The 100% Utilization Trap: Why a Fully Booked CPG Line Is Your Slowest Line

Picture the schedule every plant manager loves to see. The filler is booked to 98% of its available hours. Every shift is spoken for. On paper, nothing is wasted — you are squeezing every drop out of the asset.

Then the floor tells a different story. On-time delivery slips. Work-in-process piles up between operations. The expediting calls start, and suddenly you are running overtime and rush changeovers just to catch up on orders that were supposedly "scheduled."

Here is the counterintuitive truth every scheduler needs to internalize: as a line approaches 100% utilization, wait time does not rise gently — it explodes. A schedule that looks maximally efficient is actually the slowest, most fragile schedule you can build. This post explains the math behind that, why lumpy CPG demand makes it worse, and how to set a deliberate utilization ceiling.

The one equation every scheduler should know: VUT

Queueing theory gives us a compact way to understand this. Kingman's formula — also called the VUT equation — approximates the mean waiting time in a G/G/1 queue as the product of three terms that depend on Variability, Utilization, and Time (service time).

In plain language, the time a job spends waiting in queue is roughly:

Wait ≈ Variability factor × Utilization factor × Process time

Each term does a distinct job:

• U — Utilization. How loaded the resource is. This term takes the form ρ / (1 − ρ), where ρ is utilization.
• V — Variability. The average of the squared coefficients of variation of arrivals and of process time. Bursty orders and inconsistent run/changeover times live here.
• T — Time. The average processing (service) time per job.

The part that bites is the utilization term, ρ / (1 − ρ). Watch what it does as you push the line harder:

Utilization (ρ)	ρ / (1 − ρ) multiplier
50%	1.0
70%	2.3
80%	4.0
90%	9.0
95%	19.0
98%	49.0

That is the hockey stick. Going from 80% to 90% roughly doubles the queue-time multiplier. Going from 90% to 95% doubles it again. From 95% to 98%, it more than doubles once more. The last few points of utilization — the ones that make a schedule look "full" — are exactly the ones that generate the longest, most volatile lead times.

And this is not a rough guess in the danger zone. Kingman's approximation is known to be generally very accurate, especially for a system operating close to saturation — that is, it is most reliable precisely where planners are tempted to over-load the line.

Variability is the accelerant

Utilization is only one of the three terms. The V term is the one CPG operations tend to underestimate, and it multiplies against everything.

Variability enters from two directions:

• Arrival variability (how work shows up). Lumpy order patterns, promotions, seasonal spikes, and last-minute schedule changes all make demand arrive in bursts rather than a smooth stream. A promo that lands 40% of the month's volume in one week is high arrival variability.
• Process variability (how work flows through). Inconsistent changeover times, unplanned breakdowns, quality holds, and a wide mix of SKUs with different run rates all widen process-time spread.

The key insight the VUT equation hands you is this: you can cut lead time by reducing either utilization or variability. They trade off. A line running erratic changeovers and chasing spiky promo demand will choke at 80% utilization. A line with disciplined, leveled work can safely run hotter.

That is why the variability-fighting tools already in your kit matter here. Leveling the load with heijunka attacks arrival variability directly. Shrinking and standardizing changeovers — see the hidden cost of changeovers — attacks process variability. Reducing forecast bias so the plan matches reality, covered in forecasting that survives the floor, keeps demand from arriving as a series of surprises. Every one of these buys back effective capacity.

Why "keep the line busy" is the wrong goal

The root cause of the utilization trap is a cost-accounting instinct. Traditional plants aim to keep machines and labor near 100% of capacity because idle time reads as waste on a cost report. Quick Response Manufacturing (QRM), developed by Rajan Suri, frames this pursuit as counterproductive: chasing high utilization lengthens lead times and builds backlogs. The mindset shift is from keep it busy to keep it fast.

High utilization is not the same as high throughput. Once the queue explodes, most of what accumulates on the floor is WIP sitting and waiting, not product moving. For CPG this is doubly dangerous:

• WIP and finished goods age. Long queues mean inventory sits, which is a real problem for short-shelf-life SKUs. If you schedule perishable items, pair this with shelf-life-constrained scheduling — a full line that produces slowly can push product past its sell-by window.
• The batch-size trap makes it worse. The classic response to a busy line is to run bigger batches to avoid changeovers. But from the QRM view, large batch sizes create long waiting times, high WIP, and long lead times. You dodge one changeover and pay for it in queue. This is exactly the tension balanced in economic production quantity and every-product-every-interval: the right batch size is a balance, not a maximum.

The expediting spiral flows from the same source. When lead times balloon, orders go late, so you rush them — inserting unplanned changeovers and priority jobs that inject more variability, which lengthens lead time further. That is how a "full" schedule turns into permanent firefighting. If that sounds familiar, stop firefighting stockouts covers the operational habits that break the cycle.

Setting your utilization ceiling

So where should the line run? QRM cites queueing theory that high utilization increases product waiting time and recommends operating critical resources at roughly 80% of capacity to absorb demand and product variability. That 80% is not a law of physics — it is a sensible, quotable starting ceiling for a critical resource. Lines with low variability can run higher; lines with spiky demand and erratic changeovers should run lower.

The mental reframe is important: the 20% you leave open is not waste. It is a capacity buffer — a shock absorber that keeps lead times short and stable when reality deviates from the plan. Project production management describes three buffers you can trade against one another to protect a system from variability: capacity, inventory, and time. You can hold spare capacity (run below the ceiling), hold safety-stock inventory, or hold time (longer quoted lead times). Choosing to hold capacity slack is often the cheapest and most flexible of the three, because it protects every SKU on the line at once rather than tying up cash in stock.

A few rules for making the ceiling real:

• Measure against real capacity, not nameplate. Your utilization ceiling must sit on top of an honest run rate. A line that looks 80% loaded against nameplate speed may be 100% loaded against its actual OEE-adjusted rate. Build the ceiling on the schedulable run rate from OEE-based capacity planning.
• Protect the slack in the plan. Reserve planned open time per shift or per line and defend it during rough-cut capacity planning. If the buffer is not written into the master schedule, it will be consumed by the first rush order.
• Put the slack at the bottleneck first. The constraint governs the whole line's throughput, so its buffer matters most. This is the logic behind drum-buffer-rope scheduling: protect the drum, and lead time for the whole flow stabilizes.

A worked mini-example

Take a filling line running one SKU family. Suppose the average job (a production order plus its changeover) takes 4 hours of processing, and the combined variability term works out to about 1.0 (moderate — some order lumpiness, some changeover spread).

Using VUT, the queue-time multiplier is driven by ρ / (1 − ρ):

• At 85% utilization: the utilization factor is 0.85 / 0.15 ≈ 5.7. Estimated wait ≈ 5.7 × 1.0 × 4 hours ≈ 23 hours in queue.
• At 95% utilization: the utilization factor jumps to 0.95 / 0.05 = 19.0. Estimated wait ≈ 19.0 × 1.0 × 4 hours ≈ 76 hours in queue.

Same line, same crew, same equipment. Pushing utilization from 85% to 95% roughly tripled the average time an order waits before it even starts running. That extra 53 hours is where late orders, expediting, and stockouts come from. And because the curve steepens, a single breakdown at 95% pushes the line into the near-vertical part of the hockey stick, where recovery takes days rather than hours. At 85% the buffer absorbs the same shock.

Notice what did not change: the line's maximum output per good hour. You did not gain throughput by loading to 95% — you gained queue. That is the whole trap in two numbers.

Putting it on the board — the scheduler's checklist

• Measure real utilization against available, OEE-adjusted capacity, not nameplate speed.
• Pick a utilization ceiling per critical line or resource — start near 80% and adjust for that resource's variability.
• Write the slack into the plan and defend it; treat the buffer as protection, not idle waste.
• Attack variability to earn back capacity — level-load with heijunka, cut changeover spread with SMED, and reduce forecast bias so demand stops arriving in surprises.
• Buffer the bottleneck first, since the constraint sets the pace for the whole flow.
• Re-check the ceiling as the demand mix and SKU portfolio shift — more variability means a lower safe ceiling.

Key takeaways

• 100% utilization means maximum lead time and maximum fragility — not maximum output. The ρ/(1−ρ) curve makes the last few points of load the most expensive.
• Leave planned slack. Roughly 80% is a sensible starting ceiling for critical resources; the open capacity is a buffer, not waste.
• Cut variability to raise the ceiling safely. Leveling, faster changeovers, and better forecasts all buy back usable capacity.
• Kingman's approximation is most accurate near saturation — right where the temptation to over-load is strongest, so trust the math.

A fully booked line feels like a win on the schedule board. The queue tells the truth. Leave room to move, and the whole line moves faster.

Sources

• Kingman's formula (VUT equation) — https://en.wikipedia.org/wiki/Kingman's_formula
• Quick Response Manufacturing (utilization, lead time, 80% ceiling, batch size) — https://en.wikipedia.org/wiki/Quick_response_manufacturing
• Project production management (VUT and buffers of capacity, inventory, and time) — https://en.wikipedia.org/wiki/Project_production_management
• iSixSigma — buffer and cycle-time context — https://www.isixsigma.com/dictionary/buffer/
• Netstock — lead time and variability — https://www.netstock.com/blog/what-is-lead-time-and-how-does-it-work/