4.5 Strategic Buffers, Bottlenecks, and Capacity-Constrained Resources
Key Takeaways
- A bottleneck's capacity is less than or equal to the demand placed on it, while a capacity-constrained resource (CCR) can meet demand only if it is scheduled carefully - one needs elevated capacity, the other needs scheduling discipline.
- Strategic buffers come in exactly three currencies: inventory (stock), lead time (time), and capacity (protective or safety capacity), and they substitute for one another.
- Safety stock cannot cure a capacity shortfall and protective capacity at a lightly loaded resource buys nothing, so buffering the wrong place costs money and leaves the exposure open.
- Strategic stock belongs at the customer order decoupling point in the most generic form available; downstream of that point protection must be time or capacity because units are already customer-specific.
- A resource with 160 available hours and 148 hours of demonstrated load runs at 92.5% utilization with a 7.5% cushion; at 158 hours utilization reaches 98.75% and average queue time grows roughly sixfold.
Two resources can both look "tight" in a supply review and still need opposite responses. Labeling them correctly is the whole point of this ECM item, and it is a frequent exam discriminator.
Bottleneck Versus Capacity-Constrained Resource
A bottleneck is a facility, function, department, or resource whose capacity is less than or equal to the demand placed on it. No amount of clever sequencing makes the work fit. The plan stays infeasible until you elevate capacity (overtime, added shift, subcontract, offload, capital) or reduce the demand on it (mix change, demand shaping, re-routing).
A capacity-constrained resource (CCR) has enough capacity in total, but only if it is scheduled and managed carefully. Setups, batch policy, priority churn, or several large orders landing in the same bucket make it deviate from the planned flow even though the monthly load fits inside the monthly capacity. The response is scheduling discipline - sequencing, setup reduction, release control, protecting it with a buffer - not capital.
One-line discriminator: if perfect scheduling still leaves work undone, it is a bottleneck; if perfect scheduling makes the work fit, it is a CCR.
| Stem clue | Diagnosis | Correct response |
|---|---|---|
| Load exceeds capacity every period | Bottleneck | Elevate capacity or reduce the load placed on it |
| Load is 3-5% under capacity but dates slip when orders cluster | CCR | Sequence, cut setups, control release |
| Load is 60% of capacity and dates still slip | Neither - look elsewhere | Fix materials, release rules, or the real constraint |
Two cautions. Utilization alone does not identify a constraint: a resource can run at 100% because you over-released work nobody needed. And constraints are not always machines. Market demand, a single-source supplier, cash, or a policy ("no overtime," "no subcontracting") is often the binding constraint over the S&OP horizon.
Three Strategic Buffers
At S&OP you are not sizing item-level safety stock. You are deciding what kind of protection the plan buys and where it sits. There are exactly three currencies.
| Buffer type | Uncertainty it absorbs | What carrying it costs | When to use it |
|---|---|---|---|
| Inventory buffer (stock) | Quantity and timing variability: forecast error, supplier lateness, yield loss | Working capital, holding cost, space, obsolescence; it also hides problems | Item is standard, stable, and cheap to hold, and customer tolerance time is shorter than cumulative lead time |
| Lead-time buffer (time) | Timing variability in supply and process: transit, queue, setup, supplier reliability | A longer quoted lead time, or earlier release and more work-in-process | Product is configured or expensive to stock, the constraint must never starve, or the promise is a ship date |
| Capacity buffer (protective or safety capacity) | Rate variability: breakdowns, absenteeism, rework, demand spikes | Lower utilization and higher fixed cost per unit | Output cannot be stored (services, perishables), mix is high, or a feeding resource must be able to catch up |
Two ASCM terms to keep straight: protective capacity is extra capacity at non-constraints above the constraint's capacity so they can recover and keep the constraint fed; a capacity cushion is that reserve expressed as a percentage, 100% minus planned utilization.
The three buffers are substitutes - you will pay for protection in stock, in time, or in capacity. Buying the wrong one costs twice, because you carry the expense and stay exposed. Safety stock does not fix a capacity shortfall: you deplete it in the first heavy period and never rebuild it. Protective capacity at a resource already running at 55% buys nothing. A time buffer in front of a non-constraint only adds work-in-process.
Where the Buffer Goes
Placement is judged against two anchors.
Against the constraint. Put a buffer in front of the constraint so it never starves, and where the promise is a ship date, a second buffer after it at shipping so a downstream problem does not waste constraint output. Non-constraints protect themselves with protective capacity, not with stock.
Against the customer order decoupling point (CODP). Strategic stock belongs at the decoupling point in the most generic form available, because that is the last place inventory is still non-specific:
- Make-to-stock: finished-goods buffer plus a capacity cushion for spikes
- Assemble-to-order: module and subassembly buffer at the decoupling point, capacity buffer in final assembly
- Make-to-order: raw and component buffers, with a lead-time buffer in the quote
- Engineer-to-order: capacity and time buffers; finished stock is not an option
Downstream of the CODP the units are already customer-specific, so protection there has to be time or capacity. Where variety explodes at the last operations, push the stock buffer upstream - that is postponement expressed in buffer language.
Sizing a Capacity Cushion: Worked Example
A heat-treat cell has 160 available hours per week after planned maintenance. Demonstrated load is 148 hours.
- Utilization = 148 / 160 = 92.5%
- Protective capacity = 160 - 148 = 12 hours, a 7.5% cushion
Is 7.5% enough? Compare it with week-to-week load variability. If the standard deviation of weekly load is 9 hours, a plus-one-sigma week (157 hours) eats 9 of the 12 spare hours, and a plus-1.5-sigma week (161.5 hours) blows past available capacity and creates backlog somebody has to recover. Aggregate rule of thumb: the cushion should cover the load variability you intend to absorb without overtime.
Now demand rises and load reaches 158 hours:
- Utilization = 158 / 160 = 98.75%; cushion = 2 hours, or 1.25%
- Average queue time grows roughly in proportion to u / (1 - u), where u is utilization
- At 92.5%: 0.925 / 0.075 = 12.3
- At 98.75%: 0.9875 / 0.0125 = 79.0
- 79.0 / 12.3 = about 6.4x longer average queue time
A 6.8% increase in load - ten hours - multiplies waiting time by roughly six. Work-in-process, manufacturing lead time, and late orders all inflate while the capacity report still reads "under capacity." That non-linearity is why aggregate plans deliberately hold protective capacity instead of planning to 100%, and why sustained utilization in the high 90s on a variable resource is a warning sign rather than an achievement.
Three legitimate S&OP responses to the 158-hour case: restore the cushion (overtime, added shift, offload, subcontract), reduce the load (move mix to another resource, cut setups, rebalance routings), or accept and publish a longer planned lead time, which converts a capacity buffer into a time buffer.
Risk Feeds Buffer Sizing
Aggregate buffer sizing is a risk decision, not a formula. Inputs that push buffers up: single-source or single-region supply, long or variable supplier lead times, geopolitical and customs exposure, hazard concentration, volatile family demand, inflexible capacity, and criticality (A items, safety- or regulatory-critical parts).
Match the buffer to the failure mode:
- Distant but reliable supplier (long, stable lead time) - use a time buffer and in-transit planning, not a stock pile
- Nearby but erratic supplier - use an inventory buffer at the decoupling point
- Your own resource is the exposure - use a capacity buffer, because safety stock cannot manufacture hours
- Structural risk - redesign instead of buffering: dual-source, regionalize, or qualify an alternate route
S&OP sets the buffer budget, type, and location; item-level safety-stock statistics come later in inventory planning. Dragging item-level math into the executive meeting is a classic misuse of the forum.
A plating department has 640 available hours per month against 615 hours of demonstrated load, yet it misses due dates in any week when three large orders arrive together, because each changeover costs two hours and priorities get reshuffled mid-week. How should the aggregate plan classify and treat it?
A resource has 160 available hours per week after planned maintenance and a demonstrated load of 148 hours. What are its utilization and its protective capacity?
A make-to-order plant buys a configured component from a supplier six weeks away by ocean freight. The supplier is almost never late and transit varies by only a day or two. Which strategic buffer fits the exposure?
An assemble-to-order plant configures 400 end items from 30 common modules. Where should its strategic inventory buffer sit?