Supplier’s inventories that has been stocked in the last 10 years, but not in the last 2 years, 200 fruit chews must be ordered.’

‘For every fruit chew type that will be subject to promotion starting in the next month, 1000 fruit chews must be ordered in addition to the amount otherwise required.’

‘Fruit chews that are on the Never Sell List, must never be ordered.’

‘1000 rainbow fruit chews must be ordered.’

Remember that you will often be undertaking this decision analysis in the context of capturing processes, so in this case we know that there are other processes regularly adding and/or removing fruit chew types from the Standard Stocklist and the Never Sell List. Therefore these rules are probably more comprehensive than they may appear in the first instance. However, as you can imagine, some decisions will have extremely long lists.

  1. Put together a decision table or broader decision-making framework to most easily represent to the users the decision logic you have uncovered so that it can be repeated next time the decision-making step is required.

Now, I planned to explain decision tables and broader decision-making frameworks in a bit more detail during this post, but I’ve just noticed the word count and it is getting a bit big, so I think I’ll defer this to my next post. So stay posted. There is a lot more detail about Step 9 on the way.

But before then, what do you think?

Do you think decision analysis is severely underrated or underutilised? And do you have any ideas for another way of going about it?

What do you think?