Your data knows you’re losing money. Kea shows you where.
Kea reads the event log your systems already keep and draws the process as it actually ran, repeats and waits included. Load a CSV, get the map, click anything to see the cases behind it.
A desktop app for macOS and Windows. It reads CSV or XES event logs and needs no process model. Nothing leaves your machine.
SEE IT IN A MINUTE · A public log: 150,370 traffic fines over 13 years
THE MAP
The map is drawn from the log, not from the process document.
The video used a real public log. The rest of this page follows a smaller order-to-cash log we made up, 12,480 orders, so the numbers stay easy to follow. Kea drew this map from three of its columns. The orange loop is the rework.
This is the whole input.
An event log is a table with one row per event. Kea needs three of its columns: which case the row belongs to, what happened, and when.
Every other column becomes an attribute you can chart and filter by. Order A2052 is the one you will meet again in the walkthrough below.
| Case ID | Activity | Timestamp | Attributes | ||
|---|---|---|---|---|---|
| order_id | activity | timestamp | amount_eur | channel | region |
| A2052 | Order received | 2025-03-04 09:12 | 196.40 | web | North |
| A2052 | Credit check | 2025-03-04 11:47 | 196.40 | web | North |
| A2052 | Manual review | 2025-03-06 15:02 | 196.40 | web | North |
| A2052 | Credit check | 2025-03-10 08:30 | 196.40 | web | North |
| A2052 | Manual review | 2025-03-12 16:40 | 196.40 | web | North |
| A2052 | Credit check | 2025-03-17 09:15 | 196.40 | web | North |
| … and about 73,000 more rows for the other 12,479 orders. | |||||
HOW AN ANALYSIS RUNS
From 12,480 orders to the one loop that costs three days.
The same order-to-cash log, in the three panels you would open first. Each is redrawn from a real view in the app, and a click in any of them filters the others.
Group the cases by how they actually ran.
Kea clusters cases with similar event patterns and draws each cluster as bands: one column per step, width by frequency. C1 and C2 are the two ways an order is meant to go. C3 is 1,276 orders that went through Credit check at least twice.
Drag K to split or merge clusters. The recommended K is computed when the log is imported.
Open the cases behind the pattern.
A cluster is not a summary. It is a list of real orders with real timestamps. Click C3 and the Cases panel shows them on a timeline: A2052 went through Manual review twice and took 6.4 days longer than the median.
Selections are transient; the next click replaces the last. Pin one as a filter when you want it to stick.
Find what the slow cases have in common.
Attribute histograms show how a value is spread across the selection. Here the orders stuck in Manual review sit between €180 and €240, right on the threshold that sends them to a person. That is a rule to change, not a team to push harder.
The Comparison panel ranks which attributes separate one group of cases from another, by effect size, with p-values. Pure statistics, no AI involved.
Built for data you are not allowed to upload.
Kea is a desktop application. There is no account, no tenant and no upload step, because the logs that matter most are usually the ones you cannot send anywhere.
More detail on what the app does and does not send is on the data handling page.
| Runs on | macOS and Windows. A desktop app; nothing to host. |
|---|---|
| Reads | CSV or XES event logs. Three columns are required: case, activity, timestamp. The rest come along as attributes. |
| Network | Not required. Your log never leaves the machine. |
| AI features | Optional, and local. Without a local model the analysis is unchanged. |
| Exports | Any selection as XES or CSV, plus the analysis itself as a file a colleague can open. |
| Licence | Per machine, verified offline. No login. |
Questions people ask before a demo.
If yours is not here, ask it on the booking form and we will answer before the call.
What does my log need to contain?
Three columns: a case identifier, an activity name and a timestamp. Anything else in the file comes along as an attribute. CSV and XES are both accepted.
Do I need a process model first?
No. The map is drawn from the log.
How large a log can it handle?
Logs with millions of events. The public log in the video has 150,370 cases.
Where does my data go?
Nowhere. It stays on the machine that runs Kea, which does not need a network connection. What the app does and does not send.
Do I need the AI features?
No. They are optional and run on a local model if you install one. The analysis does not depend on them.
Which computers does it run on?
macOS and Windows. It is a desktop app, so there is nothing to host.
How is it licensed?
Per machine, verified offline. How licensing works.
What happens on the demo call?
Thirty minutes. You bring a sample of a log, or we use the public traffic-fines log from the video. We import it, draw the map and look at the first loop or wait together.
Bring a sample of your log to the call.
In thirty minutes we import it into Kea, draw the map and look at the first loop or wait together. If you cannot share data yet, we use the public log from the video instead.