OK, after creating test inventory items and running some tests, I can confirm that what suggests definitely doesn't work.
If I create new items for the colors and then do an inventory adjustment, the value of what I subtract from the original PRODUCT SKU goes to Inventory Shrinkage, but the value of the color-specific SKUs is entered as zero. Presumably that's based on an average cost of zero.
If PRODUCT-RED, PRODUCT-WHITE, and PRODUCT-BLUE have an average cost already (i.e. I've already made some purchases), then it assigns each color's cost average to the quantities adjusted into inventory. Better, but it probably won't balance.
This is where it becomes clear that the tracking of inventory and its value in QuickBooks Online was an afterthought.
Chat GPT kept suggesting journal entries to move the total inventory value for each item and then quantity adjustments to update the inventory. A journal entry can only move values between accounts, though, and all four SKUs in this scenario share the same Inventory Asset account. In other words, I can move the valuation of PRODUCT out of the Inventory Asset account, but then I'm putting it right back in the same Inventory Asset account. There is no way to credit it to a particular product. We went over that five times before I gave up, relatively comfortable that AI won't be taking over the world this week.
So the issue becomes moving both quantities and costs, and there's only one way I can see to do that:
A Vendor Credit for the total quantity and cost of PRODUCT. I'm removing all of the existing inventory and its total value.
Then, on the same Vendor Credit, negative quantity "returns" to each of the color-specific SKUs. (Remember: Since it's a Vendor Credit, positive values are inventory out while negative values are inventory in, but the rate is always positive.) The goal here is for the total of the Vendor Credit to be zero: the same total amount out of the Inventory Asset account, and then back in—but attached to new products.
If you're FIFO, that complicates things, but for cost averaging, this is close enough for me.
Until someone can tell me why this doesn't work, this is the solution.
