What does it cost to start?

The Diagnostic is a fixed price for one process, agreed before any work starts, with no platform commitment attached to it. The ongoing agreement after it is scoped from what the Diagnostic actually found, so nothing is estimated twice.

A thing your process acts on or through: an account, a trade, an invoice, an approval, a shipment. Objects are what Ammonite counts, and what it reasons about.

One moment in which something happened to one or more objects, with a timestamp. An event can touch several objects at once, which is exactly what a single-case view has to throw away.

Do I have to commit to the platform?

No. The Diagnostic stands on its own and ends in a findings report you keep. If you continue from there, that report scopes the agreement, so nothing is estimated twice.

Do I need an implementation partner?

No. If you want a human to do the work, that's the Diagnostic, at a published fixed price rather than an open-ended engagement.

Conformance Checking Without the Jargon

TL;DR: Conformance checking compares what your event log shows with what a process model or rule says should happen. It surfaces skips, extras, and wrong-order paths so you can decide which gaps to fix, allow, or rewrite in the standard.

What you will learn

A jargon-light definition of conformance checking
The difference between model-versus-log perspectives
Plain meanings of fitness and related ideas
How to classify deviations without overreacting
A practical workflow from rules to remediation backlog

What is conformance checking?

Discovery answers "what did we do?" Conformance checking answers "how does that compare to what we said we would do?" You bring two ingredients: an event log (reality) and a normative reference (intent). That reference can be a BPMN diagram, a Petri net, a checklist of ordering rules, or a golden-path model.

Definition: Conformance checking is the process mining activity of diagnosing mismatches between observed event behavior and a normative process model or rule set, highlighting deviations such as skipped, inserted, or reordered activities.

It is diagnostic, not automatically punitive. The output should help humans decide: fix the work, fix the system, or fix the standard.

Why bother if we already have a discovery map?

Discovery maps show everything that happened, including behavior you never intended to treat as standard. Conformance adds a reference frame:

Without conformance	With conformance
"We have 240 variants"	"12% of cases skip mandatory control X"
"There is a loop here"	"The loop violates the no-rework-after-approval rule"
Arguments about the real process	Shared language of allowed versus observed

Use discovery to learn; use conformance to govern and improve against an agreed yardstick.

The three everyday deviation types

Most business conversations need only three buckets:

Skips. Required activity missing (shipped without credit check).
Inserts. Extra activity not in the model (manual spreadsheet step logged as "Adjust").
Order breaks. Activities in the wrong sequence (invoice before goods receipt when policy forbids it).

Some techniques also highlight resource or data rule breaks (wrong role approved; amount over limit). Start with control-flow deviations; add data rules when the basics are trusted.

Academic conformance uses measures such as fitness and precision. Here is the hallway translation:

Term you will hear	Plain question	If it is low…
Fitness	Can the model explain what happened in the log?	Model too strict or log truly wild
Precision (related idea)	Does the model allow only behavior we actually see?	Model too loose (anything goes)
Generalization (related idea)	Will the model still fit unseen but valid cases?	Overfitting to one sample period

You rarely need to quote formulas in a steering committee. You do need to say whether mismatches mean people broke policy or the diagram is fiction.

Where does the normative model come from?

Pick consciously:

Designed standard. The BPMN from the operating model team.
Regulatory / control catalog. Must rules extracted from policy (even without a full diagram).
Agreed discovered model. Last quarter's cleaned discovery map promoted to standard.
Hybrid. Discovery for flexible regions; hard rules for controls.

Warning: Conformance against an unrealistic model produces empty theater. If 80% of cases fail, update the model or admit the standard was never implemented.

A practical conformance workflow

Write the question. For example, "How often do we approve after payment initiation?"
Encode the rule or model. Keep the first version small and testable.
Run conformance on a recent, clean log. Prefer completed cases for fairness.
Quantify. Rate, volume, and (if possible) financial or risk impact.
Sample and explain. Open 10 deviant cases with operations; label root causes.
Triage. Fix process, fix data or logging, or relax or clarify the rule.
Re-measure after changes. Conformance is a loop, not a one-off audit PDF.

Classifying deviations without blame

Use a simple triage table in reviews:

Deviation pattern	Typical cause	First response
Rare + high risk	Exception abuse or fraud risk	Control redesign and monitoring
Common + low risk	Standard outdated	Update normative model
Common + high pain	Broken upstream process	Improvement project
Clustered in one site/system	Local workaround or bug	Targeted fix or training
Appears only after release	Logging or process regression	IT and process jointly debug

Separate visibility gaps (event never logged) from behavior gaps (event logged, rule broken). Conformance cannot address what the log never saw.

Conformance and compliance claims

Process mining conformance supports compliance conversations, but it is not by itself a legal certification, SOC report, or auditor sign-off. Treat results as evidence inputs for human experts. Avoid claiming you are compliant because fitness is high. Fitness can be high against a weak model.

How conformance connects to maps and action

After reading a messy map (A4), conformance tells you which edges violate policy.
After framing discover, check, and enhance (A2), conformance is the check engine room.
After you triage, move to insight-to-action (A6) so findings become backlog, not anxiety.

Key takeaways

Conformance checking compares logs to a normative model or rules to find meaningful mismatches.
Skips, inserts, and order breaks cover most day-one conversations.
Fitness language is about whether the model can explain the log. Interpret with care.
Bad standards create fake non-conformance; validate the reference model.
Triage deviations by risk and cause; then re-measure after fixes.

FAQ

What is conformance checking in process mining?

Conformance checking compares an event log to a process model or rule set to show where real behavior matches or diverges: skipped steps, extra steps, or wrong order. Teams can then separate harmful deviations from acceptable flexibility. It turns "we have variants" into "these variants break rule X."

What is fitness in process mining?

Fitness (in plain terms) asks how well the model can explain the behavior seen in the log. High fitness means most observed cases can be replayed on the model; low fitness means many real cases do things the model cannot account for. That may mean discipline issues or a model that never matched reality.

Do I need a BPMN model for conformance checking?

Not necessarily. You can check against BPMN, Petri nets, or simpler rule sets (for example, "A must precede B"). What matters is a clear normative reference that stakeholders agree represents intended or required behavior, and that the log can observe the relevant activities.

Is every deviation a compliance failure?

No. Some deviations are sensible exceptions, workarounds for broken systems, or signs the normative model is outdated. Classify deviations by risk and value before treating them as audit findings, and involve compliance specialists when stakes are regulatory.

Attribution

Educational synthesis inspired by the open-access Process Mining Handbook (van der Aalst and Carmona, eds., 2022), https://doi.org/10.1007/978-3-031-08848-3. This article is not a substitute for the handbook. It is not legal, audit, or vendor advice.