What data you need
Bring a supported spreadsheet, table or process log that records each item’s steps and times.
The three things glasstarn needs
Section titled “The three things glasstarn needs”A supported spreadsheet, table or process log needs these three things, whatever the workflow, system or column names:
| Role | What it is | Examples |
|---|---|---|
| Item | The thing you follow from start to end | Invoice number, ticket ID, claim number, job applicant, permit application |
| Step | What happened to it | A step, status or action: “Created”, “Approved”, “Paid” |
| Time | When it happened | A date, or a date and time |
Each row is one step for one item. An invoice that went through five steps has five rows.
glasstarn works out which column plays which role, and your agent tells you what it used. Column names are a hint, not proof, so it asks when the meaning is unclear, and you can correct it at any point.
Extra columns make it smarter
Section titled “Extra columns make it smarter”Keep every column you have. glasstarn keeps them all, and you can filter and compare on any of them.
| Extra column | What it adds |
|---|---|
| Who did the step | See who handles what, and where work waits for one person |
| Amount | Put a value on the items in each finding |
| Vendor, customer, region, team, product | Find which groups are slower, through likely drivers |
| End time of a step | Separate the time a step takes from the time spent waiting for it |
| Status that changes over time | Filter on items that were ever in a status, or are still in it |
If a column holds personal data you would rather leave out, tell your agent. It leaves that column out of the project.
Files glasstarn reads
Section titled “Files glasstarn reads”- Spreadsheets: Excel (.xlsx, .xlsm, .xls, .xlsb) and LibreOffice (.ods). In a workbook with several tabs, glasstarn picks the tab that looks like an event history and says which.
- Text tables: CSV and other separated text (tab, semicolon, pipe), fixed width text and SAP list output, in any common Windows encoding.
- Zip files of any of these.
- Process mining formats: XES, Parquet and OCEL 2.0, if you already have them.
Files are read the way people export them: title rows above the header are skipped, merged cells are filled in, and dates such as 3/5/2026 9:41 AM are recognised.
The browser accepts uploads up to 100 GiB (about 107 GB). Whether a file that large can be analysed depends on its format, its contents and the memory available on your computer. Workbooks are expanded into tables in memory; a large CSV or Parquet export may be more practical. The upload limit is not a tested capacity guarantee for every format.
When the steps are codes
Section titled “When the steps are codes”If the step column holds codes such as 10, 20, 30, your agent asks what they mean, or looks for a lookup file among what you shared. It renames them for you. You do not need a new export.
When a file cannot be read
Section titled “When a file cannot be read”glasstarn tells your agent what kind of file it seems to be, without reading its rows. Your agent then tells you what would work. Common fixes:
- A password protected workbook: send a copy without the password.
- A PDF report: export the same report to Excel instead.
- A table with notes above it: usually fine, glasstarn skips them. If not, delete the notes.
When the data is spread over many tables
Section titled “When the data is spread over many tables”Systems such as SAP or Oracle keep one process across many tables. You do not need to work out which ones. glasstarn already knows how many common systems store their data, so your agent can tell you exactly which tables to export, check them when they come back, and put the history together. See Don’t know what to export?.