EE.      White Paper (A Parable of Cost Estimating in Aerospace?)

A Simple Calculation of Cost of an Aerospace Facility Compared to a Data Point

By Ester Mator, NASA KSC

The phone rings…

Question: How close would an expert cost estimator, or the tools they use, come to a more detailed, backed up cost figure if trying to estimate the cost of a major aerospace facility such as the Kennedy Space Center (KSC) Space Station processing Facility (SSPF)? They need an answer by 2pm. It’s 1pm.

An estimator sometime in the past has been asked to predict, with very little information, the “kind of cost” that might be incurred to build a Space Station Processing Facility at KSC. The estimator is given very little information, as the project is still in its conceptual phases, which is to say that only vague notions of what the building is and what it will do are known. Many inches of paper already abound about the project, none of it of any use in figuring out what the thing will cost. It will process space station elements. This is akin to processing payloads. It will handle anywhere from one to a few of these elements, possibly more, per year, of varying size, function and complexity. These payloads will be launched on the Space Shuttle.

The estimator (being an expert, and knowing that he/she is skilled enough to estimate anything, even in the absence of data or detail) begins by reducing the known to plainly a few factors. After all they want a 1 hour answer.

1.         The business will not re-invent itself too far from what it may have done in any similar case recently. Seek a data point.

2.         It’s payloads.

3.         It’s Shuttle payloads.

4.         Shuttle payloads must fit in a bay that measures 15 ft X 15 ft X 60 ft, so the size of the “things” is known. Size must be a factor.

Historical data?

The estimator (being an expert, here “expert” meaning knowing where to find the answer and claim credit as if one invented it oneself) is aware of a cost estimating relationship (CER) that was documented some years before.

For Payload/Cargo Processing:

(Equation 1)
CofF (Construction of Facilities) $M= 11.99*(202+L)*(56+W)*(45+H)/10^6*Esc

Where:

L=Length of Payload

W=Width of Payload

H=Hieght of Payload

Esc=Escalation for inflation as required, starting at 1987 Year $.

(Equation 2)
GSE (Ground Support Equipment) $M=CoF*2.06

The estimator reads on about the CER and finds that the CER relationships come from a Vertical Processing Facility (VPF) built at KSC many years earlier. Historical data had been used to make up the CER. This data was:

VPF=$19.4M to construct a facility that was 217 ft X 71 ft X 105 ft in dimensions, and that outfitting the facility with GSE had cost $39.96M (all in fiscal year $1987 dollars).

Hence, the Shuttle having a known payload set of dimensions would allow one to reverse engineer the CER factors. The CER of “11.99” is actually just a cost per cubic foot of $19.4M divided by the buildings volume [19.4*10^6/(217*71*105)=11.99]. The CER of 2.06 is actually just the $39.96M divided by the facility cost, or that is a GSE ratio of cost as compared to the facility cost (39.96/19.4=2.06).

Simply enough, the previous estimator had simply subtracted out the Shuttle payload dimensions to leave “space” factors in calculating for a new size payload. Surely, this broad assumption that the space that we required when we did this once is space we’ll likely require again would lead to errors in the future using such a relationship far from the Shuttle payload dimensions? No problem - in this case, the Space Station facility question, it is Shuttle payloads, exactly, so the estimator proceeds.

The CofF and GSE simply is the “same” as a VPF, except that escalation for inflation is required.

CofF (SSPF) =19.4*1.15=$22.3M

GSE (SSPF) =39.96*1.15=$45.96M

The escalation factor of 1.15 for taking 1987 dollars to 1990 dollars is taken from NASA inflation indexes.

The estimator knows one other factor to consider in the calculation before declaring an answer. The SSPF is planned to process payloads horizontally, not vertically. Perhaps an adjustment is required in the calculation? In either case, what would be the difference and how would it be known? There are 22 minutes left before the hour is gone. The estimator turns the equation on its side.

The equation would have been:

CofF (Vertical)=11.99*(202+15)*(56+15)*(45+60)/10^6*Esc = $19.4M

It is now (horizontally), with the same escalation:

CofF(SSPF Horiz.) =11.99*(202+60)*(56+15)*(45+15)/10^6*1.15 = $15.4M

And GSE is:

GSE(SSPF Horiz) =15.4*2.06=$31.7M

The estimator reports back (on time) the following:

An SSPF processing a single Shuttle payload at a time, horizontally, assuming it will be roughly 18,602 sq-ft (a product of 262 ft and 71 ft, L and W from above) will cost, in $FY 1990 dollars:

CofF (SSPF Horizontal) =$15.4M

GSE (SSPF Horizontal) =$31.7M

Upon further questioning of the estimate the estimator further adds that the number is an “as built” cost, such as what a major contract would come to, not including the costs on the government side of oversight or activation. The VPF historical data was for what was easily accountable and published after all, not the overhead and in-direct costs that would be more debatable and less traceable.

Some weeks pass, and the estimator must now refine his estimate for processing a “few” shuttle size payloads at a time. It begins to appear the facility will be quite larger than 18,602 sq-ft!

Now, being an expert estimator, and since, after 1 month of the estimate being passed to Washington, the White House and the Office of Management and Budget someone realized the facility was the wrong size,  the estimator undertakes the complex task of an “adjustment”.

How large will the SSPF facility be? The estimator receives guidance only that up to 3 payloads may be in flow at a time. Being an expert estimator, and this time having a whole week to document the new analysis, the estimator multiplies the previous estimate by 3 and increases his documentation to 60 pages of tables, graphs and explanation.

The cost is now, for duplicating the previous capacity facility, at 55,806 sq-ft:

CofF (SSPF Horizontal) =3*$15.4M=$46.2M

GSE (SSPF Horizontal) =3*$31.7M=$95.1M

And some months pass…

As some more months pass, the estimator sees his initial numbers come and go in discussion, reports, briefings, and general misunderstanding. Teams are formed to derive a better estimate, 60 pages of documentation from the cost estimator being deemed insufficient, as well as that contracts will soon be floated that require actual definition of the what the SSPF will be.

It is documented the SSPF facility will likely be in the range of “465,000 sq-ft and 3 stories high”. This would be a volume of about 155,000 sq-ft X 45 ft = 6,975,000 cubic-ft.

The estimator realizes this is finally useful information as the CER was volume based!

Easily enough, using Equation 1 again:

CofF = 11.99*(202+L)*(56+W)*(45+H)/10^6*Esc

Replace the L, W and H with the expected “box” volume and, (with Esc=escalation to year 1990, a factor of 1.15)…

CofF = 11.99*(6,975,000)/10^6*1.15

CofF = $96.2M

And GSE can be taken, using Equation 2 again:

GSE = 96.2 * 2.06 = $198M

The estimator happily reports back that the SSPF facility project is a “roughly $300M project to bring about”. He increases his documentation to 173 pages, with 172 pages of boilerplate referencing or copying other useless program documents (as required of the reviewers of his work, and to establish credibility) and 1 page for the previous calculation.

Not believing the result, a very high-up manager forms a “red team” to “hammer out the real numbers, in detail, from the bottoms up, without using these models and stuff no one believes…”.

And what actually happened…

In December of 1992 a case study Cost Engineering Report was published of the SSPF C-100 government estimate.

“This report is an important cost engineering tool for construction, activation, and GSE design, estimating, fabrication, installation, testing, termination, and verification of this over $380,000,000 (including GSE and activation) project.”

A bottoms up government cost estimate placed the SSPF CofF at ~$57M. This is shown below (estimate #2 on the list).

We are also aware from the same document of total costs for the SSPF, with GSE, including activation at ~$380M.

How well did our estimator do?

For comparison the estimator generated the following values for the SSPF query, based solely on knowing it was Shuttle, and it was horizontal:

Estimate 1:

An SSPF processing a single Shuttle payload at a time, horizontally, assuming it will be roughly 18,602 sq-ft (a product of 262 ft and 71 ft, L and W from above) will cost, in $FY 1990 dollars:

CofF (SSPF Horizontal, Single Capacity) =$15.4M

GSE (SSPF Horizontal, Single Capacity) =$31.7M

Later, a simple multiplier for quantity was used, 3 payloads at a time:

Estimate 2:

CofF (SSPF Horizontal, 3 Payloads at a time) =3*$15.4M=$46.2M

GSE (SSPF Horizontal, 3 Payloads at a time) =3*$31.7M=$95.1M

Lastly, as the building had actual dimensions attached, the estimator fudged a new result, based on the building (not the vehicle / payload dependency as a driver):

Estimate 3:

CofF (SSPF Horizontal, knowing only building volume)= $96.2M

GSE (SSPF Horizontal, knowing only building volume)= $198M

Estimate 1 and 2, for the caveat attached, can be compared against the eventual numbers reported.

}        $46.2M deviates from the $56.9M by an under-estimate of $10.7M, or ~19%.

}        The total of Estimate 2 (46.2+95.1=$141.3M) can not be compared to the known total of $380M, as Estimate 2 did not include activation and this quantity is unknown.

}        At a rough level, estimate 3, if building information is available even sketchily, shows how a knapkin and a good estimator can within minutes provide a major program factor of roughly a size that can give a “heads up” on where money will (or may) go. 96.2+198=$294M is a sizable factor to consider as “usable” even in the most conceptual of program phases. Later estimates and refinement using orders of magnitude more resources may only change such as value by less than 25%.

In closing…on the usefulness of early rough conceptual level cost estimating…

}        Estimates can be incorrect but still useful.

}        Moderate enhancement of an incorrect estimate can make it roughly correct for an infinitesimally small amount of effort as compared to sub-sequent efforts to refine an estimate using bottoms-up approaches.

}        A high level estimate can still be incorrect, as it’s easy to miss a detail when only a few details exist.

}        Find a good, honest estimator who’s done it before.