The CFO's Missing Thirty: Why Manufacturing Savings Plans Realize 70% of the Deck

Opening Insight

The labor plan went to committee on a Tuesday. Eight heads across two crews, sized against the current SKU mix, mid six figures in annual savings, approved clean.

The packaging sourcing negotiation wrapped on Thursday. New film supplier, tighter threading, better unit cost. Procurement hit their number.

The automation capex presented the following week. Case packer on Line 2, ROI case built on 14 productive hours per day, 22-month payback. Committee approved.

The scheduling initiative was already in flight, optimizing changeover sequences against the original labor pattern that existed six weeks earlier.

Four decisions. Four owners. Four clean business cases. Four approvals. The CFO added up the savings: 2.1 million in year one.

Twelve months later, the P&L showed 1.4 million. The question was not where execution failed. The question was why four defensible projects, each individually sound, compound into a gap that shows up only in the aggregate.

Most mid-market manufacturers pull cost-savings levers one at a time. The real savings are in pulling four together. The cost is not execution. It is the missing model that shows how four capital decisions interact once they hit the floor.

System Context

In our work with mid-market manufacturers, especially food and CPG, the pattern is close to identical.

The plant has four decisions running concurrently at any given moment. Labor, how many people, where, at what utilization. Automation, where equipment replaces or augments labor. Scheduling, how work falls across shifts, lines, and time. Packaging sourcing, what flows through, in what formats, from which suppliers.

Each decision has a natural owner. Operations owns the scheduling initiative. Finance owns the automation capex case. Procurement owns the sourcing negotiation. Human resources owns the labor plan.

Each owner builds a business case that treats the other three decisions as stable. The labor cut is sized against the current product mix. The automation case assumes a full schedule at current changeover cadence. The sourcing change is optimized against this quarter's volume profile. The scheduling optimization runs against this year's labor availability.

None of those assumptions survive the floor once multiple approvals land at the same time.

Mechanism

When four capital decisions are approved in isolation, each is optimized against a static picture of the other three. The floor is not static. The decisions interact.

Take that labor cut sized against last year's SKU mix. It assumed the line runs a specific blend of products, each requiring a specific changeover sequence, each demanding a specific labor footprint per shift. The cut was defensible at that product mix.

Now the packaging sourcing change lands. The new vendor's film runs tighter on the threading. Changeovers lose 8 to 12 seconds each. The plant runs 40 to 60 format changes a week. That is 320 to 720 seconds per week of added micro-stop exposure. The labor cut was never sized for that added variability.

Now the automation capex runs live. The ROI case assumed 14 hours of productive time per day. That number was built on the scheduling pattern that existed before the labor cut. With fewer operators covering the same shifts, the schedule adapts to labor availability. The automation runs 10 to 11 hours instead of 14. Payback stretches from 18 months to 36.

Now the scheduling initiative tries to optimize against the new reality. Its original optimization was modeled against the original labor plan. With the cut in place and the automation running short of capacity, the scheduling fix has a different feasible frontier than the one modeled.

Each project is doing its job. The interactions between them are where the savings leak.

This is not execution failure. It is approved math running on unmodeled reality.

System Interaction

The primary mechanism is compounded by three second-order effects that only show up when the four decisions land on the same floor.

First, variance amplification. When a labor plan is tuned for a specific schedule and the schedule shifts, the system does not settle at the new steady state. It oscillates. Quality holds rise because inspectors cover cross-trained gaps. Quality holds pull capacity from cold storage and staging. Cold storage fills up. Product waits for disposition. Disposition latency chokes upstream flow.

Second, rework-loop amplification. A sourcing change that introduces a new material viscosity or a new film tolerance creates rework events that consume first-pass capacity. The plant does not lose capacity when it stops. It loses capacity when it runs the same product twice. Rework loops become invisible because they consume the same lines as first-pass production.

Rework consumes the same line capacity as first-pass production. A 4 percent rework rate is a 4 percent capacity tax that never shows up on the downtime log.

Third, decision-distortion cascade. Hidden loss gets misattributed. Overtime rises and looks like a labor shortage. Capital is approved to solve what looks like a capacity problem. The system adds steel while the underlying instability remains. Next year's capital request is sized against a symptom that the integration model would have named differently.

Economic Consequence

Approved on clean math. Running on messy reality.

In our Outpost engagements, the gap between approved savings and realized savings lands between 20 and 35 percent in the first year. The range varies by plant. The pattern does not. The more interlocked the four decisions are, the wider the gap.

Translate to dollars. A two million annual savings plan lands at 1.3 to 1.6 million. Over a three-year horizon, that 400 to 700 thousand per year compounds into 1.2 to 2.1 million of margin that never materialized. For a mid-market manufacturer running single-digit operating margins, that compounding gap is meaningful against enterprise value.

The second-order effect is worse. Next year's capital request is sized against what the CFO sees, not against what the plant could have realized. The plant has fewer easy levers remaining. The largest savings from labor, automation, scheduling, and packaging sourcing have all been pulled. What remains are the harder levers. The CFO asks for another two million. The plant struggles to find one million.

This is the Silo Tax. It accumulates year over year as an erosion on capital efficiency. It never shows up as a single line item on the P&L. It shows up as a persistent delta between deck math and actual performance.

Diagnostic

The diagnostic is simple. Ask what percent of last year's planned cost savings actually hit the P&L.

If the answer is 100 percent or more, the plant runs an exceptionally well-integrated capital planning process and this article does not apply.

If the answer is 80 to 99 percent, the plant has modest Silo Tax. Manageable.

If the answer is below 80 percent, or nobody tracked it formally, the Silo Tax is present and likely material.

The integration savings are bounded by how many of the Four Decisions were approved in the same 12-month window. Two or fewer, modest gap. Three or more, significant gap.

Decision Output

• Decision type: Capital plan integration
• Trigger: CFO annual savings target above 5 percent of COGS, with three or more concurrent capital projects touching Labor, Automation, Scheduling, or Packaging sourcing
• Action: Build an integrated system model of the four decisions before committee approval. Score each project's stand-alone math against the integrated math. Approve only to the integrated number.
• Tradeoff: Integrated modeling adds two to four weeks to the approval cycle. The cycle time cost is usually less than the Silo Tax it eliminates.
• Evidence: Compare last year's approved savings to realized. If the gap exceeds 15 percent, the modeling cost is recovered in the first integrated cycle.

Framework Connection

This is where Nobody Owns the Seams plays out in capital planning.

Operations owns utilization. Finance owns payback. Procurement owns unit cost. Human resources owns the labor plan.

Nobody owns the interactions.

The capex committee approves four projects. What the committee did not approve, and cannot have approved, is the system those four projects will operate inside together. There is no accountable owner for the seams between Labor, Automation, Scheduling, and Packaging sourcing once the approvals land. Each project's owner delivers their piece. No one's piece is the integration.

The Silo Tax is not a failure of any single project. It is the systematic cost of a committee process that approves components but does not approve the system.

The fix is upstream of the committee. It is a model, not a better presentation.

When the four decisions are modeled together, the committee sees the real number rather than the sum of four optimistic stand-alones. The approved savings and the realized savings converge. The Silo Tax compresses.

This is the shape of A Factory That Thinks. Not a plant that has perfectly optimized every unit operation. A plant where the four decisions that matter most to cost structure were modeled as one system before any of the capital committed.

Strategic Perspective

A mid-market manufacturer making four capital decisions per year in isolation is paying a recurring tax that does not appear on any invoice.

The tax is proportional to the ambition of the cost-savings plan. A plant pulling 500 thousand a year in isolated levers pays the tax in thousands. A plant pulling three million a year pays it in hundreds of thousands, compounding.

For a mid-market CFO, the leverage is at the capital committee, not on the floor. The integration model is cheap. The leak is not.

Factories that think model the Four Decisions together before the checks are cut. The approved number and the realized number converge. The compounding gap closes. The next capital request starts from honest baseline math.