How EDGEBIC Composes Step Timing: Queue, Overlap, Transit, in Order

EDGEBIC by User Solutions builds the handoff time between two operations in a fixed three-stage sequence: queue time first, lot streaming second, transit days third. The order is an invariant, not a preference, and knowing it explains almost every surprising gap on a Gantt chart. This post walks the sequence, shows what each stage does to the clock, and runs one routing step through all three with real numbers.

For what each field models, read queue and transit times explained. For how to enter them, see how to set queue, flow, and transit times. This is the mechanism underneath both.

The Sequence

When an operation finishes, the engine holds one value: the moment the last hour of work was consumed. That value then passes through three stages before it becomes the successor's gate.

operation end
   ↓  stage 1: add queue time (shift-aware)
queue-adjusted end
   ↓  stage 2: lot streaming, if configured (REPLACES the value)
streaming gate
   ↓  stage 3: add transit days (calendar or working)
published handoff
   ↓
next operation's earliest start

Two properties of this pipeline do most of the explanatory work.

Stage 2 replaces rather than adds. Streaming does not extend the queue result, it overwrites it. That is what allows the handoff to land earlier than the operation's own end, which is the entire purpose of overlapping.

Stage 3 always applies last. Transit is the outermost wrapper, so anything streaming gained is still subject to the truck leaving. Reversing stages 1 and 2 would let a queue buffer cancel the overlap, and moving transit earlier would let a streamed handoff jump the shipment. Neither is physically sensible, which is why the order is fixed.

Stage 1: The Shift-Aware Queue Walk

Queue time is not added as flat hours. The engine walks the work center's shift calendar forward and consumes queue hours only from time that falls inside an active shift.

The walk is straightforward to follow by hand:

1. Start at the operation's end time with the full queue balance.
2. If the current date has no shifts (a holiday, a weekend, an idle day), jump to the next day's first shift start and try again.
3. Inside a shift, take whichever is smaller: the queue balance, or the hours remaining in that shift. Subtract, advance the clock.
4. Repeat until the balance is zero.

Four hours of queue starting Monday at 14:30 on an 08:00 to 16:00 calendar:

The overnight gap contributes nothing. That is what "shift-aware" means, and it is why a four-hour cure entered on a Friday afternoon can look like a three-day gap on the chart while being exactly right.

If queue time is zero, this stage returns the operation end unchanged. It does not skip forward past a holiday, and it does not need to: the capacity allocator will find the first working slot later in the pipeline.

Stage 2: Lot Streaming Replaces the Value

This stage runs only when the routing step carries a transfer batch size or a flow lag. When it runs, it computes a gate from the upstream operation's start rather than its end, then hands that back as the answer.

The piece-count model computes:

gate = upstream start + setup + min(transfer batch, order quantity) x hours per piece, then plus transfer delay

The flow-lag model computes:

gate = upstream start + flow lag hours, then plus transfer delay

Neither result is compared against the queue-adjusted value from stage 1. It is simply replaced. Take a welding step that starts Monday 08:00 and ends Monday 10:00, with two hours of queue for cooling and a six-hour flow lag:

• Stage 1 walks two queue hours from Mon 10:00 and produces Mon 12:00.
• Stage 2 computes Mon 08:00 plus six hours and produces Mon 14:00, which becomes the answer.

The cooling queue is gone. Not overruled by a later value, gone: had the flow lag been two hours instead of six, the gate would be Mon 10:00 and the successor would be released before the queue buffer had elapsed. This is the interaction the configuration audit flags whenever a step carries both a flow lag and a queue time, and the fix is always to move the wait into transfer delay or onto the downstream step. The model detail lives in transfer batch versus flow overlap.

Streaming also reads actuals rather than plan when a job is already running. If the upstream step is complete, the gate is its actual end plus handling. If it is running and logged piece counts have crossed the batch threshold, the gate is the end of shift on the crossing day, a deliberately conservative choice given that pieces trickle in through the day.

Stage 3: Transit Days on Top

Transit is the last thing applied, and it behaves in one of two modes chosen per routing step.

Calendar days add flat days. A handoff at Thursday 12:00 with three calendar transit days becomes Sunday 12:00. The result may land outside working hours, which is correct for a vendor whose calendar is not yours.

Working days count only days that have at least one active shift on the sending work center, then resolve to the end of the last shift on the final counted day. From the same Thursday noon: Friday counts as one, the weekend is skipped, Monday is two, Tuesday is three, and the handoff resolves to Tuesday 16:00.

That end-of-shift snap is the detail worth internalizing. Working-day transit never releases a successor mid-morning: it releases it after the last shift of the final transit day, so in practice the next operation starts the following morning. Which is why a single working transit day is worth roughly a full day of lead time, and why it is the most expensive checkbox on the routing step.

Material steps skip this stage entirely, because they already absorb their own transit when their dates are built. Setting transit days on both a material step and the operation that follows it counts the same delay twice.

Publication and the Join Rule

The composed value is published as the handoff for that step, and the next operation reads it when deciding its own earliest start.

When several branches feed one operation, each publishes its own handoff and the engine keeps the latest of them. That is the correct behavior for a join: an assembly cannot start until every feeding branch has delivered. It also means shortening a non-binding branch changes nothing at all, which frustrates a lot of improvement work aimed at the wrong sub-assembly. Find the branch whose handoff matches the assembly's start before you optimize anything. Sub-assemblies in the designer shows how those branches are drawn.

One Step, All Three Stages

A 500-piece machining order feeding deburring. Machining starts Monday 08:00 on a Monday-to-Friday, 08:00 to 16:00 calendar.

Stage 1. Machining's own end is late Tuesday. One hour of queue is walked from there, producing a queue-adjusted end.

Stage 2. Streaming fires and discards that value. Fifty pieces exist 1 + 50 x 0.05 = 3.5 hours after the machine starts, so the gate is Monday 11:30. Transfer delay adds half an hour: Monday 12:00. The queue hour never reaches the schedule.

Stage 3, calendar mode. One calendar day from Monday 12:00 is Tuesday 12:00. Deburring's first feasible slot is Tuesday at noon. The job still gains most of the overlap.

Stage 3, working mode. One working day from Monday counts Tuesday, resolving to Tuesday 16:00. Deburring cannot start until Wednesday morning. The 22-hour overlap the transfer batch just bought has been spent entirely on the transit setting.

Same routing, same batch size, same delay. One checkbox moves the downstream start by 20 hours. That is the practical reason to know the sequence: the fields do not contribute equally, and the last stage is the loudest.

Reading a Gap on the Chart

When a gap looks wrong, work the stages in order and stop at the first one that explains it:

1. Does the step have a queue? Convert it to calendar time by walking the shifts. Four hours from Friday afternoon is genuinely three days of chart.
2. Does it have streaming? If so, the queue is not in the plan at all, and the gate came from the upstream start rather than its end.
3. Does it have transit days, and in which mode? Working mode snaps to end of shift. This is the single most common source of an unexplained extra day.
4. If none of the above accounts for it, it is capacity. The composed handoff is an earliest start, not a booking, and a loaded work center supplies the rest of the gap. Why a schedule has idle gaps covers that case, and finite versus infinite capacity scheduling covers why the allocator behaves that way at all.

What to Read Next

Queue and transit mistakes catalogs the configurations that produce plans the floor cannot hit. Lot streaming explained covers stage 2 in depth. For where step timing sits among the engine's other decisions, see the complete guide to EDGEBIC.

Bring one routing with a queue, an overlap, and a vendor send on it to a demo of EDGEBIC and we will walk the three stages against your own dates.