An operations case asks why the machine is underperforming and how to fix it. A turnaround case adds urgency and cash stress, because the business is losing money and may run out of time. The signature failure is to treat every margin problem as a cost problem and start cutting, when the lever that matters is often somewhere else entirely.
How do you tell a cost problem from a throughput problem?
The first question that sorts these cases is cost problem or throughput problem? A cost problem means each unit costs too much to make or serve, through input prices, labour rates, overhead or waste. Its lever is cost reduction and pass-through. A throughput problem means the system can't produce enough good units fast enough. Think of a bottleneck, a quality/rework loop, low utilisation, a demand-supply mismatch. There the lever is elevating the constraint. Goldratt's Theory of Constraints, from The Goal, says a system's output is governed by its single tightest bottleneck. Improve anything other than the bottleneck and you raise cost and inventory without raising output. Mistake a throughput problem for a cost problem and a plant cuts maintenance to save money. That worsens the bottleneck that caused the trouble in the first place, and Worked Example 4 takes that trap apart.
Advertisement
How do you find the root cause of falling margin?
The workhorse tool is a margin bridge that reconciles. Start from the prior-period profit and break the whole change into named, additive buckets: price/mix, volume/throughput, input costs, labour, overhead deleverage, one-offs. Then make them add up to the current profit with nothing left over. A bridge that reconciles turns "margins are down" into "here are the four things that moved it, in order of size", and you start with the biggest bucket. Next comes root-cause analysis, the five whys or a fishbone, which takes you past the proximate cause to the one underneath. Margin is down because throughput is down. Throughput is down because one line is the bottleneck. The bottleneck machine keeps failing, and it keeps failing because maintenance was frozen two years ago to preserve cash. The "cost problem" was a capital-allocation problem wearing a cost-problem mask.
What order should a turnaround follow?
Finally, a turnaround has a sequence. Get the order wrong and you waste time and cash, the two things you're shortest of. (1) Stabilise: stop the bleeding, secure liquidity, triage. (2) Diagnose: bridge and root-cause before acting broadly. (3) Fix the constraint: attack the biggest, fastest, cheapest lever first (often the bottleneck) and bank quick wins that fund the rest. (4) Restructure the deeper cost and capital base. (5) Grow, but only once the machine is stable and profitable. Growth before stabilisation is how a company runs out of cash while chasing revenue. Worked Example 4 builds the bridge, traces the cause, and sequences the fix.
{
"skill": "Scaffolding a pricing case's value-based ceiling — the economic-value-to-customer (EVC) build",
"use": "Have an AI assistant produce a first-cut list of the value drivers a product creates for its customer (time saved, defects avoided, revenue enabled, risk reduced) and a starting competitive-price range, so you begin the EVC build from a checklist instead of a blank page — then you quantify and verify each driver yourself.",
"tools": ["Chat assistants — Claude, ChatGPT, Gemini (for the driver checklist and structure)", "Research/browsing assistant (for a first-pass competitive-price range, source to be verified)", "Spreadsheet or code tool (to build and re-check the EVC and contribution math)"],
"prompt": "For <PRODUCT> sold to <CUSTOMER TYPE>, list the distinct ways it creates or destroys economic value versus <NAMED ALTERNATIVE>, each expressed as something I could put a rupee/dollar figure on (cost avoided, revenue enabled, time saved, risk reduced). Then name the primary public sources where I could find the reference (competitor) price. Do not invent specific prices or savings figures — give me the driver structure and where to verify each.",
"verify": "Every quantified value driver and every competitor price must be traced to a primary source before it enters the recommendation — the customer's own cost data, a published price list, a filing, a regulator's tariff — never the model's guess (Guardrail 1-5: Trace, Match unit/scale/period/scope, Recompute the EVC and contribution arithmetic yourself, Date-check the price, Cite). A hallucinated saving figure does not just embarrass you; it sets a real price.",
"diy": "You must be able to build the EVC ladder and the contribution/elasticity math by hand — that is exactly what Format 1 and Worked Example 1 install. AI proposes the driver list; you dispose by quantifying, checking the floor, and pricing to contribution.",
"market": "IN"
}