03/Process telemetry

Know whether the documented process is what actually happens.

Real run data, checked against the process you designed.

Bring in run data from the systems the work happens in, compare it against the process you documented, and see precisely where the two have drifted apart.

Onboarding v2.4 · runs
Live
1,533 runs · last 30 days
Snowflake · webhook
Submit
41% out of order
Credit check
2d 06h
Manager approval
Fulfillment
On the designed path 1,284 · 84%
Undocumented path taken 212 · 14%
Required approval skipped 37 · 2%
SOC 2 Type 2 compliant Trust center

Connecting run data

Bring in run data from the systems you already use

Run data arrives from the systems the work already happens in and is matched to the process model, so each execution is a path across a map you recognize.

  • Warehouse, lake, S3, SQL, spreadsheet and webhook sources
  • Runs matched to the designed process automatically
Runs · Onboarding v2.4
1,284 runs matched to the design On path
last 30d
212 runs took an undocumented path Deviation
14%
37 runs skipped a required approval Breach
2%
1,533 runs ingested · Snowflake + inbound webhook

Gaps against the design

See where real runs differ from the documented process

Deviations are shown on the process itself, not in a table of exceptions, so the gap between the documented path and the real one is visible at a glance.

  • Live data compared against the designed process
  • Skipped, reordered and undocumented steps surfaced
Designed path vs observed path

Designed

Submit
Credit check
Approve
Fulfillment

Most common actual

Submit
Approve
Credit check
Fulfillment
Credit check runs after approval in 41% of cases

Time and bottlenecks

See which steps are actually consuming the time

Volume, duration and rework across thousands of runs point at the steps worth changing, rather than the ones people complain about loudest.

  • Bottleneck, rework and cycle-time patterns across runs
  • Findings feed straight into process improvements
Step timing · median
Credit check 4h 12m
Manager approval 2d 06h
Document prep 1h 40m
Fulfillment 3h 55m
Manager approval is 71% of total cycle time

Documentation you can prove is accurate

The map is checked against real runs, not assumed.

Drift spotted while it is still cheap to fix

You see the process diverge while it still matters.

Effort aimed at the real bottleneck

Evidence decides what to change, not the loudest voice.

Common questions

Where does the run data come from?

From wherever it already lands: a data warehouse, a data lake or S3, a SQL database, a spreadsheet export, or an inbound webhook posted by the system doing the work. Nothing has to be re-instrumented first.

Is this process mining?

It covers the same ground as conformance checking, comparing real executions against a process model, with one difference that matters: the model is a process you captured and maintain in Sapeum, not one reverse-engineered from logs alone.

Do we need event logs already in place?

You need some record of what happened, in some system. If runs are already landing in a warehouse, a database or a spreadsheet, that is usually enough to start.

What happens when telemetry disagrees with the map?

That disagreement is the point. A deviation is either a map that is out of date or a process that is not being followed, and both are worth knowing. Either can be taken straight into a future-state design.

Can we see this per team or per region?

Runs carry the attributes they arrive with, so timing and deviation patterns can be read by whatever dimension your source data already has.