Overlapping Operations with Lot Streaming: An EDGEBIC Example

Lot streaming is the scheduling technique that lets a downstream operation begin before the upstream operation has finished the whole order, by moving work forward in transfer batches instead of one monolithic lot. The payoff is measured in makespan: in this walkthrough, a 1,000-piece cut-and-drill order finishes Thursday 15:00 scheduled serially, and Wednesday 14:15 with a 100-piece transfer batch. Same machines, same hours of work, one configuration field, one day earlier. EDGEBIC by User Solutions supports two distinct overlap models, and this example runs the same order through both so you can see exactly where the numbers come from.

The Order: 1,000 Brackets, Two Operations

Product Bracket-HD-1000 routes through two work centers, each a single machine on one Day Shift, 08:00 to 16:00:

The order is MO-77: 1,000 pieces, starting Monday June 22 at 08:00, due Friday June 26.

Cutting needs 10.5 hours against an 8-hour shift, so it spans Monday plus Tuesday morning, ending Tuesday 10:30. Drilling needs 20.5 hours, more than two and a half shifts. Where drilling starts is the whole game, and it is exactly what lot streaming changes.

The Serial Baseline: Thursday 15:00

With no streaming configured, the drill waits for the last cut piece. It starts Tuesday 10:30 and burns its 20.5 hours across Tuesday afternoon, Wednesday, and most of Thursday, finishing Thursday 15:00. Total span: three and a third days.

Notice what the drill is doing Monday: nothing. A thousand pieces are queuing up at the saw while the longer of your two operations sits idle. That idle Monday is the cost of moving the lot as one block, and it is pure schedule waste, no machine ran faster or slower because of it. This is the classic transfer-batch insight from the Theory of Constraints literature: the batch you process and the batch you move do not have to be the same size. If the concept is new, our primer on what a transfer batch is covers the theory; here we watch it move real dates.

Model 1: The Piece-Count Transfer Batch

Configuration: on the cut step, set the transfer batch size to 100 pieces and a transfer delay of 0.25 hours, the 15-minute forklift move from saw to drill bank.

EDGEBIC derives the drill's earliest start from the physics of the upstream step:

Overlap gate = setup + (transfer batch x run rate), measured from the cut's start

= 0.5 h + 100 x 0.01 h = 1.5 hours after cutting begins

Cutting starts Monday 08:00, so the first 100 pieces exist at 09:30. Add the 15-minute handling move and the drill's earliest start is Monday 09:45, while the saw is still on roughly its 175th piece. The multi-shift allocator then books the drill's 20.5 hours from that gate:

Drilling finishes Wednesday 14:15. The two machines overlap for nearly 25 hours of the schedule, and the order lands a full day earlier than serial. Because the drill is the longer operation, every hour of overlap converts one-for-one into an earlier finish.

Model 2: The Start-to-Start Lag

The second model skips piece counting entirely. You assert a fixed lag in hours: the downstream step may start exactly N hours after the upstream step starts, independent of quantity, rate, and setup.

Set a lag of 1.5 hours on the cut step (deliberately chosen to equal Model 1's derived gate) and the drill's earliest start is Monday 08:00 plus 1.5 hours: Monday 09:30. No transfer delay in this configuration, so drilling runs 09:30 to 16:00 Monday, all day Tuesday, and finishes Wednesday 14:00.

Fifteen minutes apart, the two models look interchangeable. They are not, and the differences decide which one belongs on which step:

The reschedule row is the one that matters most in practice. Suppose drilling is underway and you reschedule mid-job: with a transfer batch, EDGEBIC reads the cut's actual start and its logged piece counts and recomputes the gate from reality. If the saw actually started Tuesday instead of Monday, the drill's gate slides to Tuesday 09:45 on its own, no planner intervention. A hand-picked lag does none of that; assert 1.5 hours and the engine trusts you at 1.5 hours, even if the saw's setup ran long and the first piece is not through. Pick the lag model only where pieces genuinely cannot be counted, and note that EDGEBIC enforces this boundary itself: mark a work center as continuous-process and the piece-count model is switched off for steps routed through it, falling back to the hours lag.

The Three Schedules Side by Side

One day of lead time recovered, and the capacity picture is unchanged: both machines do exactly the hours they always did. Lot streaming spends no capacity; it removes waiting. On a finite capacity schedule that distinction is everything, because the one resource you cannot buy back is the calendar.

Pushing Further: One-Piece Flow

Shrink the transfer batch to 1 and the gate collapses toward the setup time:

0.5 h setup + 1 x 0.01 h = 0.51 h, plus the 0.25 h move = drill starts Monday 08:45

Forty-five minutes after the saw's first stroke, the drill is running. This is the lean one-piece-flow ideal, and it is maximum overlap and minimum work-in-process. It is also a material-handling commitment: someone or something is ferrying pieces continuously between the machines all day. Choose a batch size your handling can actually sustain; a 100-piece batch that moves reliably beats a 1-piece batch that queues on the floor beside the saw.

How Streaming Composes with Queue, Handling, and Transit

Routings carry other between-step times, and EDGEBIC composes them with lot streaming in a defined order. Three rules keep you out of trouble:

The streaming gate replaces the upstream queue wait. Queue time on an upstream step means "downstream waits this long after I finish" in a serial schedule. When lot streaming fires, its earlier gate takes precedence and the upstream queue time becomes irrelevant, a combination EDGEBIC's built-in anomaly check flags so you can confirm the intent. If you want a guaranteed minimum wait, put it in the transfer delay or as queue time on the downstream step.

Transfer delay is for same-day handling. The 0.25-hour forklift move rides on top of the piece-count gate. Use it for cooling, moving, or inspection measured in minutes and hours.

Transit days are applied after streaming, and they are bigger than the win. Give the cut step one working day of transit, the send-out-to-heat-treat pattern, and the handoff lands at end of shift Tuesday; the drill's first feasible slot becomes Wednesday 08:00. The entire overlap gain is absorbed. Streaming buys you hours; a transit day costs you a day. Reserve transit days for genuine overnight sends, and keep same-building moves in the transfer delay.

Reading It on the Gantt

Filter the Job View to MO-77 and the overlap is visible at a glance: the drill bar begins Monday morning while the cut bar is still running, with roughly a day of the two bars stacked in parallel. The schedule grid confirms drill resources booked on Monday, 6.25 hours, the direct evidence that streaming fired. And the engine's decision log records the branch it took for the step, the derived gate, and the transfer delay applied, so the "why does drilling start at 09:45?" question has a written answer.

The arithmetic in this walkthrough is small enough to check by hand, which is precisely why we publish it this way in the EDGEBIC worked examples series. On a real routing with six steps, three shifts, and shared work centers you will not do this by hand, and that is the point of EDGEBIC: the same piece-count gate, the same composition rules, applied consistently across every step of every job, every time you press Schedule All. It is one field on one routing step. Manufacturers have been buying lead time this way since the Theory of Constraints made transfer batches famous, and User Solutions has been putting that math in planners' hands since 1991.

Have a two-step pair of operations where the second machine always waits? Bring that routing to a demo, and we will set the transfer batch together and watch the finish date move.