The concept of workflows comes from the time when work was rather mechanical. People (and later machines) were supposed to repeat precisely defined steps in a fixed order (or sequence). Any deviation from prescribed path, any thinking and improvising was not welcome. Assembly lines or microprocessors - they were all realizing rigid sequences.

In today's business world processes are present everywhere. They are also driven by humans. This means that process automation is not sufficient anymore. It's reducing people taking part in the process to a simple machine-like element. We can, however, support people in the process by using state machine workflows.

State machine and sequential workflows

Before explaining advantages of state machine workflows over sequential ones, let's try to understand the difference between them. Let us take a look at an example, one that is probably known to all (or most) of us: an in-car GPS navigation. When you set your destination, the GPS shows you the way with detailed instructions where to turn, where to go straight ahead. There is only one itinerary (may have variations if there is traffic jam) and one sequence of moves. You follow them by mechanically performing instructions (turns) defined by your navigation system. You don't think, don't make decisions. It's like sequential process. Now - imagine a map with a starting point and your target. You want to reach your final destination, but on your way you make some decisions that may change the way. You may decide to visit another city, take scenic route instead of highway, stop for lunch or go back to the city you already visited. Of course, choices are limited in order to let you finally come to the end point, but you decide how to compose your itinerary. It's similar to a state machine workflow: it moves from state to state when an external event (in this case your decision) moves it to the next state. You can control at least some aspects of the process.

Coming back to our business processes, imagine a Contract Approval that has to have three approvals: Legal Department, Finance Department and your boss. Let's assume that you may get them in any order. In sequential process the order of approval is predesigned, for instance: first Legal Department, then Finances, and finally your boss. You can't change it - you have to follow it. In the state machine process, however, it's the user who decides to which department he wants to send it first (assuming that you want to leave this option to the user). The only requirement is that at the end he has to have all three approvals.

State machine model clearly separates states in which the process awaits a specific action to be performed (approval of a document, its verification, rejection, asking for correction etc.), and transitions which may be a single action or a sequence of actions. Usually states are places for human decisions, while actions are programmed and performed by machine. Sequential model mixes it all in one schema. Sequential workflows are predictable and in control of the process. State machine workflows are controlled by external environment (people, for example) and may have many paths.

How does it affect business?

Now, understanding the difference, let's move on to the question: WHAT DIFFERENCE DOES IT MAKE IN BUSINESS? Today companies need to be agile and their processes need to change, constantly adapting to rapidly changing conditions. Also, in most cases there exist a lot of varieties of a single process. When we look at Expense Approval or Purchase Order, usually we have one path with many different variations depending on the type of order or expense. In the sequence world we have to foresee and program all possible paths. That makes the schema completely unreadable and the process absolutely unchangeable. A small change in the process definition requires changing of all the variations. In the state machine world, however, you have only to draw the main logic of the process and account for all the possibilities. If the logic changes slightly you change only the main part. It's the user who decides which variation to take.

Let's take a look at the first example - Contract Approval - to better understand the problem. As mentioned before, we have to receive three approvals (Legal, Finance and Boss).

The sequential process will look like this:

The state machine process will be:

At this point the sequential process is a better choice. It clearly guides you from start to end. But in many cases user may have his own preferences about the order in which he gets the approvals. He might, for example, know that the lawyer is making many changes and wants him to be first, or knows that boss is going tomorrow on holidays so wants boss to be first). This situation may be different in various departments, so you might like to put all these choices in the process. Now things get complicated. With state machine process we don't have to change anything -the order of approvals is not built in the design.

Sequential process has to have all the choices and alternate paths hardwired, so the final schema will look like below:

Well, now we have all possibilities and the same flexibility as in the state machine workflow.

But what will happen if we add a fourth Approver (2nd Boss, for example)? In state machine we add just one state:

In sequential we have to add fourth person in all branches, plus fourth condition ("if") and new branches. Try to imagine how the chart will look like.

Advantages of state machine workflows

This gives us first very important advantage of state machine workflows: flexibility. Both in the way the workflows are run and in implementing changes.

A quick look at both diagrams instantly shows us also the next difference: readability. Business processes require cooperation of many different people from within the company, as well as from outside (consultants, customers, partners…). Some of them are technical, some are business users. A common platform understandable by everyone is the first important thing to enable cooperation. As mentioned before, businesses today need constant adaption and constant improvement of processes. Not having a possibility for everyone to read the process makes this task very difficult, if not impossible. Business analysts, users and programmers must speak the same language. In order to do it with sequential process, you are forced to add a layer understandable by all. Every change in the process definition requires change in the business layer as well. State machine processes are naturally divided into business layer (states and transitions between them) and technical layer (programmed actions). There is no need to synchronize them or to check whether they fit - it's all in the nature of state machine process.

It's also critical for the user to see and understand the whole process in order to make a decision about next actions. Here readability is the key to management of the process. User, enabled to take control of the process can, and has to, see and understand the whole order, as well as where the process is at the time and what options does the user have. Managers may have a broad picture of the current state of the actions. They can see in which phase all the processes are at any given time. Then they can analyze them, finding states in which the processes get stuck. They can optimize them by finding shortest (or quickest) paths based on current behavior and history.

State machine is the human approach to the workflow world. When we think about processes, we are usually wondering about their current status (state of the workflow) and what will happen next or what should be done next (action). When one draws the process representation it looks like in this picture:

That's exactly what state machine process diagram looks like. Why to translate this into a difficult to read and comprehend sequential workflow:

When we can leave it in the same form it was designed?

To summarize: state machine workflows are the only logical choice in rapidly changing business environment. They provide great flexibility needed for constantly changing workflows. They are clear and relatively simple, which results in readability sought by all participants from users, through business analysts, consultants, partners, customers, to programmers that have to implement the automation of the process. Finally, state machine workflows enable management and optimization.

And, most of all, they are human-centric, representing the way we think of processes and act during their execution.

In this article I have simplified things in order to better show and explain the matter. Most (if not all) workflows may be implemented using either of the process types, but state machine have clear advantages, unless you want people to follow only the narrowly defined paths and be controlled by the process - which in some cases may be desirable.

Paweł Misiurewicz

Datapolis, CEO

Datapolis are already signed up as Silver Exhibitors for the ESP Conference 2013. Why not join them? Contact sponsor@sharepointeurope.com or call +353 91 514501 to learn more about sponsorship & exhibition opportunities.

Stay tuned for more SharePoint content by joining our community or by following us on twitter or  facebook.