Throughput

There are many P-KPIs that the platform could report to the dashboard for a given station. Depending on the type of station, a specific subset of P-KPIs might be more appropriate. If your station is reporting test data by using SDK functions that track the unit in and out of the station, throughput metrics help you understand how much work your stations are performing.

Absolute Throughput

Absolute throughput is the count of unit events tracked in and out of a station within a given time range.

Every track-in and track-out event is counted individually. If a unit enters a station three times (e.g., due to retries), that counts as three inputs. This reflects the total workload the station handled — its processing capacity.

This is distinct from the unique unit count used by Yield and Retry Rate, which deduplicate by unit ID.

UPH (Units Per Hour)

UPH is the rate of production — how many unit events are processed per hour.

$$\text{UPH} = \frac{\text{Total IO events}}{\text{hr}}$$

On the dashboard, the UPH chart shows total inputs and outputs per hour as separate series. Each track-in event increments the input count and each track-out event increments the output count for that hour.

Why total events, not unique units?

UPH counts every event rather than unique units for two reasons:

1. Stability — A unique-unit UPH would change retroactively. If a unit fails at 01:30 and retries at 02:15, you'd have to decide which hour "owns" that unit. Assigning it to the first hour means the second hour's count drops; assigning it to the last means the first hour's count changes after the fact. Total events avoid this — each event has a fixed timestamp and never moves between buckets.

2. Capacity — Total events reflect the station's actual processing capacity. A station running 60 cycles per hour is doing 60 cycles of work, regardless of whether 5 of those were retries. This is the number that matters for capacity planning and bottleneck analysis.

Relationship to other metrics

The UPH chart and Yield chart will show different denominators when retries are present — this is expected. UPH reflects workload; Yield reflects quality.

Note that this metric can be used at the station level, station group level, or line level.

Example

In this example, we'll look at throughput for one station over a one-hour window.

Each bar below represents a single station run — spanning from the track_unit_in() SDK call to the track_unit_out() SDK call. These are the events that Autumn counts when computing throughput.

• Total Inputs: 6 (every track-in event)
• Total Outputs: 6 (every track-out event)
• Unique Units: 4

$$\text{UPH} = \frac{6 \text{ events}}{1 \text{ hr}} = 6$$

For comparison, the Yield calculation for the same data uses 4 unique units as its denominator — not 6.