Re-cutting a batch-automation business case to survive an 18-month hurdle
A specialty-chemical plant runs three 12,000 L batch reactors producing 14 product families on a five-day campaign cycle. The automation lead proposes replacing recipe handling with Glacier Fabricant plus closed-loop batch control, citing a six-week pilot on Reactor 2 that cut average cycle time from 9.4 h to 7.3 h. Capital is USD 1.85 M with USD 210 k/yr recurring licence and support. Projected benefits are USD 620 k/yr from cycle time and USD 150 k/yr from a drop in off-spec batches from 4.2% to 1.9%. The site's capital hurdle is an 18-month simple payback plus a 12% discounted return test; the project as written returns 3.0 years. Finance notes the pilot ran with a dedicated day-shift operator, that no campaign changeovers occurred during the trial, and that the off-spec credit is already booked in the quality budget. Two of the three reactors share a common CIP header, so they cannot clean simultaneously. The plant currently runs at roughly 78% of demonstrated capacity.
1. The plant manager wants the business case re-cut until it passes the 18-month hurdle. Which benefit claims do you remove or re-baseline, and what payback do you report honestly?
Read the worked answer
Work through the claims individually rather than applying a blanket haircut.
Remove the off-spec credit. Finance has already booked it in the quality budget, so including the USD 150 k counts the same batches twice. Removing it takes annual benefit to USD 620 k.
Re-baseline cycle time. The 22% figure was measured with a dedicated operator and no changeovers. Changeovers and cleaning transitions are exactly where recipe handling earns its money, and you have no data for them. Also, two reactors share a CIP header, so any cleaning-related saving on those two is not additive. Model the constrained resource, not a three-reactor average.
Test whether released hours are cash. At 78% loading the plant can already make the demand. Released reactor hours convert to money only if the site can sell incremental batches, avoid overtime, or defer a capacity project. Value it as incremental batches x contribution margin, not hours x blended rate. Under the stated loading the defensible cycle-time value may be close to zero unless a debottleneck case is made.
Rebuilt case. USD 1.85 M capital plus the PV of USD 210 k/yr recurring. Realistic recurring benefit probably sits in the USD 150-250 k/yr band once yield, rework and manual-intervention savings are counted, giving a payback well beyond 18 months. Say that plainly and present the sensitivity band; a case that survives scrutiny at 30 months beats one that collapses at the first audit.
2. What pilot or staged evidence would you require before releasing full capital, and how would you define the gate criteria?
Read the worked answer
Convert the business case from a promise into a measured experiment with defined gates.
Gate 1 - representative baseline (4 weeks, no capital). Instrument the current process and capture cycle time on all three reactors across at least two full campaign cycles including changeovers and CIP. Publish the baseline distribution, not a mean. Add a second operator shift so the pilot is not flattered by a dedicated engineer.
Gate 2 - single-reactor pilot on Reactor 2 (8-10 weeks). Same measures, same crew rotation, changeovers included. Success = median cycle time reduction of at least X% (pick the number that makes the case work, not the number the pilot happened to deliver) with no increase in off-spec or cleaning failures, and operator intervention count not rising.
Gate 3 - constraint test. Deliberately run the shared CIP header conflict: schedule two reactors for simultaneous cleaning and confirm the new sequence resolves it without extending the campaign. This is the risk Finance has not been shown.
Gate 4 - value realisation before Phase 2. Release Reactors 1 and 3 capital only when the Phase 1 savings appear in the plant's actual cost reporting for two consecutive months, using an agreed benefit-tracking method signed by Finance.
Document each gate's pass/fail criteria and the person who signs it before the project starts. Staged funding is not a compromise; it is how you keep a 12% return test meaningful when the pilot sample is six weeks long.
3. Which non-financial benefits are legitimate to present, and how do you keep them from becoming disguised financial claims?
Read the worked answer
There are real benefits that should be stated, but each needs a label that prevents it being re-added to the cash case.
Legitimate, non-monetised:
- Electronic batch records and automatic reconciliation reduce transcription errors and make deviation investigation faster - state this as reduced investigation hours per deviation, with the current count, and leave the money out.
- Recipe version control and change audit trail address a known compliance finding; cite the finding reference, not a fine-avoidance number.
- Reduced operator cognitive load at changeover, which is a safety and quality argument.
- Shorter, more predictable changeover makes campaign scheduling more reliable, improving service level even at the same total hours.
What to avoid: 'avoided cost of a future second plant', 'avoided regulatory penalty', and 'avoided headcount' unless there is an approved headcount reduction plan. These are the classic items that turn a defensible case into one that fails a post-implementation audit.
Method. Present them in a separate section headed 'Benefits not claimed in the financial case', with the reason each is excluded (not convertible to cash, already counted elsewhere, or outside the project's control). Have Finance initial that section. If the plant manager later wants one monetised, it must enter the case through the same scrutiny as the cycle-time claim - with a measured baseline and a conversion mechanism.