The 30% Mandate at Northwind Logistics

The plan treats a cost number as the objective and makes three decisions that each discard something the Framework protects:

• 80% of compute locked into 3-year all-upfront RIs prices commitment against a demand curve nobody has measured. If any meaningful share of that compute is dev/test, or sits behind an architecture that will change, the commitment converts a variable cost into a stranded asset and removes the option value the variable cost model exists to provide.
• Friday-night termination is a decision made about someone else's workload by a team that does not own it. The two batch teams are the diagnostic: what looks like idle in a utilization chart is production demand that happens to be scheduled at night. Business value, not a utilization threshold, decides whether a resource should exist.
• Consolidating to one account optimizes the invoice and destroys the allocation boundary Finance would need for any recharge that means anything.

Reframe the mandate as a savings pipeline with owners, not a single number: split the target into waste (idle, orphaned, over-provisioned, unowned), rate (commitments and discounts on proven steady state), and architecture (re-platforming, storage tiering). Tell the CFO which portion is achievable in 90 days with confidence and which requires owner decisions that will take longer. A 30% cut achieved by removing customer-facing capacity is a cost reduction that shows up as an availability incident two quarters later.

Days 1-15 - establish ownership and a baseline. Get a named owner against every account, resource group and tag key you can. Publish a first showback to the 14 product teams, imperfect, with the gaps labelled as gaps. Start an idle/orphan inventory (unattached volumes, old snapshots, oversized instances, non-prod running 24x7) and route each item to its owner for a keep/kill/resize decision.

Days 16-50 - harvest the uncontested savings. The owner-confirmed idle and non-prod items go first; the batch teams confirm their windows and get scheduling guardrails rather than shutdowns. In parallel, measure the actual steady-state floor per service and buy commitments against that floor - Compute Savings Plans or 1-year partial-upfront before any 3-year all-upfront, so the commitment tracks the measured curve rather than a guess. Rightsize the top 20 resources by cost.

Days 51-90 - re-baseline and prove sustainability. Restate the run-rate against the original $1.8M, publish a savings register with owner, action and realised amount, and pilot one unit-economics metric with a product team.

Refuse: blanket night-time shutdowns of workloads whose owners have not been consulted; 3-year all-upfront coverage at 80% of compute; account consolidation before an allocation model exists. Each is a one-way door taken on data you do not have.

Cost per unit of business value - for Northwind, cost per quote generated or per shipment processed - shown alongside the absolute run-rate, not instead of it.

Why this one: an absolute reduction answers 'did we spend less?' but not 'did we spend less because we are smaller, or because we are better?'. Volume at a logistics company moves with seasonality and contract wins; a 30% absolute cut can coexist with unit costs rising. A unit metric is the only one that survives a change in volume and therefore the only one that tells the CFO whether the savings are structural.

Present it as three lines on one chart: total spend, unit cost, and volume. If total spend is flat while unit cost falls, the business grew and the practice worked. If both fall together, you may simply have shrunk.

Add a second, supporting number for the practitioner's own accountability: percentage of spend allocated to a named owner. At the start it is roughly the 41% tag coverage; the target is 90%+. That number predicts whether the next 30% is achievable at all, because unallocated spend has no one to optimise it.