Backflush Tooling Cost Variance in a High-Mix Shop
Caldera Fabrication runs a high-mix job shop with 38 CNC machines. Insert pockets are backflushed at operation completion: six inserts per cutter for every operation, regardless of whether the operator changes all six. The crib issues sealed packs of six to the job but accepts partial packs back without a transaction. Last quarter, tooling expense was 18% over standard, and cycle counts found 1,900 inserts of variance across four grades. Two cutters with the same tool ID are used on different jobs: one runs 40 pieces per edge, the other 110 pieces per edge. The controller wants to stop backflushing entirely and expense all tooling at purchase, which would simplify transactions but reduce job visibility. Operations wants to keep backflush because manual issue takes too long.
1. Assess the controller's proposal to expense all tooling at purchase. What would be gained and lost, and what is a better control design?
Read the worked answer
Expensing everything at purchase would simplify the ledger but destroy job-level visibility, make variance invisible, and weaken traceability for customers who ask which tool grade ran their parts. It also removes the link between tool life and standard cost.
A better design is segmented: keep backflush for low-value, low-risk consumables where the transaction cost exceeds the control benefit; use actual issue and return for high-value inserts, customer-owned tooling, and grades with large variance. For backflushed items, set the quantity from the expected replacements per tool life, not a fixed six per operation. Add a return or scrap transaction for partial packs. The controller's goal should be controlled consumption, not zero transaction detail.
2. What changes to tool IDs, backflush quantity, and crib returns would correct the 18% variance? How should the existing 1,900-insert variance be handled?
Read the worked answer
Three changes address the 18%: first, split the tool ID where the same ID serves a 40-piece-per-edge and a 110-piece-per-edge application; one standard cannot represent both. Second, calculate backflush quantity from pocket count and expected edge life, and allow a variable replacement quantity when the operation completes. Third, require the crib to transact returns of partial packs or declare them scrap, so issued quantity does not become consumed quantity. Then compare actual issue, backflush, and cycle count by grade and work center.
The 1,900-insert variance should be recounted by grade, investigated for a specific cause such as a wrong backflush quantity or untransacted returns, and adjusted through an approved count adjustment, not buried in a purchase expense account. Pilot one cell before global change.