Capex Mistakes

Common Capex Portfolio Mistakes and How to Catch Them

The wrong assumptions, unit slips, and process failures that make capex payback and portfolio scores lie, plus how to catch each one before approval.

The most expensive capex mistake is treating the equipment price as the total investment. Symptom: payback lands at 1.8 years on the Project Payback tool but actual cash recovery takes 3.4 years. Root cause: install, rigging, controls integration, and validation are excluded, and those routinely add 25 to 60 percent on top of the machine tag. A 400,000 dollar cell often costs 560,000 dollars installed. Fix: build the denominator from a full landed-and-commissioned number before you touch Capex ROI. If you cannot itemize freight, foundation, electrical, and 30 to 90 days of ramp scrap, your investment base is wrong and every downstream ratio inherits the error.

Double-counting savings across projects is the silent portfolio killer. Symptom: the sum of individual project benefits promises 4.1 million dollars in annual savings but the plant P&L only moves 2.3 million. Root cause: three separate approvals each claim the same 6 operators freed by automation, or two projects both bank the same energy reduction on one line. Fix: net benefits at the portfolio level, not the project level, before you trust Project Portfolio Value or Project Benefit Realization. Tag every saving to a specific cost center and headcount line, then dedupe. If two projects touch the same 12-person crew, only one can book that labor.

Mixing nominal and real cash flows corrupts multi-year payback. Symptom: a 5-year project shows a healthy internal return but the finance team rejects it. Root cause: benefits are stated in today's dollars while the hurdle rate carries a 6 to 8 percent inflation and cost-of-capital assumption, so you are discounting real cash with a nominal rate. Fix: pick one convention. If your discount rate is 12 percent nominal, escalate labor savings at 3 percent per year and material at your actual PPI trend before running Capex ROI. A 300,000 dollar annual benefit held flat understates year-5 value by roughly 13 percent.

Unit and basis errors quietly inflate throughput benefits. Symptom: the model claims 18,000 extra units per year but the line physically cannot exceed 14,500. Root cause: someone multiplied a cycle-time gain by 8,760 calendar hours instead of scheduled hours, ignoring a 68 percent utilization ceiling and planned downtime. A 2-second cycle improvement on a 30-second part is a 6.7 percent gain, not 20 percent. Fix: always convert per-part savings through real available hours. Take scheduled shifts, subtract changeover and PM, apply historical availability, then compute incremental parts. Feed that grounded number into Project Payback, never a theoretical nameplate rate.

Ranking projects by a single metric misallocates the whole budget. Symptom: capex is fully committed by March, then a mandatory safety or obsolescence project appears with no funds left. Root cause: the list was sorted purely on ROI, so a 240 percent return software tweak beats a 40 percent return machine replacement that prevents a line-down risk. Fix: score on multiple axes using Capital Request Score and Project Risk Score together, and reserve 15 to 25 percent of the annual budget for compliance and Equipment Replacement Priority items. Track commitments live with Capital Budget Utilization so you never discover you are overcommitted after the fact.

Ignoring project delay cost makes slow approvals look free. Symptom: a project sits in review for 4 months and no one records a loss. Root cause: the benefit clock is treated as starting at go-live regardless of when it could have started. A project worth 600,000 dollars a year in savings burns roughly 50,000 dollars for every month it slips, so a 4-month stall quietly costs 200,000 dollars. Fix: run Project Delay Cost on every item stuck in the pipeline and put that figure in the approval packet. Delay cost above the analysis cost of a fast decision means you are optimizing the wrong variable.

Stale or single-point inputs make risk invisible. Symptom: the Project Risk Score reads low, then the project overruns by 35 percent. Root cause: estimates used a vendor quote from 14 months ago, a single deterministic benefit figure, and no contingency, so variance never entered the model. Fix: refresh quotes inside a 90-day window, state benefits as a range with a low case at 60 percent of target, and carry 10 to 20 percent contingency on capital and 3 to 6 months on schedule. A project whose best case and worst case differ by more than 2 to 1 is not ready for approval regardless of its headline ROI.

Never reconciling forecast to actual lets the same errors repeat. Symptom: every approved project hits its numbers on paper but the portfolio underdelivers by 30 percent year after year. Root cause: no post-audit loop, so Project Benefit Realization is never measured against the original Capex ROI case. Fix: audit each completed project 6 and 12 months after go-live and compute realized benefit divided by promised benefit. World-class shops land above 85 percent; if yours sits near 60 percent, the fix is not a new formula but feeding actuals back into next year's assumptions. Also verify Project Labor Load matched the plan, since staffing shortfalls are the most common reason benefits arrive late.

Published 2026-07-01.