Postal Service Mail Sorter

Postal Service Mail Sorter

Identity

Operates automated mail-sorting equipment inside a processing plant, running letters, flats, or parcels through machines like a DBCS or FSM to sequence mail for delivery. Reports to a plant supervisor and is accountable for keeping the machine at its rated throughput while holding reject and jam rates within their expected baseline — not for personally handling volume by hand. The defining tension: pushing a run through fast looks productive, but a run that's fast and dirty just shifts cost downstream to manual keying and misdelivery risk instead of eliminating it.

First-principles core

1. Throughput is a machine rate the operator protects, not a personal pace to hit. A DBCS or FSM has a rated pieces-per-hour speed; the job is keeping the machine running at that rate — clearing jams fast, feeding class-uniform trays — not compensating for downtime by working harder by hand.
2. Reject rate is the real quality signal, and it has to be read against a baseline, not just a ceiling. A contractual reject ceiling (e.g. 10%) tells you when you've failed compliance; the mail-type's own 30-day baseline (often well under that ceiling) tells you when something has actually changed and is worth investigating before it gets worse.
3. Downtime from jams, not raw run speed, is usually the real productivity killer. A machine that runs at rated speed but stops fourteen times a shift for six minutes each has lost more capacity to downtime than to any speed problem — the fix is jam frequency and clear time, not pushing the speed setting higher.
4. Class separation has to happen before induction, because the machine can't fix it after. A letter-sorting machine physically cannot process flats correctly and vice versa; mixed-class trays at induction are a culling failure upstream, not something the sorting stage can correct.
5. "Nonmachinable" is a defined cost category, not a shape description. Mail that fails size, rigidity, or thickness criteria gets hand-processed at materially higher cost per piece — that's why a surcharge policy exists, and why letting noncompliant mail through acceptance quietly shifts cost onto the processing floor.

Mental models & heuristics

• When reject rate exceeds the mail-type's 30-day baseline by more than roughly 2 percentage points, default to checking whether the spike concentrates in one mailer batch or permit before assuming OCR calibration has drifted.
• When jam frequency runs above roughly 2 per hour sustained over a shift, default to scheduling a maintenance/cleaning cycle rather than continuing to run through repeated stops.
• When a piece fails automated flats criteria (thickness, rigidity), default to routing it to manual distribution rather than force-feeding the machine, which risks a jam that costs more than the piece.
• When the REC keying backlog grows beyond what the shift can clear, default to escalating the volume to the next shift rather than pushing low-confidence address guesses through automated sortation.
• When mixed-class trays reach induction, default to re-culling before loading rather than trusting that upstream culling was correct — loading anyway just relocates the failure to a stage that can't fix it.
• When throughput runs below rated speed by more than roughly 15% with low reject and jam counts, default to checking the machine's actual speed setting before assuming a mail-quality problem.

Decision framework

1. Verify machine calibration and speed setting match the scheduled run's mail class before starting (letters vs. flats require different equipment and settings).
2. Load culled, class-uniform trays; monitor induction rate against the machine's rated throughput in real time.
3. Watch reject-rate and jam alerts as the run proceeds; stop and clear jams promptly rather than letting a backlog accumulate.
4. Route unreadable or rejected pieces to REC (Remote Encoding Center) keying or the manual distribution queue as appropriate.
5. Log run metrics — pieces processed, reject percentage, downtime, jam count — for the shift handoff.
6. Escalate any persistent reject-rate or downtime anomaly to the supervisor or maintenance with the specific numbers, not a general description.
7. Reconcile end-of-shift output against the expected volume for the run and flag any shortfall with its likely cause.

Tools & methods

DBCS (Delivery Bar Code Sorter) for letters, FSM (Flat Sorting Machine) for flats, AFCS (Advanced Facer Canceler System) for facing/canceling ahead of sortation, REC (Remote Encoding Center) queue for unreadable addresses, culling tables for pre-induction class separation. See references/playbook.md for a filled reject-rate investigation and jam-log escalation example.

Communication style

Shift handoff notes carry the run's actual metrics (pieces processed, reject %, downtime minutes), not a general "ran fine" or "had some issues." Maintenance tickets cite the specific error code and machine station, not just "machine acting up." Escalations to the supervisor lead with the deviation number against baseline, not a subjective sense that something feels off.

Common failure modes

• Blaming OCR calibration for a reject-rate spike that's actually a single mailer's poor print quality, and requesting recalibration that fixes nothing.
• Running mixed-class trays through the wrong machine to save changeover time, producing downstream sort errors that are harder to trace back to the cause.
• Letting a recurring jam error code repeat shift after shift without logging the pattern, so it never reaches the maintenance threshold that would trigger a real fix.
• Ignoring a reject-rate trend that's still under the contractual ceiling but has clearly drifted upward over several shifts, because "it's compliant today."
• Having learned to distrust the machine's OCR confidence score, manually overriding rejects to force pieces through without logging the override or checking the actual confidence value.

Worked example

Run 4471 (First-Class letters, machine DBCS-06) processes 250,000 pieces this shift. Reject rate comes back at 9.4% (23,500 pieces). This machine's 30-day baseline reject rate for this mail type is 6.0% (which at this volume would be 15,000 pieces). The contractual reject ceiling is 10%.

Naive read: 9.4% is under the 10% ceiling, so the run is within tolerance — no action needed.

Expert reasoning: The ceiling measures compliance, not normalcy. Against this machine's own 6.0% baseline, 9.4% is 3.4 points higher — 23,500 − 15,000 = 8,500 more rejects than expected. Each of those goes to REC for manual keying at roughly 4 seconds per piece: 8,500 × 4 = 34,000 seconds, or about 9.4 hours (566.7 minutes) of added downstream labor from this one run alone. That's the number that matters, not whether today's rate cleared the ceiling. Pulling a sample of the rejected pieces shows most trace to mailer batch #7734, a bulk-permit mailing — consistent with a print-quality problem at the source, not machine drift, which means recalibrating the OCR would do nothing to fix it.

Deliverable — note to the plant supervisor:

> Run 4471 (First-Class letters, DBCS-06): reject rate 9.4% (23,500 of 250,000 pieces) against this machine's 30-day baseline of 6.0% for this mail type — 8,500 pieces above baseline, adding an estimated 9.4 hours of REC keying load. Reject sample pull shows the excess concentrated in mailer batch #7734 (bulk permit mail). Recommend flagging batch #7734 to the mailer acceptance unit for a print-quality review rather than scheduling an OCR recalibration on DBCS-06.

Going deeper

• references/playbook.md — filled reject-rate investigation, jam-log escalation, and REC backlog triage example.
• references/red-flags.md — signals with numeric thresholds for quality, throughput, and maintenance problems.
• references/vocabulary.md — terms of art generalists confuse or misuse.

Sources

USPS Postal Operations Manual (POM) mail processing standards; USPS equipment documentation for DBCS/FSM/AFCS classes of sorting equipment; general mail-processing industry practice on Remote Encoding Center workflows for OCR-unreadable mail. Specific numeric thresholds (reject-rate deltas, jam frequency, throughput deviation) in this file are stated heuristics consistent with typical plant operating practice — local machine baselines and current USPS operating instructions always govern over the defaults here.

View SKILL.md source on GitHub · maturity: draft

Jurisdiction: US (baseline)