The problem as I see it is – generally, there is no such thing as a process. When you map a process, you have just that, a process map. It is a piece of paper or screen view of what happens in the business. It is a graphical representation of routine, of behaviour, of what people do on a daily basis to turn inputs into outputs. An application form turns into a new member.

The issue here is – if processes don’t exist, how do you make them better.

An easier question is - what does exist. The answer is ‘routines and habits’. People come to work and complete their daily tasks using the same routine – day after day. The routine (processing sequence) is almost ingrained in them. They do not need to refer to operating manuals or ask colleagues for help. They know what to do – and when to do it. This routine is their daily habit.

(I do acknowledgethat Business Process Management (BPM) solutions substantially automate processes and remove the reliance on habit and routine).

A wise man (either Steve Rigby or David Lipschitz; they are both wise ) gave me the phrase – “Tell me how you are measured and I will show you how you behave” . This phrase is brilliant as it unlocks the difficulties associated with business process reengineering. If you redefine process from being “a sequence of tasks completed in a sequential manner” to “a sequence of habits completed in a routine manner” then you have the answer. Change the habits to change the process. How do you change habit – you change the measures. Tell me how you are measured and I will show you how you behave. Change the measures = change the behaviours = change the habits = change the process.

I am not so one eyed as to believe that measurement is the cure to all things process, but I am a firm believer in the idea that if you can’t manage it, you can’t measure it.  So measurement is a good start for changing the process.

Many of you will be thinking that your process measures (KPI’s) are reasonable and there is no need to change them. This is probably correct. The answer is to extend them, either on the same score sheet or in another sheet. Include measures of the specific points in the process where the process needs redefining (ie bad behaviour happens). Often these points are not process related at all. Rather they are decision points within the process – points in the process where the worker has to make a call on the best way forward, where they have to break routine. This can be for a multitude of reasons eg. Not receiving all the information they need, the customer has asked for something out of the ordinary, the computers are down, there is scheduled maintenance on a need piece of machinery etc.  Decisions cost money, increase headcount requirements, decrease the predictability of the process outcome and impact gross margins.  To reengineer a process is to minimise decision making in the process.

 A process is never a single entity – rather it is a collection of paths. Each path is created by a decision. Consider a decision with a 80/20 split.  Immediately you have two paths, each carrying a different volume. One has 80% of the volume, the other has 20%. Now imagine the 20% path has a decision with a 70/30 split. Now there are three paths, with one carrying 30% of 20%. Not much volume at all. Every decision introduces another path. Here is an example.

Paths 1,2 and 3 all complete – they all end with a common output.  Path 4 does not. Often path 4 will be modelled as rework to force completion.  In this example we will assume it is actually scrap.

Once the process is mapped with the decision splits in place, the next piece of data to consider is the work and cycle time associated with each step or collection of steps. Eg step 1 takes 30 minutes to complete and the hand off between step 2 and 5 has a built in 4 hour delay. So you plug in all this data. Then you need to attribute role costs to each step in the process.

Now you have volume x effort x cost. This will give you a weighted average of each path in the process.

Path 1 has an activity based cost of $100 vs the weighted average cost for the process of $149. Path 1 also carries 60% of the volume. The challenge is now to maximise the opportunity represented by this decision. The ultimate goal is to remove it from the process. If that can be achieved, then the process becomes path 1 and the weighted average cost becomes $100. This represents a savings of $49 per process run. Assume a volume of 10000, then this is a savings of $490,000.

A further benefit of this analysis is that it provides a foundation for practitioners of lean and 6sigma. Instead of applying those methodologies to an entire process, they need only focus on path1 as it provides the biggest bang for the buck.

To get rid of the decision is to take the decision before the process starts. This can be achieved with policy or behavioural changes, or it can be forced with technologies such as Imaging or workflow.

For the purposes of this blog I am defining data as predominantly time based data. The reason is that you need the Time element to make a decision. Consider if someone asked you to buy a BMW for $10,000. One of your first questions would be what is the model/year. A 2010 BMW for $10k is attractive. A 1990 BMW is not.

If a person told you that a factory ran at 1000 000 units you would ask – is that a day/week/ month or year. It gets less impressive as the time extends.

So the upshot is that it the time dimension is important for making decisions. Data without time is a merely a reference point. Eg Your name or the name of a company or your bank account number. These cannot have time attached to them and the data point seldom if ever changes. (you would hope your bank account number stayed constant).

To manage the whole, you manage the parts. Once a process is mapped, the business analyst should ask the process owner how long each step in the process takes to complete. It is not the vital to get the time right to the exact minute. As long as the time estimate is within reason, then it is acceptable. The process owner can often accurately estimate the time for each step. The role of the business analyst is to establish the weighted average time for the process based on the times given for each step. This approach makes the input data sufficiently accurate for decision making. If the business analyst knows their stuff, then they will convert the time into a weighted average total for the whole process. The process owner can validate this total as being a reasonable reflection of the process.

When analysing data at the enterprise level, what you are doing is aggregating many detailed processes into one step. This makes it is impossible to find a person who can estimate the time for an enterprise step accurately. Example: the high level enterprise process might be Buy – Make - Sell - Deliver.

To solve the data issue at the enterprise recommended techniques include multi dimensional preferential analysis or multi variate analysis. These techniques are excellent when there are hundreds of thousands to millions of records. This type of analysis provides substantial insights into data sets and should deliver ‘clean’ trends based on the bell curve. Outliers can be readily dealt with. This type of analysis is dramatically simplified in the following schematic.

 This type of data analysis will provide the data segmentation required to cost, evaluate and streamline a process at the enterprise level. The problem is having the skill set available that understands how to do this type of analysis – and the analysis type depends on the objective.

There are two possible objectives; 1. I have problem, what is the cause; 2. I have lots of data, what can it tell me.

The methodologies to address each question are similar but with different start points. The first question is best solved using hypothesis based problem solving and the second is initially addressed through statistical evaluation of the data and then overlayed with hypothesis based problem solving. An important issue in dealing with the second question is to remember to ask “so what”. A large data set will throw up lots of findings. The so what question will help sort the finding into nuggets of gold and nuggets of coal.

I was reviewing some data sets with a client the other day and he had a KPI report with lots of numbers on it. Monthly results for each indicator. He asked me what I thought. My observations were two fold. Both based on the so what question. My first observation was that he did not have a planned figure for each metric. He had lots of actual, but no planned figure against which to compare them. So how did he know if the actuals were good or bad. My second observation was that he had no ratios. A good management report will have a list of key ratios that show how the metrics are moving in relation to each other. Consider an inventory metric that is reducing. This is good as inventory holdings are getting smaller and cash is being freed up. Now consider a metric that says sales are increasing. An inventory to sales ratio will force the reader to realise that they may run out of inventory. This is very simplistic I know, but I have seen very good managers miss the obvious many times. What about purchase orders placed to inventory. This should show that purchasing is in place to address the impending shortfall.

 Performance is best described as plan v actual. What did I set out to achieve versus what did I achieve. This is quite a simple topic, but it is remarkable how many companies get it wrong, or just simply do a woeful job. I have never quite managed to get my head around why companies don’t do this well, especially as there is a huge industry selling business intelligence software that purports to do exactly this.

My experience has shown that supervisors are not taught to supervise. Let me explain.

The process of management is the same for all companies. At a high level managers need to translate demand into organisational requirements and then allocate these resources to work. Then they need to determine if the resources produced the output they were expecting.

Schematically is looks like this.

The management process is common for all companies. The business process obviously differs by process, by company.  The trick is to link the two processes together. First  you need to calculate the weighted average effort required to complete 1 iteration of the process. Eg to complete an invoice payment takes me 25minutes. You can calculate this figure or you can guess it. One is more accurate than the other, but the main thing is to have consensus on the number.

Then (simplistically speaking) you multiple this number by the annual volume eg 10,000 invoices and divide the total by work hours and you have your Full Time Equivalent headcount.  In this case it will take 1 person 555 days, working 7.5hrs a day to complete the work.  Or more rationally it will take 2.5 full time staff to manage the volume.

Now you have your staffing level. The next step is to allocate them to work. They should be working at the standard of 1 invoice every 25 minutes. This assumes a constant flow of invoices.

In the management process the step for allocating work is ‘schedule’. This is a vital step in effective management and is the primary responsibility of the supervisor or team leader as it is directly responsible for managing cost and revenues.

Once the work is completed, then the measurement step comes in. This should simply be a comparison of plan vs actual. The manager/supervisor then deals with the variance. Variances can be caused by poor process, low skills, laziness, pacing and a range of other issues. It is quite fascinating how businesses miss this step. They have scorecards that measure ‘actuals’, but seldom do they compare these actual to the plan. You need to ask yourself, why bother measuring actual if you are not comparing them to what you set out to achieve.

Once you have Plan vs Actual in place, you can build a report that generates ratios. While it is important to measure specific transactions and it is equally important to cross reference these transactional measures against each other to drive further insights. For example the sales measure is going up – that is good. Costs are going up – that is not so good, but easily explained. But an analysis of the gross margin ratio may show that the GM percentage is declining sharply. That is terrible. It is only once you compare sales to costs do you get the true impact of the trends of the individual measures.

While the above is reasonably obvious, it is quickly forgotten that you can’t fix a ratio. You can only fix the individual measures that make up the ratio – hence the need to manage the individual transactional lines.

A key question is – how does the supervisor manage the individual transactions in a way that any surprises are avoided and variances to plan are minimised. The answer is ‘Short Interval Control’.

Short interval control (SIC) is the primary weapon in the supervisors’ kitbag. It is broadly defined as – “measuring the process at an interval that allows the supervisor to take action on an operational variance before the variance causes customer promises to be broken and/or increased operational costs”

Consider a supervisor working the floor of a bottling plant. Bottles are whizzing around at a rate of hundreds or thousands a minute. Short interval control requires the supervisor to be checking his schedule every 15 minutes. If he waits an hour, the magnitude of scrap from a processing error would be enormous.  Contrast this to a claims processing department. Each claim may take 90 minutes to complete. Short interval control may only require checks at 2 hourly intervals.

This principle is shown in the following graphic.

The day starts at point A. This should manifest as a start of shift meeting. At this time the supervisor or team leader reviews the days production targets with the team (point ‘B’) S/He confirms staff attendance, set up / tear down requirements, processing sequence and in the event where different lines are processing different product, then this work is allocated out. The start of shift meeting should be no longer than 10 or 15 minutes.

Depending on the production environment, the most effective tool I have seen for managing SIC is a large whiteboard secured firmly to the wall. It has time interval columns pre-drawn on it with permanent ink. The supervisor then updates it with the days production as discussed at the start of shift.

The white board will have 3 rows for each production line. 1 for plan, the second for actual, the third for variance. Then on the appropriate SIC the supervisor will review the ‘shop floor’ and update the white board. This allows everyone to see how they are progressing. There are two scenarios – 1. Staff that are working too fast. This may be good, but it may also mean that they are sacrificing quality for speed. Consider an operator on a line preparing chicken skewers. If they work to fast, they will not clean all the meat off the bone and the waste will be high. On the other side, there is the operator working too slowly. This may be because of operational issues. The operators station is being fed raw product too slowly, it may be a skills issue or it may be attitudinal or a host of other problems.

In both situations the Supervisor is required to take corrective action to bring the production output back onto plan. This is point ‘C’ in the graphic. Ignoring the problem will result in a substantial variance at the end of the day; point ‘E’ vs. point ‘D’.

The most important point in all of this is that if you want to substantially increase productivity then adopting SIC in some form or other is mandatory. It is often critiqued as being Taylorism in disguise. In the wrong hands, this critique may be true, but with commonsense, it should never become a tool for driving punitive behaviour.

SIC does not only apply to supervisors and team leaders. It is applicable for all operational managers in all departments. The difference between the COO and the floor supervisor is that the COO has a longer interval between control points.

The all important question is how the COO should be advised of the daily production. The answer is HODO – the Hierarchy of Dependant Objectives (HODO). The HODO was a term given to me by my late father to describe operational communications between levels of management. It ties most of the concepts I have written about over the last few months together.

Consider a business that has an aggregate forecasted demand of 240,000 units for the next 12 months. In the simplest of terms, the business will need to produce 20,000 units per month to meet demand.

To increase the complexity slightly, consider that the business has two production plants. This means that each plant should produce 10,000 units per month.

Now consider that the business sells 4 different product types. Each plant is responsible for two lines. This means that each plant should produce 5000 units of each line per month.

The plants must produce 1,250 units per week or 250 units per day or 34 units per hour (7.5 hour shift).

To summarise – if the business produces 34 units per hour across 4 products, across 2 plants, then it will meet the operational requirements of the business.

To overlay this with a typical operations organisation structure – the Chief Operating Officer (COO), Production Managers, Supervisors, Team Leaders and the workforce. The workforce is not treated as management.

 As indicated in my earlier writing, the management process comprises of; Forecast, Plan, Schedule, Measure and Control, (inc reports) and Action. (Each of these steps are equally applicable to each of the management levels in the operations organisation structure. The scale and scope change by level of management, but not the intent.)

When you put it all together you get the following. The picture on the right is the traditional hierarchy. The picture on the left is the flow of information. The red line is the disaggregation of information down the hierarchy. Note that Schedule goes to Forecast. This is vital as you cannot have a lower level management working on an operational time line longer than that of the next level of senior management.

 The green line is the aggregation of information up the hierarchy.

These two lines (red and green) represent the hierarchy of dependant objectives. Operationally it may look like this:

There are a number of key points here.

1. 1. The management process is common for all levels of management
2. 2. Short Interval control is applicable to all levels of management
3. 3. If you do not have a formal schedule, then your measures may be a waste of time
4. 4. Without a formal review of Plan/Actual/Variance it is almost impossible to say if you have had a good day

It is also important to note that the HODO is applicable to departments, silos and divisions. It does not apply to the value chain. For further comments on performance management of the value chain, please see my comments on KPIs, SLAs and other measures.

Click here for a related article of this text. It is not a copy of the above.