Strategy Decision Cases

Make or buy: Boeing sold Wichita for efficiency and bought it back for $8.3 billion

Thirty-three companies on record decided what to own and what to outsource. The five reversals were all argued on cost, and undone by something the cost comparison did not contain.

The comparison that gets made, and the one that decides

Make-or-buy arrives as a cost question, because that is the form it can be answered in. Two columns, a unit cost each, a difference. The record contains 33 verified decisions of this kind, and 64% were never reversed — the second-lowest persistence of any decision type here.

The five that were undone were all argued and won on the cost comparison. Each was reversed by something the comparison did not have a row for.

Boeing, Ericsson, and two reshorings

Boeing sold its Wichita fuselage operations in 2005 to become a lean large-scale systems integrator. In July 2024 it agreed to reacquire Spirit AeroSystems at a total enterprise value of about $8.3 billion, undoing the divestiture. The thing that reversed it was not labor cost. It was that the specification had left the building with the factory.

Ericsson single-sourced radio-frequency chips from Philips in 2000 to simplify its supply chain. The Philips plant fire and its consequences ran through a handset business that had no second source. Within eighteen months Ericsson had folded its standalone handset unit into the Sony Ericsson joint venture.

Two reshorings went the other way and ended the same. GE Appliances brought GeoSpring water-heater production to Louisville in 2012 to prove lean assembly could beat cheap overseas labor; production stopped at the end of 2016 on low sales, affecting about 300 hourly workers. Master Lock reshored a sliver of production to Milwaukee in 2011 as Chinese costs rose; the Milwaukee plant closed in March 2024 after more than a century.

The test that separates them

The decisions in this record that stuck are the ones where the company could still specify the thing it did not own. TSMC's customers do not own fabs and can still say precisely what they want built. Zara owns design, production planning, logistics and the IT backbone, and does not own every factory — what it kept is what lets it decide on Tuesday to cut a slow style and double a fast one.

The question is not how much of the chain you own. It is whether a supplier who has never seen your product could build to the specification you hold today. If the answer is no, the capability leaves with the work, and the cost comparison you are looking at is measuring the wrong year.

How to run this decision

Drawn from what separated the records that stuck from the ones that were undone. A step nobody in the record can be seen taking is not here.

  1. 01

    Write the specification for the thing you are about to stop making

    Before the decision, not after. If a supplier who has never seen it could not build to what you have written, the knowledge is in the people and the building, and outsourcing is a transfer of capability, not a transfer of cost.

  2. 02

    Identify the layer, not the percentage

    Make-or-buy gets argued as degree — more in-house, less in-house — and degree does not sort the outcomes here. One of the most successful companies in the record deliberately owns almost nothing; one that owned everything could not leave the layer it owned. Name the specific layer that carries the thing customers are paying for.

  3. 03

    Price the buy-back

    Assume you are wrong and have to bring it home in ten years, at a price the counterparty sets. Boeing's was $8.3 billion. Put that number, discounted, against the annual saving. Most cost comparisons in this record would have inverted under that one line.

  4. 04

    Require a second source before you require a saving

    Ericsson's single-source decision was made to simplify. The simplification was real. The cost of the day it stopped was the handset business. Where a single source is genuinely cheaper, say what the second source would cost and treat the difference as an insurance premium you are choosing not to pay.

  5. 05

    Check whether the layer was chosen because it is visible

    Companies reshore and outsource the parts that appear on the delay report and in the press, which are not reliably the parts that matter. Both reshorings in this record were small, symbolic and short-lived.

  6. 06

    Set the review at the capability, not the margin

    Revisit the decision when your ability to specify the work degrades, not when the unit cost moves. The cost tells you about this year. The specification tells you whether you still have a choice.

What it looks like when this is going wrong

  • The case rests on a unit-cost delta and contains no buy-back price.Boeing, 2005.
  • Nobody can produce a current, complete specification for the work being moved.Boeing, 2005.
  • A single source is justified by simplification.Ericsson, 2000.
  • The layer being moved is the one that shows up in reporting, not the one customers pay for.GE Appliances, 2012. Master Lock, 2011.
  • The decision is stated as a degree of integration rather than a named layer.The seven partial verdicts on record.