Operating perspectives on inventory, supply chain, and AI — written by practitioners who are still doing the work. No vendor-sponsored content. No "thought leadership" as lead-generation theater. Just what we've actually seen.
Every CFO conversation about supplier relocation starts with tariffs. That's the visible number. It's also the smallest line item in the actual decision.
We've now run or supported supplier diversification programs at five mid-cap consumer-products brands. In every single case, the leadership team meaningfully underestimated the cost and duration of qualification. Here's the operational math.
A new injection-molding partner in Vietnam doesn't ship Friday. They submit samples. You test samples. Samples fail, samples succeed, samples need tooling adjustments, tooling adjustments require new samples. Then first-article inspection. Then small-lot validation. Then capacity ramp. Across 8 SKU families that's not 60 days — it's a year of overlapping qualification streams that need to be project-managed.
The press-release version of relocation is "unit cost is 12% higher but tariffs save 25%, so we're ahead 13 points." The actual landed cost has freight (different lanes, different carriers, different transit times that force more safety stock), duty drawback unwinds, FX exposure shifts, and a one-time investment in tooling that nobody put in the model.
In every case we've seen, the all-in cost of relocation in year one is between 4% and 9% worse than the original tariff regime. Year-two through year-five it gets better — but only if you executed year one well.
If your goods used to land in Long Beach and now they land in Houston, your DC network is wrong. We've seen brands quietly pay 8–12% in incremental freight for 18 months because nobody re-ran the network model post-relocation. The network needs to be redrawn at the same time as the supplier shift, not afterward.
This is the conversation nobody documents. Your incumbent Chinese supplier — who has been a partner for 7–15 years — is reading the same news you are. They will respond with some combination of: aggressive concession offers, quality slippage on the ramp-down volume, and reluctance to release tooling. The negotiation arc is roughly six months and requires senior relationship management. It is not a procurement task.
Three principles, repeated across every successful diversification we've supported:
Sequence by margin contribution, not by tariff exposure. Move your highest-margin SKUs first. They absorb relocation cost most easily and give you political cover for the harder ones.
Dual-source from day one. Single-source moves trade one risk for another. Even a 70/30 split into two new geographies dramatically de-risks the next disruption.
Redesign the network in parallel. Don't treat sourcing and distribution as sequential workstreams. They share data, they share constraints, and they share trade-offs.
If you're in the middle of this work right now, the Supply Chain Resilience practice is built around it. The diagnostic gives you a sized, sequenced relocation plan in four weeks — including the freight, tooling, and qualification math most teams skip.
Spoiler: it's not the model. SAP IBP's underlying statistical engines are perfectly capable. The problem is upstream — in how promotional events are categorized, encoded, and fed back into the forecast.
A CPG operator runs a national feature-and-display promotion. Sales spike. Six weeks later, the planning team sits in S&OP and watches the forecast call for similar volume next quarter — except there's no promotion next quarter. The forecast carries the lift forward. Inventory ships. Inventory sits. DSI climbs. Finance gets unhappy.
Or the inverse: a promotion is run, sales lift, and the planning system doesn't recognize the event — it just sees an unexplained spike, flags it as noise, and dampens future forecasts to compensate. You end up short on a re-run six months later.
The root cause in both cases is the same: the system can't distinguish baseline from lift because the event metadata isn't clean.
For every promotional event, the planning system needs to know:
What it was (TPR, feature, display, multi-buy, coupon, end-cap), where it ran (retailer, region, store cluster), when it ran (start, end, ad date, in-store date — these are not the same), and what it cost (trade spend dollars, slotting fees).
Most CPG shops have this data. It lives in trade-promotion management systems, in retailer portals, in account-team spreadsheets, and in marketing's heads. The work isn't generating the data — it's getting it into the planning system in a form the forecast can use.
Once event metadata is clean, you can layer machine learning on top of the statistical baseline to estimate lift coefficients per event type, per retailer, per category. You're not replacing IBP. You're feeding it better inputs and pulling out better outputs.
We've seen forecast accuracy on promoted SKUs improve by 25–40 points using this approach. Not by switching planning platforms. By cleaning the data and adding a thin ML layer.
Most "AI forecasting transformation" projects fail because they're sequenced backwards. They buy a new ML platform first, then try to feed it the same dirty data, and produce the same bad forecasts — now with more steps.
The order that works:
1. Fix the event taxonomy. 2. Get clean event data flowing into the planning system. 3. Layer ML on the highest-lift SKU clusters. 4. Roll out by category. 5. Reassess platform investment after you know what you actually need.
This is exactly the work the Inventory Intelligence practice does — and it's why we don't push platform migrations as the answer. The platform you have is usually fine. The data into it usually isn't.
When a consumer-products company separates a business unit — through spin-off, divestiture, or strategic sale — the press release talks about strategic focus, the proxy filing talks about financial structure, and the deal lawyers talk about transition services. Almost nobody talks about inventory. That's where the worst problems hide.
Most diversified CPG companies have inventory items that are used across business units — a packaging component, a raw material, a finished good sold under multiple brands. When the separation happens, who owns the residual inventory? Who pays for write-downs? Who is on the hook for the supplier contract that the spinco no longer wants but the parent doesn't need?
These questions get punted in the merger agreement to "good faith negotiation post-close" — which is consulting code for "operations teams will fight about this for 14 months."
Pre-separation, demand planning was done at the enterprise level. Post-separation, two planning teams need to forecast independently — but they share a supplier base, share manufacturing capacity (under TSA), and share retailer relationships. Forecast bias on either side spills onto the other.
We've seen carve-outs where the parent and spinco independently over-forecasted the same shared retailer — and both built inventory against the same expected demand. The miss compounded.
On day one, both entities have the same item master, supplier master, and chart of accounts. On day 90, they don't. New SKUs get added in different ways, supplier codes get reassigned, taxonomies drift. By the time someone notices, six months of decisions have been made on incompatible data — and reconciliation is brutal.
A good Transition Services Agreement covers IT, HR, facilities, and finance. A great TSA also covers the operational mechanics of separation:
Inventory rebalancing rights and obligations. Who can pull shared inventory, under what conditions, at what transfer price. Forecast sharing protocols. Who sees what demand signals during the separation period and what governance is in place to prevent independent decisions on shared capacity. Master data governance. Who owns the shared item master during transition, who approves changes, and how the eventual split happens cleanly. Supplier contract bifurcation. Which contracts go where, which get reassigned with consent, and who eats the breakage fees.
The companies that get carve-outs right have a defined operational separation playbook that runs in parallel with the legal/financial separation. The playbook covers the four areas above and is owned by a small joint team that reports to the COOs of both entities — not to the deal team.
The companies that get it wrong assume operations will figure it out. Operations does figure it out, eventually, at a cost of 8–15% of one year's combined working capital.
If a separation is on your roadmap — whether you're the parent, the spinco, or evaluating the option — the Carve-Out & Stand-Up offering inside the Supply Chain Resilience practice is built for exactly this. We've sat on both sides.
The four-week assessment is exactly that — a structured, fixed-fee diagnostic on the highest-leverage moves in your operation. Output is one document your CFO can read in 20 minutes.
Request Assessment