Key takeaways
Short answer: Push scheduling releases work based on a forecast and MRP — make it because the plan says so. Pull releases work only when downstream demand consumes something — make it because it was used. Push tends to overproduce and build WIP; pull keeps inventory low but needs reliable signals and buffers. Most real plants blend the two. See also oee for manufacturing.
Push starts from a forecast. MRP explodes the plan into orders and releases work to the schedule regardless of what the floor has actually consumed. It is well suited to stable, forecastable, long-lead-time demand where planning ahead is both possible and necessary.
Pull starts from consumption. Downstream usage signals upstream replenishment — a kanban returns, a slot opens — and only then is more work released. WIP is capped by design, so the system makes what was taken rather than what was forecast.
A plant on pure push builds to a monthly forecast. The forecast misses, and it ends the month with three weeks of finished goods nobody ordered and a shortage of the variant that actually sold. Switch the final assembly to pull — build only what the next process consumes — and WIP drops sharply while the right mix flows. But the long-lead castings upstream still need push, because you cannot wait for a pull signal on a twelve-week part. The answer was not push or pull but push for procurement and pull for assembly.
Push amplifies forecast error into excess inventory and the bullwhip effect. Pull minimises inventory but stalls if buffers are mis-sized or demand spikes beyond them. Neither is universally right — the choice depends on demand stability and lead time at each stage.
Push for planning and long-lead procurement; pull for shop-floor execution and replenishment. The plan sets the envelope; consumption sets the pace inside it. Most successful systems are this blend rather than a pure doctrine.
1. Calling MRP "pull." MRP is push — it plans from forecast, not consumption.
2. Pull without sized buffers. Stockouts or overproduction result.
3. Forcing one model everywhere. Long-lead procurement rarely suits pull.
4. Never re-sizing kanban. Caps go stale as demand changes.
Pull smooths flow and reduces the starve-and-block that hurts OEE Performance and the bottleneck Availability. Push can keep a constraint fed but at the cost of WIP and the losses it hides — so the scheduling choice quietly shapes the loss tree.
Fabrico makes the starving, blocking and WIP build-up visible, so you can see whether your scheduling is actually smoothing flow. Book a demo to see flow losses on your lines.
No — push suits forecastable, long-lead demand; pull suits steady, repetitive demand.
MRP is push — it plans from a forecast rather than from consumption.
Yes — a hybrid of push planning and pull execution is the most common real-world setup.
It reduces the starving and blocking that cause Performance loss and lost bottleneck time.
Reliable consumption signals and correctly sized buffers; without them it stalls.
Programați o întâlnire individuală cu experții noștri sau înscrieți-vă direct în planul nostru gratuit.
Nu este nevoie de card de credit!
Login
Blog
