HPU Forecasting for Proposals & EACs

Turn learning curves into defensible forecasts: forward pricing rates, lot buys, proposal HPU, T1 and learning rate sensitivity, and links to EAC.

22 min read 5 sections Guide 6 of 6, HPU Forecasting Updated

FORECAST →
FPR
Forward Pricing Rate
P50
Point Estimate Target
T1×LR
Sensitivity Drivers
EAC
Estimate at Completion

Forward Pricing Rates and Labor Cost Estimation

HPU forecasts predict hours. To convert hours to dollars, you need labor rates, and the rates must match the time period when the work will be performed. This is where forward pricing rates (FPRs) come in.

WORK PACKAGESPLANNING PKGUBMR
Budget baselineThe budget stacks as layers: work packages, undistributed budget and reserve on topThe Budget baseline guide
Rate ComponentWhat It IncludesTypical Annual Escalation
Direct labor rateBase wages for touch labor by skill category2 to 4%
Fringe/benefitsHealth insurance, retirement, FICA, PTO3 to 6%
OverheadSupervision, facilities, indirect support, depreciation1 to 3%
G&ACorporate management, legal, finance, IR&D1 to 2%

The total burdened rate (direct + fringe + overhead + G&A) is what converts HPU to cost. On a typical aerospace program, the burdened rate is 2.5 to 3.5 times the direct labor rate. A $40/hour direct mechanic costs $100 to $140/hour fully burdened.

Rate Timing Matters

If your program spans 2027 to 2032, you cannot use 2027 rates for all five years. Each lot must be priced at the FPR for its delivery year. A 3% annual escalation compounds to 16% over five years. On a $500M direct labor program, using a single year’s rate instead of escalated rates understates the estimate by $40 to $80M. Always map lots to fiscal years and apply the corresponding FPR.

Lot-Buy Estimation

Most aerospace production contracts are structured as lot buys: a fixed quantity of units purchased in each contract year. The HPU estimate for each lot is built from the learning curve using lot midpoints (covered in the Aerospace Application guide).

80% CURVEHPU
Learning curveEach time quantity doubles, hours per unit fall to the rateThe Learning curve guide
Lot-Buy Estimation Workflow End to End

Step 1: Define lot structure (units per lot, delivery schedule)

Step 2: Calculate lot midpoint for each lot

Step 3: Apply learning curve at each midpoint to get representative unit hours

Step 4: Multiply midpoint hours × lot quantity = total lot hours

Step 5: Apply rate adjustments, disruption factors, and engineering change impacts

Step 6: Multiply adjusted lot hours × FPR for the delivery year = lot cost

Step 7: Sum all lots = total program direct labor cost

LotUnitsMidpointMP HoursLot HoursFPR YearRateLot Cost
LRIP 11 to 125.87,34088,080FY28$112$9.86M
LRIP 213 to 3622.45,420130,080FY29$115$14.96M
FRP 137 to 8457.24,280205,440FY30$119$24.45M
FRP 285 to 148113.63,640232,960FY31$123$28.65M
FRP 3149 to 200172.33,320172,640FY32$126$21.75M

Total program direct labor: $99.67M across 200 units and 5 lots.

Proposal HPU Development

Developing HPU estimates for a proposal is a structured process that must withstand customer review, DCAA audit, and internal management scrutiny. The estimate must be traceable, defensible, and realistic.

BREAK
Rate disruptionThe learning curve is running down when a break bumps hours back up, then it recoversThe Rate disruption guide
Proposal PhaseHPU ActivitiesKey Outputs
Basis of Estimate (BOE)Document T1 source, learning rate rationale, rate assumptions, disruption factorsBOE narrative with supporting data
Cross-checkCompare to analogous programs, parametric models, independent estimatesComparison matrix showing consistency or explaining differences
Risk assessmentIdentify T1, learning rate, and rate change uncertainties; assign probability distributionsRisk register with quantified impacts
Management reviewPresent point estimate, confidence interval, risk-adjusted rangeApproved HPU estimate with documented assumptions
Customer submissionFormat per RFP requirements (CSDR, FlexFile, or contract-specific)Compliant cost volume with learning curve backup

The “Should-Cost” Trap

Management often pushes for a “should-cost” estimate that reflects aspirational learning rates or optimistic T1 values. This is dangerous in proposals because it creates a baseline you cannot achieve. The proposal should reflect the most likely outcome (P50), not the best case. If management wants to bid below P50, that is a business decision, but the technical estimate must be honest. Document the P50 estimate and the delta between it and the bid price as quantified risk.

Sensitivity Analysis: T1 × Learning Rate Uncertainty

The two parameters that drive HPU forecasts (T1 and the learning rate) both carry uncertainty. Sensitivity analysis quantifies how forecast uncertainty grows as these parameters vary within their plausible ranges.

P80
Monte CarloThousands of runs fall into place as a histogram of finish dates, with the P80 markedThe Monte Carlo guide
Two-Variable Sensitivity Matrix Unit 100 Hours

Base case: T1 = 10,000 hours, Learning rate = 85%

T1 ↓ / LR →83%85%87%
9,0002,5702,8493,142
10,0002,8563,1653,491
11,0003,1413,4823,840

Range: 2,570 to 3,840 hours (±20% around the base case)

Key insight: A 2-point change in learning rate has more impact on unit 100 than a 10% change in T1. At unit 200, the learning rate effect is even more dominant. This means that for later lots, getting the learning rate right matters more than getting T1 right.

For formal proposals, extend this to a Monte Carlo simulation. Assign distributions to T1 (typically normal or triangular) and learning rate (typically triangular or beta). Run 10,000 iterations. The output is a probability distribution of total program hours that yields P10, P50, P80, and P90 values directly.

Connecting HPU Forecasts to EAC

The Estimate at Completion (EAC) is the living forecast of total program cost. The HPU learning curve is the engine that drives the direct labor component of the EAC. Keeping the EAC current requires updating the learning curve parameters as actual data accumulates.

BACEAC
Estimate at completionOne method: EAC = BAC ÷ CPIThe Estimate at completion guide
EAC ComponentSourceUpdate Frequency
Actuals to dateAccounting system (incurred hours × actual rates)Monthly
Forecast remaining hoursLearning curve projection from current regressionMonthly or at each unit completion
Forecast remaining ratesForward pricing rates by fiscal yearAnnually or when FPRs are renegotiated
Risk/opportunity adjustmentsIdentified risks and their quantified impactsQuarterly at program reviews
EAC Variance Analysis Proposal vs. Actual

Proposal baseline: T1 = 10,000 hours, 85% learning rate, 200 units = 832,000 total hours

Current actuals (50 units complete): Regression gives T1 = 10,400 hours, 86.2% learning rate

Revised forecast:

MetricProposalCurrent EACVariance
T110,000 hrs10,400 hrs+4.0%
Learning rate85.0%86.2%+1.2 pts
Total program hours832,000912,000+80,000 (+9.6%)
Remaining hours (units 51 to 200)512,000576,000+64,000 (+12.5%)

The 4% T1 variance and 1.2-point learning rate variance compound to a 9.6% total program hour overrun. On a $100M direct labor program, this is a $9.6M cost growth, and it will only be visible if the learning curve is updated regularly with actual data. Programs that do not update their curves discover overruns too late to take corrective action.

The Bottom Line

HPU forecasting connects learning curve analysis to dollars and decisions. Forward pricing rates must match the work period. Lot-buy estimation uses midpoint hours multiplied by lot quantity and year-appropriate rates. Proposals must document the basis of estimate, cross-check against analogies, and present risk ranges, not single points. Sensitivity analysis shows that learning rate uncertainty dominates T1 uncertainty for later lots. The EAC must be updated monthly using the latest regression, not the original proposal assumptions. A 1-point error in learning rate can produce a 10%+ cost growth on a large program. The entire HPU Forecasting track (from theory through disruption modeling) builds toward this: turning data into defensible, accurate cost forecasts that support winning proposals and realistic program baselines.

Frequently asked questions

What is the difference between a forward pricing rate and a billing rate?

A forward pricing rate is a projected future labor rate (dollars per hour) used for estimating costs on proposals and contract modifications. It includes anticipated wage increases, benefit changes, and overhead rate projections. A billing rate is the actual rate charged to the contract based on incurred costs. Forward pricing rates are negotiated with DCAA/DCMA and are typically locked for 1 to 2 years. The distinction matters because using current billing rates on a 5-year program will understate costs by 10 to 20% due to inflation and rate escalation.

How do you present risk ranges to a customer?

Present three scenarios: point estimate (most likely), management reserve range (P50 to P80), and risk-adjusted range (P20 to P80 or P10 to P90). Use a tornado chart to show which inputs drive the most uncertainty: typically T1 and learning rate dominate. Never present a single number without context. Customers and program managers need to understand that the forecast is a distribution, not a point. The point estimate should be the P50 (median), not the optimistic or ‘should-cost’ value.

How do you connect HPU forecasts to EACs?

The EAC for direct labor is: Actuals to date + (Forecast HPU for remaining units × Forward pricing rate). Update the HPU forecast each month using the latest regression from actual data. The EAC should reflect the current best estimate of T1 and learning rate, not the original proposal values. If the actual learning rate is 87% and the proposal assumed 85%, the EAC must reflect 87%. This is variance analysis: identifying whether the program is performing better or worse than planned and by how much.

Share this guide: LinkedIn X Follow SymplProcess on LinkedIn

Take this to a room

The running order

For proposal and EAC teams. They should leave committed to P50 with a documented basis of estimate and year-matched rates.

7 beats · 13 min
  1. 1

    Match the rate to the year

    A programme spanning five years cannot be priced at one year's rates.

    • Each lot is priced at the forward pricing rate for its delivery year.
    • Three per cent annual escalation compounds to 16 per cent over five years.
    • On a 500 million dollar direct labour programme, that is 40 to 80 million understated.

    Ask the room Are our lots mapped to fiscal years with matching rates?

  2. 2

    Lot-buy estimation uses midpoints

    Not the arithmetic mean of the lot. The lot midpoint, which is a different number.

    • Midpoint hours multiplied by lot quantity, at the year-appropriate rate.
    • Using the arithmetic mean overstates early lots and understates later ones.
    • The error grows with lot size.
  3. 3

    Bid the most likely, not the best case

    The proposal should reflect P50. That is a technical statement, not a commercial one.

    • Should-cost estimates built on aspirational learning rates create a baseline nobody can hit.
    • If management wants to bid below P50, that is a business decision, and a legitimate one.
    • But the technical estimate stays honest, and the gap is documented.

    Ask the room Does our proposal distinguish the technical estimate from the bid?

  4. 4

    Document the basis of estimate

    Every number in a proposal will be questioned. The BOE is the answer you prepared earlier.

    • Where T1 came from, which model, which rate, which data.
    • Cross-check against analogies from comparable programmes.
    • A BOE written after the award is worth nothing.
  5. 5

    Sensitivity: rate dominates T1

    When you run the two uncertainties together, one of them matters far more.

    • T1 uncertainty affects early units most.
    • Learning rate uncertainty compounds, so it dominates for later lots.
    • A one-point error in learning rate produces very large errors on a long programme.

    Ask the room How confident are we in our rate, to the nearest point?

  6. 6

    Give ranges, not points

    A single number in a proposal becomes a commitment the moment it is read.

    • Present a risk range with the point estimate inside it.
    • Show which assumptions move the range most.
    • This is also your defence when the customer challenges the number.
  7. 7

    Update the EAC monthly

    Not from the proposal assumptions. From the latest regression on actual data.

    • The proposal was the best estimate before any units were built.
    • Once you have production data, that data is better than the assumption.
    • An EAC still running on proposal assumptions at unit 40 is not a forecast.

    Ask the room What is our EAC based on right now?