Table of Contents
A last-time buy is a commitment to a remaining product life, not a multiplication of last month’s consumption by an arbitrary number of years. Production demand, service demand, yield loss, and material already owned must be expressed in the same component units before the purchase quantity is meaningful.
The following calculation is editorial scenario analysis for OEM and EMS teams. Its numbers are assumptions, not a forecast for a particular device or a report of a customer purchase.
The last order date determines when a supplier will accept an order. The final delivery date determines when material must ship or arrive under the applicable notice and agreement. The service horizon determines how long the buyer needs the component. These dates can be years apart.
Use the actual product discontinuation notice for the complete manufacturer part number. TI’s lifecycle definitions, checked on September 15, 2026, distinguish last-time-buy status from obsolete status. They illustrate why a lifecycle label is a planning trigger, while the specific notice supplies the controlling dates.
The broader EOL sourcing guide covers strategic options. This calculation begins after the team has decided what unchanged design must remain supportable.
A net-demand model in component units
Define:
- P: remaining production units multiplied by component usage per unit.
- S: component units expected for service and repair over the chosen horizon.
- E: engineering, qualification, and destructive-test consumption.
- a: expected loss fraction applied to the component flow being modeled.
- U: verified usable on-hand stock available to this program.
- F: sufficiently firm inbound component units that arrive when needed.
For a simplified model applying one common attrition factor:
Net purchase = max(0, ceil((P + S + E) / (1 − a)) − U − F).
Then round up to the agreed order multiple. If production, service, and qualification have different loss mechanisms, calculate them separately. Do not apply the same loss twice to a demand figure that already includes scrap.
“Usable” excludes quarantined, reserved-for-another-program, unapproved-suffix, or otherwise unavailable material. Physical possession and planning availability are different fields.
A worked example with visible assumptions
Assume 3,600 remaining boards, two devices per board, 500 devices for service, and 100 devices for engineering and qualification. Assume 2% attrition across that combined flow, 1,200 usable units on hand, and 800 firm inbound units.
| Step | Calculation | Component units |
|---|---|---|
| Production requirement | 3,600 × 2 | 7,200 |
| Total before attrition | 7,200 + 500 + 100 | 7,800 |
| Gross requirement | ceil(7,800 / 0.98) | 7,960 |
| Net purchase | 7,960 − 1,200 − 800 | 5,960 |
| Order multiple of 250 | Round up to next multiple | 6,000 |
The order multiple creates 40 additional units above the modeled net requirement. Show that surplus explicitly; it is not evidence that the uncertainty allowance is sufficient.
A second check is time-phased coverage. An inbound 800-unit delivery scheduled after a build cannot prevent that earlier shortage, even though it reduces the lifetime total.
Sensitivity to service demand and redesign timing
Use at least three demand cases, but vary named inputs rather than adding an unexplained percentage.
For example, holding other assumptions constant, reducing service demand to 250 units gives a net need of 5,705 units, rounded to 5,750. Increasing service demand to 1,000 units gives 6,470 units, rounded to 6,500. These are sensitivity cases, not confidence limits.
The team can then compare the incremental 500-unit commitment between the base and high-service cases with the consequences of a repair shortfall. If a redesign is delayed, model the added production separately. If a successful redesign reduces service consumption, model its adoption schedule rather than assuming every installed unit changes immediately.
Financial approval should consider storage, periodic verification, financing, programming, and disposal exposure alongside purchase price. Document who owns the downside if demand falls.
Make the quantity maintainable after purchase
Keep the model linked to a controlled BOM and sourcing package, including board revision, usage quantity, approved ordering code, service assumptions, and stock allocation. Reconcile consumption against each demand bucket.
Storage qualification and repair capability need owners for the entire holding period. A sealed carton is not a maintenance plan, and a component reserve cannot help if programming tools or replacement boards disappear.
When an alternative is technically feasible, the component alternatives process should have dated milestones. Review the buy quantity at each milestone before the order becomes irrevocable, then retain the approved assumptions with the purchase record.
Frequently Asked Questions (FAQ)
Should service demand use the original production volume?
No. Estimate service demand from the remaining installed base, service horizon, repair policy, and component usage per repair. Separate whole-board replacements from component-level repairs to avoid double counting.
Can all open purchase orders reduce the last-time-buy quantity?
Only count receipts that are sufficiently firm, usable for the approved design, and expected before the relevant need dates. Keep uncertain, late, or unapproved receipts outside the base calculation.
Is a last-time buy the same as safety stock?
No. A last-time buy covers a finite remaining lifecycle when replenishment is ending. Safety stock normally buffers variability within a continuing replenishment process.
