Framework makes modular laptops whose memory and storage can be replaced instead of soldered down. In July 2026, founder and chief executive Nirav Patel faced the ugly side of that promise: a supplier quoted more than twice the previous cost for the new LPCAMM2 memory used in one Laptop 13 Pro configuration. Framework changed some preorder configurations, raised new-order prices, then cut prices and issued refunds when cheaper supply appeared in September.
The useful lesson is not that a transparent blog post makes a shortage painless. It is that a small hardware company needs a written rule for which customer, batch and margin absorbs a component shock. Framework exposed enough of its rule to build one.
The forecast missed by an entire multiple
Framework expected a low-to-mid double-digit percentage increase from the second to the third quarter. The actual supplier quote was more than double the cost of prior inventory, according to Patel's update and Tom's Hardware's July 22 report. The problem was not only price. Framework had enough modules in aggregate for existing orders, but not the right mix of 16GB, 32GB and 64GB capacities.
That distinction matters. A spreadsheet that says "500 modules available" can still hide a customer-level shortage when 300 buyers ordered the scarce capacity. The unit of planning is not a component family. It is each sellable configuration.
Framework first protected the company. It said absorbing the quote would put its ability to operate at real financial risk. It fulfilled as many high-capacity orders as available supply allowed, then moved affected orders to lower capacities at the corresponding original price. New orders received higher prices. Customers rightly objected that a preorder is a promise about a configuration, not just a place in line. A July community thread complained that buyers learned about changes indirectly. That is a separate failure from the supplier shock, and an avoidable one.
The September reversal was a policy test
By September, Framework had found a limited quantity of cheaper Micron modules. Its running pricing update says it automatically refunded shipped buyers who had paid the higher price and reduced pending preorders through Batch 10. The September 29 update extended the decrease to remaining Laptop 13 Pro preorders and new orders.
That move turns "we pass through costs" into a two-way promise. Many sellers raise prices when inputs rise and quietly keep the difference when they fall. Framework used weighted average cost, blending expensive and cheaper inventory, and applied the result to identifiable batches. Tom's Hardware confirmed that reductions covered shipped orders and a subset of pending orders.
Weighted average cost is simple in principle. If 100 modules cost $180 each and a new shipment of 50 costs $120, the illustrative inventory cost is:
(100 x $180 + 50 x $120) / 150 = $160 per module.
That is not Framework's unpublished cost. It shows why a new cheaper shipment does not instantly return the whole catalog to the old price. The business still owns expensive units. Using the blended figure avoids giving the newest buyer all the benefit while earlier buyers fund the difference.
A four-box preorder rule
A founder can write the decision before the next quote arrives. Put every response in one of four boxes:
- Existing paid order, cost rises: honor price and specification until doing so crosses a predeclared survival threshold.
- Existing paid order, supply disappears: offer a wait, an equivalent substitute, or a full refund. Do not silently downgrade.
- New order, cost rises: reprice immediately and timestamp the applicable batch.
- Cost falls before or shortly after fulfillment: define whether the difference becomes a refund, credit or margin recovery.
The hard part is the survival threshold. Illustrative arithmetic makes it concrete. Suppose 400 preorders each contain a part whose cost jumps by $140. Honoring all orders costs $56,000. If the company's unrestricted cash after payroll and tax reserves is $240,000, the shock consumes 23 percent of that cushion. The answer may still be to absorb it, but the decision is now attached to runway, not pride.
What transfers, and what does not
Framework's modular product made substitutions unusually visible. A food brand can change a carton supplier without altering the product a customer selected. A laptop buyer chooses memory capacity explicitly. That limits how directly the case transfers.
The stronger transfer is the ledger. Track old inventory, incoming purchase orders, customer batch, promised configuration and the exact communication sent. Update buyers before a forum discovers the change. The same discipline belongs in any supplier map, especially when a single input can hold the whole product.
Within an hour, add five columns to the preorder sheet: component revision, quantity committed, replacement approved, maximum absorbable variance and customer-notification owner. Then run a 2x quote through it. If the sheet cannot tell you who gets what, the business does not yet have a shortage policy.
Framework's repairable design is its public idea. The quieter operating idea is more useful during a shortage: price transparency needs a formula, a batch boundary and a refund rule. Otherwise it is only narration after the damage.
Separate the promise from the estimate
A preorder page contains at least four different statements: the selected specification, the price, the estimated ship window and the order in which batches will be served. Label them internally as promises or estimates. A component capacity and paid price usually belong in the promise column. An early-Q3 shipping window may be an estimate. If the team has not agreed on the distinction, customer support will invent it during the crisis.
Next, add a communication clock. A material supplier change should trigger an internal decision deadline, a direct buyer message and a public update in that order. Forum discussion can still be useful for questions, but it should not become the first notification system. Write who owns each message and prepare the downgrade, wait and refund options in advance.
Finally, track the cost of remedy. Automatic refunds require transaction work and cash; configuration swaps require inventory and support; cancellations change the batch forecast. Include those costs when setting the survival threshold. A policy that cannot be funded is only a statement of intent.
The result is a small but durable operating asset: a preorder constitution. It tells the team what stays fixed, what can move and who must hear first. Supplier quotes will remain volatile. The business does not have to be.
Review it after every launch. Record the largest forecast variance, days to direct notice, cancellations and remedy cost. Automatic reductions can repair economic harm, but they cannot erase surprise. Both results belong in the review. The strongest trust claim is not that prices are fair. It is a process customers can observe when fairness becomes expensive.
The same discipline applies when the gate is testing rather than price, as the GameTank production delay shows.
Discussion
Sign in with Google or just a name. No email link, no password to remember.