FreeNo sign-upYour records stay on your computer

Process Mining Tool

Read the event log your systems already keep, rebuild the paths items really take through the process, and see which paths carry the errors. No log? Type in the paths of items followed by hand.

Where the paths come from

Read the paths from
What is a path? An illustration with three made-up items

The event log

  • Columns: the item's number, what happened, and when. Optional: an error column, and one to compare by, such as the supplier.
  • Errors can be events of their own (a return, a correction) or values in an error column. The file stays in your browser.
Or paste rows from a spreadsheet

Your records never leave your browser

The log is read on your computer. Only the paths it found, with their counts, are kept in your progress file. To write the diagnosis, only counts and rates are sent to our server, by place in your list. How it works

About the tool

Why it exists

Find where errors start before asking why

Errors rarely start everywhere. This tool reads the records your systems already keep, rebuilds the path each item really took through the process, and shows each path's share of the work and of the errors. Paths your written procedure does not allow for are workarounds, and they often carry most of the errors.

It works on its own, and as stage 2 of our playbook for reducing errors and defects, where the paths with the most errors go on to stage 3, to prove why the errors happen there.

The method

How the tool works

The event log. One row per event: the item's number (an order line, a ticket, an invoice), what happened, and when. The tool guesses which column is which from the headings and the values, and you can change it.

Paths. Each item's events are put in time order (the order of the rows when times are equal or missing), and repeats in a row are counted once. Each distinct sequence of steps is a path. The ten most common are shown; the rest are grouped as other paths.

Errors. You tick the activities that show something went wrong, such as a return or a correction. They mark the item as one with an error, and are left out of its path, so a returned item and a good one that went the same way share a path. If your log has an error column instead, every value other than empty, no, 0 or OK marks the item, and each value is listed as an error of its own.

Stage 1's defects. In the playbook, each error is tied to one of stage 1's defect types (the tool matches them by name, and you can change it). The tool then compares each type's share of items in the records with stage 1's, and shows which path carries most of each type.

Which path carries the errors. A path stands out when its items go wrong at least 1.5 times as often as all items together. The same test picks out single steps linked to errors, and values of the column you compare by, such as one supplier.

Written procedure. You tick every path your written procedure allows; there can be several, such as one per sales channel. A path with many errors that nobody wrote down is a workaround.

Without a log. Follow 50 to 100 items through by hand, group them by the steps they took, and type in each path with its count and how many went wrong. The paths must add up to the items followed, since each item took one path.

What leaves your computer. The log is read in your browser and never uploaded. Only the distinct paths with their counts are kept in your progress file, and only counts and rates, by place in your list, are sent to our server to write the diagnosis; the server keeps nothing.