A grade-out can look like a quality event when it is really a line setup signal. Both produce the same immediate result: product lands in a lower-value category than intended. The response, however, is different. Changing suppliers will not correct a setup problem on the line.
Across multiple shifts at one facility processing the same incoming lot, grade-out rates can vary beyond what product variation explains. The same product may run four hours apart on the same line and produce different rates. The lot did not change. The line configuration did, or it failed to change when the product required it.
Three grade-out signals that indicate setup
A high grade-out rate does not always point to setup. Certain patterns, though, indicate the line is the main variable, not the product.
Grade-out tied to speed
Near the upper end of designed throughput, each item spends less time beneath the sorter's cameras or sensor array. A sorter calibrated for standard line speed can misclassify acceptable items as the belt accelerates because its detection window no longer supports a confident call. Each item is being judged with less data.
When grade-out follows belt-speed changes during a shift, treat the pattern as a setup signal. The product is not poorer at high speed. The sorter is less accurate against its current calibration. Raising the threshold or slowing the belt to a range that preserves classification accuracy addresses the line, not product quality.
Grade-out tied to temperature
Optical sorters and vision systems model color, texture, and surface profile within an expected temperature range. Product characteristics change with temperature. Leafy greens arriving at the sorter three to four degrees warmer than the calibration baseline can appear different to the sensor while still meeting a human quality standard. The sensor is applying a profile built for another temperature range.
If grade-out rises each afternoon on warm days while incoming lot quality remains steady, check sorter calibration temperature first. The product is not necessarily worse later in the day. The sorter may be using the wrong profile for the temperature reaching its sensors.
Grade-out tied to feed rate
Excessive product density on the feed belt makes items overlap or crowd the sensor field. Crowded items interfere with one another's profile signature, so the sorter misclassifies them more often. If grade-out rises after feed rate is increased to recover from a slow start or handle a volume surge, capacity is mismatched. The sorter has a lower effective ceiling than the belt's raw capacity, and exceeding it reduces yield.
Why the pattern is hard to spot
Most grade-out reports lack the line-condition data present when the events occurred. A shift report may give a batch grade-out percentage without showing belt speed, feed rate, intake product temperature, or the point in the shift when counts peaked.
Without that context, the grade-out figure reaches the quality team, and quality becomes the available explanation. That is a reasonable first hypothesis, but not a complete one. Acting on it sends attention upstream to suppliers instead of to the line configuration producing the loss.
When grade-out counts are paired with concurrent line-condition data, the pattern can emerge within a few weeks. Consider a facility processing consistent product through summer and fall that blamed higher afternoon grade-outs on field heat stress in incoming lots. Overlaying line-speed and feed-rate records showed another pattern: a significant share of elevated events clustered in the 90-minute periods after belt speed rose to recover from a slow start, rather than across the entire afternoon. The same field lots, run earlier at standard speed, stayed within the normal grade-out range.
What this excludes
Incoming product quality does affect grade-out. Lots harvested under heat stress, product handled roughly during transit, or varieties more mature than the sorter calibration profile can produce higher rates regardless of setup. We are not claiming setup is the dominant factor in every case, nor that incoming quality is irrelevant. This is not a choice between product and line in every situation.
The point is that product quality and line setup are separate signals, while most grade-out reports combine them into one figure. Without line-condition data linked to each event, the two cannot be separated. The result is an ambiguous number and a default explanation, usually the product.
Grade-out as an ongoing diagnostic
Grade-out rate is useful because it records a quality outcome near real time, at the sorter, while the run is active. The sorter already makes the classification, and SCADA already logs the counts in its historian. What is generally absent is the link between those counts and the line conditions active during the same window.
In most facilities, that link does not call for additional sensors. Belt speed is already in the historian, feed rate is already tracked, and intake product temperature is already logged at the cold room. The missing work is to align those streams and flag grade-out increases that correlate with a specific line condition rather than a shift-wide average.
At Terrantic, grade-out is among the first data sets we connect at a new facility. It quickly shows where product and line settings diverge, and it separates quality issues from setup issues more clearly than shift averages can. That separation identifies which operating decision deserves attention.