Quick answer: Process mining rebuilds how a process ran from the timestamps your systems already hold. For a mid-market company it pays when one named process is costing time or money and runs through a system that keeps an event history.
These are the questions operations leaders at mid-market companies ask us before they commit to process mining, answered as directly as we can. Where an answer has a number in it, the number is the range we publish and explain in the guide linked beneath it.
What it is, and whether you need it
What is process mining, in plain terms?
Process mining reads the timestamps your systems already record and rebuilds the route each order, invoice or patient actually took. Every ERP, CRM and ticketing tool logs when a step started and finished. Put those records in sequence and you get a map of the real process, with the time spent between each step. The value is in the gap between that map and the process people believe they run.
Is process mining worth it for a mid-market company?
It is worth it when one process is costing you something you can name, such as late deliveries, rework or cash tied up in stock, and when that process runs through a system that keeps an event history. It is a poor fit when the work happens mostly in email and spreadsheets, because there is no log to read. Size matters less than volume: a few thousand cases a year is enough to show where the delays collect.
Do I need a process mining platform, or just better process documentation?
Documentation tells you how a process is meant to run. Mining shows how it ran last quarter. If your question is “what are the steps and who owns them”, documentation answers it for far less money. If your question is “why does the same order take four days one week and eleven the next”, only the event data can tell you. Many mid-market teams need the second answer for one or two processes and documentation for the rest.
More detail: Process mining vs business process management.
Time, cost and data
How long does process mining take to implement?
Platform-led programmes typically take 3 to 9 months from kickoff to a first measurable outcome. A service-led engagement is shorter because nobody has to select, buy and learn a tool first: ours target a first measurable result by week 8, after a discovery of 2 to 3 weeks. Complex multi-system landscapes, poor data quality or choosing a tool at the same time can stretch a programme to 9 to 18 months.
More detail: Process mining implementation timeline.
What does process mining cost, and when does it pay back?
Three things drive the cost: the platform licence if you buy one, the work of extracting and cleaning the event data, and the time your own people give to acting on what is found. The last is the one most budgets leave out. On return, programmes typically pay back within 3 to 9 months, at a first-year return of 3:1 to 10:1. Treat any quote without a named process and a measurable target as a guess. A Zenotris engagement on one process costs between $10,000 and $50,000, from discovery to the first measurable result.
More detail: Process mining implementation cost for mid-market companies.
What data do we need before we start?
You need three fields for each event: an identifier for the case, the name of the step and a timestamp. An order number, “goods issued” and a date and time is a complete record. Most ERP and warehouse systems hold this already, though it is rarely in one table. Expect the first weeks to go on finding it, joining it and checking that the timestamps mean what their column names say.
Can we measure process efficiency without a process mining tool?
Yes, for a first look. Export the start and end timestamps for a few hundred recent cases, calculate the elapsed time for each and sort them. The slowest tenth usually shares a cause you can find by reading the cases. This will not show you rework loops or the order in which steps happen, and it stops scaling quickly, but it is enough to decide whether a deeper study would pay.
Build, buy or hire: how the three routes compare
Most teams reach the same fork early: build the capability in-house, buy a platform, or bring in a consultancy. The right route depends mainly on how soon you need a result and on the data skills you already have.
| Route | Time to a first useful result | Suits | Main drawback |
|---|---|---|---|
| Build in-house | 12 months or more | Teams with strong data engineering and unusual data sources | Ongoing engineering cost and no vendor support |
| Buy a platform | 6 to 9 months of capability building | Organisations with an analytics team already in place | A slow ramp-up before anything changes |
| Hire a consultancy | 3 to 9 months | Teams without process intelligence expertise that want the capability handed over | You depend on the firm until the handover |
The full comparison is in Process mining: build vs buy vs hire.
Getting it approved and making it last
How do I build a business case for process mining or automation?
Start with what the delay costs today. Pick one process, count the cases per month and put a price on each day of delay, each rework and each missed deadline. Then set a target you can measure in 90 days, for example cutting the time from order to dispatch by a fifth. A case built on one process and one number gets approved; a case built on “visibility” does not.
More detail: Process mining ROI calculator.
What should I ask a process mining consultancy before hiring them?
Ask which process they would start with and why. Ask what they will need from your IT team in the first month, what you will own at the end and how the result will be measured. Ask to see an example of a finding that led to a change, with the before and after figures. A firm that leads with the platform it resells is selling you a licence.
More detail: How to choose a process mining implementation partner.
How do we stop improvements being reversed after six months?
Improvements slip when nobody watches the number once the project ends. Give each fixed process a named owner and one measure that is checked every week. Keep the measurement running on live data so a drift shows up in days. Most reversals trace back to a workaround that returned quietly under pressure, and a weekly figure is what exposes it.
More detail: Real-time process monitoring.
See what it finds in your operation. See how it works →