All articles
Business Analysis
5
min read
How to write a process description

Thom Uffing
Process Innovation

A process description explains a whole process in text: its purpose, scope, trigger, roles, steps, inputs and outputs, decision rules, exceptions, owner and version. It does not replace the process model. It is the text around the model, covering what a diagram cannot show on its own, such as why the process exists and who maintains it.
The question at the first review
The process model is on the screen and the client recognizes the flow. Then the process owner asks: "Where does it say who may approve an exception? And which version is this?"
The model shows that an exception exists and which lane handles it. It does not say why the rule is there, what counts as an exception or who decides when the rule changes. That information usually ends up in a process description: a document the client can read without knowing BPMN and that an auditor or a new team lead can use as a reference.
Process description, process model, SOP and work instruction
The terms overlap and organizations use them differently. A common distinction is:
Process model: the flow across roles, with tasks, decisions and hand-offs. Usually in BPMN 2.0.
Process description: the text that goes with the model. It describes the whole process, its context and the agreements around it.
Procedure or SOP: who does what and when to carry out the process, often across several roles. See how to write an SOP from your process model.
Work instruction: how to carry out one task in detail, sometimes with screens and field names.
Read this way, the model and the description both describe the process itself: the model shows the flow and the description adds the purpose, the reasons and the agreements. Procedures and work instructions sit below them and describe how the work is carried out.
What a process description contains
Purpose: why the process exists and what result it delivers.
Scope: where the process starts and ends, and what is explicitly out of scope. A SIPOC made at the start of the project already holds most of this.
Trigger: the event that starts the process, such as an incoming request, a date or a signal from another process.
Roles: who is involved and what each role is responsible for.
Steps: the main steps in order, at the level of the tasks in the model. The detail of how each step is done belongs in procedures and work instructions.
Inputs and outputs: which documents, data and systems the process uses and what it produces.
Decision rules: the conditions behind each decision in the model, written out in full.
Exceptions: what happens when something deviates from the normal path and who handles it.
Owner: who is accountable for the process and approves changes.
Version: the version number, the date and the version of the model it belongs to.
Some organizations add performance indicators, risks and controls or related processes. That is common in quality management and audit settings.
How to gather the input
Start with what already exists. Policies, older procedures, forms, system screens and email templates often contain the decision rules and the inputs and outputs. Reading them first makes the interviews shorter and more specific.
Interview the people who do the work. A process interview with each role gives the steps, the hand-offs and the exceptions that are not written down anywhere.
Talk to the process owner separately. Purpose, scope and the reasoning behind decision rules usually come from the owner rather than from the people who carry out the steps.
Write down contradictions as open questions. When two sources describe the same step differently, note both and settle it in the validation round instead of choosing one yourself.
Writing it alongside the model
A common order is to build the model first and then write the description section by section, using the model as the outline. The task names in the text match the task names in the model. Each decision rule refers to the gateway it belongs to. Each exception refers to the path in the model where it is handled.
This keeps the two aligned. A reader can switch between text and model without having to translate, and a reviewer can check whether every gateway has a written rule and every step in the text exists in the model.
Keeping it in sync
Process descriptions often go out of date because they are kept in a different place than the model. A few practices help:
One owner for model and description together, so a change to one leads to a check of the other.
The model version in the description, so a reader can see which model the text describes.
A change log with the date, the change and who approved it.
A review moment, for example once a year or after every change to the process, whichever comes first.
Common pitfalls
Repeating the model in prose. A description that only lists the steps adds little. Its value lies in the purpose, the decision rules, the exceptions and the ownership.
Mixing levels. Field-by-field instructions belong in a work instruction, not in the process description.
Decision rules without a source. If nobody can say where a rule comes from, it is worth asking the owner before writing it down as fact.
No version link to the model. Without it, it is hard to tell later whether the text and the model still describe the same process.
When you can skip it
For a simple process with one or two roles and few decisions, a well-annotated model can be enough. A separate process description is most useful when the process is audited, when several departments are involved, when people without BPMN knowledge need to read it or when the process changes regularly and needs a record of its agreements.
How ModelMatic does it
In ModelMatic you start from stakeholder interviews and existing documents. ModelMatic writes up the input and builds a BPMN process model from it, in which every step points back to its source. Points that are still unclear are listed as open questions for the validation round. From the validated model it generates the SOP, the RACI matrix and the requirements, so the documents around the model start from the same version.
At Bovemij, a process description that normally took about four weeks was finished in about two with ModelMatic.
Book a demo to see how an interview becomes a validated process model and the documentation around it.




