Demand and Benefit Forecasts by the Numbers
Chapter 8 of 10 · 10 min read
Tracking benefits against a phased plan
A benefits register for the canning plant, a plan phased by season, quarterly realisation from 0.060 to 0.692, leading indicators, and a 16.441m run-rate that was really 14.193m.
The board’s first condition was quarterly benefit tracking with leading indicators. It is easy to agree to and easy to do badly. A tracking report that compares each quarter’s result with a quarter of the annual plan will show a seasonal business failing every winter and succeeding every summer, and a report that only counts money in the bank will show a shortfall months after anything could have been done about it.
This chapter sets up the plant’s benefits register, phases the plan by season so that each quarter is judged against what it should have produced, reads the first seven quarters, chooses indicators that move before the accounts do, and shows how to state a run-rate for a seasonal business. It ends with a quarterly report that annualised the busiest months of the year.
The benefits register
This is an illustration. All money is A$ excluding GST. In January 2025, before the first commercial run, the finance manager turned the approved business case into a register of the benefits the plant was meant to deliver, the dis-benefit already identified, and who would answer for each. The register followed the shape of public-sector practice in Australia, such as NSW Treasury’s Benefits Management Guide (TPG24-31), which asks for a plan and a register kept up to date from appraisal through delivery, but in a form a business of this size can maintain in a spreadsheet.
| Owner | Measure | Baseline | Target | Position at month 21 | |
|---|---|---|---|---|---|
| B1 Plant contribution | managing director | contribution, A$k a quarter | nil, new plant | A$2,241.1k in 2025; A$3,041.5k in 2026; A$3,201.6k a year from 2027 | run-rate A$2,109.0k a year, 0.659 of steady state |
| B2 Sale of mobile rig 2 | finance manager | proceeds and date | rig in service | A$320,000 in month 6 | A$255,000 in month 14 |
| B3 Can reject rate | operations manager | rejected filled cans, % | 2.5% on the mobile rigs | 0.8% by month 12 | 0.9% |
| B4 Diversification | sales manager | non-beer share of volume, segments B and C | mobile rigs fill craft beer | 75.8% | 85.0% |
| Dis-benefit: mobile contribution lost | sales manager | contribution of mobile customers moving to the plant | nil | no more than A$57,600 a year | four former mobile customers moved in month 1 |
Three features of the register matter more than its layout. Every measure is something the business already records: contribution from the job costing, rejects from the line counter, volume by customer from the invoices. Every target comes from the business case the board approved, not from a revised figure, following the rule from Chapter 5 that a forecast is scored against the version the decision used. And the dis-benefit identified in Chapter 2, the A$57,600 a year of mobile contribution lost when existing customers move to the plant, has its own line and owner, so that it cannot quietly disappear from the accounting. Mapping benefits back to business changes and enablers in Outcome-First Projects by the Numbers shows how each such line can be traced to the business change that has to happen for it to arrive.
Phasing the plan by season
The business case gave annual figures: 70% of steady-state contribution in 2025, A$2,241.1k, and 95% in 2026, A$3,041.5k. The mobile business’s records give the seasonal pattern of canning in the region: shares of the year of 0.290 in January to March, 0.210 in April to June, 0.200 in July to September and 0.300 in October to December. Summer and the run-up to Christmas are busy; winter is quiet.
A flat plan, one quarter of the annual figure each quarter, would have set the target for January to March 2025 at A$560.3k. The phased plan, 0.290 of A$2,241.1k, set it at A$649.9k. In the winter quarters the phased target is lower than the flat one, and in the summer quarters higher. Without phasing, the report for April to June would have flattered the plant and the report for October to December would have alarmed the board, with neither telling it anything about the forecast.
The phased plan is still the inside view’s plan, with its flat 70% first year. Chapter 4 showed that the rig 2 record pointed to a year-1 average nearer 0.367, and the reviewer had argued for a ramp-shaped plan. The register nevertheless tracks against the approved plan, because that is the forecast the decision rested on; a ramp-adjusted comparison can be shown beside it as a second line, clearly labelled, but it does not replace it.
Seven quarters of realisation
Realisation here is actual contribution divided by the phased plan for the same quarter. A value of 1.000 means the quarter delivered what the business case said it would.
The pattern has two parts. Through 2025, realisation rose steadily as customers arrived and the line learned, from 0.060 in the first quarter to 0.702 in the fourth. For 2025 as a whole the plant earned A$882.8k of contribution against a plan of A$2,241.1k, a realisation of 0.394. Then in January to March 2026 realisation fell to 0.574, although actual contribution rose from A$471.7k to A$506.4k. The fall came from the plan, not the plant: the business case stepped from 70% to 95% of steady state on 1 January 2026, and the phased target jumped from A$672.3k to A$882.0k. A reader who saw only the ratio would conclude that the plant had gone backwards. The register’s practice of showing the actual and the plan beside every ratio prevents that misreading.
By the third quarter of 2026, realisation had recovered to 0.692 and was rising slowly. Cumulative realisation to September 2026 was 0.507: in 21 months the plant had earned about half of what the business case said it would. Schedule, benefits and the cost of delay in Reference Class Forecasting by the Numbers found first-year realisation ratios with a median of 0.62 across five past automation projects; the plant’s first-year 0.394 is well below that, and its ramp is the main reason.
Leading indicators
Contribution is a lagging measure: by the time it shows a shortfall, the causes are months old. The board asked for leading indicators, measures that move before contribution does and give time to act. The onboarding chain from Chapter 4 supplies most of them, because every new brand passes through the same steps: a signed supply agreement, an order for printed cans with a 14-week lead time, three trial runs over four weeks, and then the first commercial run, about 18 weeks after signing.
The register tracked seven indicators each month:
- Signed supply agreements, by segment, against the number the plan needed. An agreement signed this quarter becomes volume about four months later.
- Printed-can orders placed by customers. A brand that has ordered printed cans has committed real money; one that has not, 14 weeks before its planned start, will not start on time.
- Trial runs completed, which show that a product and can format run on the line.
- Quotes issued and won, the earliest signal of demand in each segment.
- Deseasonalised monthly volume, described below.
- Line effective rate, against the learning curve of 5,000 cans an hour in month 1 rising to 9,000 by month 9.
- Reject rate, benefit B3, which is itself a leading indicator of cost.
The indicators told the story early. By month 12 the plant had 24 active customers, A 12, B 3, C 8 and D 1, against 50 in the business case, and the pipeline of agreements and can orders made it plain that the gap would not close in 2026. That is what leading indicators are for: the board knew in early 2026 that the plan’s 95% year would not happen, rather than learning it from the 2026 accounts.
Benefits B2, B3 and B4
B2, the sale of rig 2, was planned at A$320,000 in month 6, June 2025. The rig sold for A$255,000 in month 14, February 2026, because craft customers kept booking the mobile rig while the plant ramped up and the business could not release it. Discounted monthly at 9% a year from January 2025, the plan was worth A$306,504 and the result A$230,609, about A$75.9k less. B2 is a reminder that benefits interact: a slower ramp delayed a benefit that seemed independent of it.
B3, the reject rate, started at 2.1% in month 3, fell to 1.6% by month 6 and 1.2% by month 12, missing the 0.8% target for that date, and reached 0.9% by month 18, where it stayed at month 21. At A$0.30 per rejected filled can, each percentage point of rejects on 16.950m cans a year costs A$50,850 a year. The remaining 0.1-point gap to target is worth A$5,085 a year; the fall from the mobile rigs’ 2.5% to 0.9% is worth about A$81,360 a year at the same volume. B3 was late, nearly achieved and worth having.
B4, diversification, exceeded its target: 85.0% of run-rate volume came from RTD and non-alcoholic brands, against 75.8% in the plan. That looks like the clearest success in the register, and it is the line most in need of explanation, because it rests largely on one non-alcoholic customer. Upside surprises and honest accounting takes it apart.
A run-rate for a seasonal business
A board wants to know what the plant is running at now, in annual terms, so that the number can be compared with the business case’s 25.32m cans and the switching volume of 19.311m from Chapter 7. For a seasonal business the obvious method, the last three months multiplied by four, is wrong in a predictable direction: it overstates in summer and understates in winter.
The register used a deseasonalised run-rate. Each month’s volume is divided by its seasonal index, the month’s share of the year multiplied by 12, which runs from 0.72 in June and July to 1.38 in December. The result is what the month would have been in an average month. The average of the last three deseasonalised months, multiplied by 12, is the run-rate.
This is an illustration. In July, August and September 2026, months 19 to 21, the plant filled 1,013.8k, 1,103.5k and 1,273.3k cans. With indices of 0.72, 0.78 and 0.90, the deseasonalised months were 1,408.0k, 1,414.8k and 1,414.8k cans, and the run-rate was 16.950m cans a year: 66.9% of the business case, 110.3% of the outside view’s expected 15.367m, and below the switching volume of 19.311m. The raw quarter multiplied by four would have given 13.562m, understating the plant by 3.4m cans because July to September is the quietest quarter of the year.
Deseasonalised actualBusiness case plan
Share of 25.32m cans a yearMonth of operation
| Month of operation | Deseasonalised actual | Business case plan |
|---|---|---|
| 1 | 0.013 | 0.70 |
| 2 | 0.032 | 0.70 |
| 3 | 0.054 | 0.70 |
| 4 | 0.138 | 0.70 |
| 5 | 0.204 | 0.70 |
| 6 | 0.283 | 0.70 |
| 7 | 0.334 | 0.70 |
| 8 | 0.368 | 0.70 |
| 9 | 0.451 | 0.70 |
| 10 | 0.485 | 0.70 |
| 11 | 0.500 | 0.70 |
| 12 | 0.534 | 0.70 |
| 13 | 0.545 | 0.95 |
| 14 | 0.561 | 0.95 |
| 15 | 0.576 | 0.95 |
| 16 | 0.606 | 0.95 |
| 17 | 0.621 | 0.95 |
| 18 | 0.633 | 0.95 |
| 19 | 0.667 | 0.95 |
| 20 | 0.670 | 0.95 |
| 21 | 0.670 | 0.95 |
Deseasonalised, the plant’s progress becomes a smooth curve that can be read month by month: steep through 2025 as customers arrived, flattening through 2026, and level for the last two months. The flattening is itself information. It says that the growth left in the plant depends on new customers, not on the existing ones settling in.
Tetlock and Gardner, in Superforecasting, stress keeping score and updating forecasts in small steps as evidence arrives, rather than holding a view until it collapses or swinging with every new figure. A deseasonalised monthly run-rate supports exactly that: each month moves the estimate a little, and a change of direction shows up within a quarter. Benefits move after approval: tracking erosion, surprises and dis-benefits describes the kinds of movement a register should expect to record.
The error: annualising the busy season
This is an illustration. The managing director’s report for the quarter to March 2026 stated a run-rate of 16.441m cans a year: January to March volume of 4.110m cans multiplied by four. The report presented this as evidence that the plant was already at two thirds of the business case and on track for the switching volume within the year.
January to March is the business’s busiest quarter after October to December, with seasonal indices of 1.26, 1.14 and 1.08, an average of 1.160. Dividing each month by its own index and averaging gives a deseasonalised run-rate of 14.193m cans a year, 2.248m lower. The finance manager detected the error when reconciling the report with the register, which stated the run-rate deseasonalised every month, and saw that the two figures for the same quarter differed by about 16%.
The correction was a reporting rule: the plant’s run-rate is reported only in deseasonalised form, with the seasonal indices used stated in a footnote, and any raw annualised figure is labelled as such. The same rule protected the plant six months later, when the raw July to September quarter would have understated it at 13.562m. A rule that removes the season works in both directions, which is why it is easier to agree to in advance than when the error happens to flatter the person reporting.
Exercise: the winter quarter
The plant’s raw volumes for April, May and June 2026 were 1,228.3k, 1,101.5k and 962.1k cans. Calculate the raw annualised run-rate and the deseasonalised run-rate for the quarter, and say which should go in the report.
Worked answer. The quarter’s volume is 1,228.3 + 1,101.5 + 962.1 = 3,291.9k, about 3.292m cans, and four times that is 13.167m. The seasonal indices are 0.96, 0.84 and 0.72, summing to 2.52, so the deseasonalised run-rate is 3.292m ÷ 2.52 × 12 = 15.676m cans a year. Dividing each month by its own index and averaging gives 1,279.5k, 1,311.3k and 1,336.3k, and 12 times their average is 15.708m. Either method is defensible if it is used consistently; the register uses the month-by-month version. The raw figure understates the plant by about 2.5m cans because April to June is a quiet quarter, so the deseasonalised figure goes in the report, rising from 14.193m in the March quarter to about 15.7m in the June quarter.