The Project Manager’s Corner: How Much Will TMT Cost?
Gary Sanders, TMT Project Manager

August 2006

We are working on the answer to this important question. Very hard. Price quotations arrive every day. New requirements documents and requests for quotes go out, as well, every day. Cost-estimating tally sheets, with narrative descriptions of the design and estimating method, are stacking up. A data base is filling. Some items are coming out higher than expected; others leave a bit of room within our expectations.

Later this year we will have laid out a detailed cost estimate, and outside experts will pore through it and render an opinion on what has been left out, or what’s been judged with too much optimism or—perhaps—too conservatively. We will refine some things and then we will have a basis for planning adjustments to what’s included in the TMT design, and how the project can be carried out within budget.

We have all heard of those megaprojects that turn out to cost much more than originally stated. The tunnels in big cities or under the English Channel, or the Superconducting Super Collider, or even just that freeway through town, all described in the newspapers to be 2 or 3 or 4 times the original cost. How can that happen? Are there reliable methods to estimate and plan the costs of big, one-of-kind projects? Do the methods that apply to airports and bridges also apply to big science projects?

The first thing to ask: how can such large overruns happen? Generally, it is due to changing the project requirements or its design in midstream. Politics, changing customer desires, or lack of focus on building what was agreed upon are invitations to cost growth. The costs change because the project has been changed. But even if you stick to the original project, there is risk. There are, however, methods to prepare a cost estimate and to plan for those risks.

There are many reliable methods that can point to reasonably accurate costs. There are also methods to keep on track to work within those costs as the project is carried out. Farther on, we will describe some of these methods and how we are going about this in TMT.

Airports and cruise ships and airliners and big shopping centers and major bridges and tunnels are all quite risky, and they tend to be one-of-a-kind projects. Experience with previous projects provides only a rough guide to the current project. Requirements and circumstances and design solutions are often sufficiently different to limit past experience as a reliable guide.

Science projects seem to be even more challenging and therefore more risky. They often involve new technologies and new materials, and even outright inventions in the common sense of the word. We scientists like to think that what we do is so special that the methods used on those “conventional” projects in the previous paragraph cannot fully apply. But this is not true.

As challenging as TMT is, a new generation of cruise ships, or the English Channel tunnel, or that 169+ story skyscraper in Dubai are just as novel and risky. Cost risks can be treated in the same way in TMT as in those projects. The best practices used in these megaprojects apply to TMT.

There are several ways to estimate project costs. Generally, the first estimates are done by finding out all of the costs of the most closely related projects, and estimating the new project costs item-by-item or subsystem-by-subsystem. To use a specific analogy: we know the costs of the Keck and Gemini telescope mirrors. Can’t we just scale those costs by part count or mirror area or pounds of glass or some such parameter? Can’t we do that for every part of TMT that has an analog in those projects? TMT might even use similar designs and the same suppliers.

We can do that and we would learn a great deal. But this technique will not capture the cost of all of the project risks and the cost impact of significant differences in requirements. TMT mirror segments are more aspheric. They are different in size. We may have little guidance whether we can polish twice as fast if two or more segments can go on the same machines, compared to the larger Keck segments. Or the steeper curvature may make the polishing and testing much more laborious. Making 740 segments for TMT, instead of Keck’s 36 segments, means we have a big chance to take advantage of the savings from “almost mass production”.

Concrete and steel prices have zoomed upward in the last few years across the globe. Though Keck and Gemini were built only a decade ago, the world market in commodities, energy, labor, and transport are dramatically different. TMT may ship about 5,000 tons of fabricated steel elements across oceans to its site. Ocean freight is a brave new world.

Analogies to the past and parametric scaling, though quite useful, leave too much uncertainty. We have to do the estimate for TMT in excruciating detail, using what is called a “bottom-up” estimate.

Every piece of work in the design and fabrication and assembly and testing of TMT must be collected in a giant list, organized according to all of the delivered pieces of TMT. This list is called a Work Breakdown Structure (WBS). Having done a conceptual design of TMT, we know what most of the deliverable elements of TMT are. And we know the steps to deliver each.

We prepare a table with a tree-branch structure (a list of lists) with the TMT as a whole at the top. The main branches of this tree are the main subsystems (summit facilities, enclosure, telescope, adaptive optics, the instruments, and so on…). Each main branch has smaller branches representing the delivered sub-subsystems or components. The tree diagram is worked down to as many as nine levels of branches.

Every piece of work or item to be delivered in order to build TMT is represented. Right now, this WBS, including all project phases and items, fills more than 700 pages. The lowest-level pages in this tree structure describe the estimate for delivering a single item in the WBS, including needed labor, purchased items, contracted fabrication, travel and a tabulation of cost, schedule and technical risks for that item. The complete tree-structured estimate is captured in a big database.

The engineer or scientist responsible for each of the main or mid-level branches in the estimate is responsible to prepare the estimate for all of their deliverable items or tasks. They generate the estimate at the lowest level in this tree of lists. This is called “bottom-up” estimating.

All of the estimates are added up to produce the larger delivered and testable items. Science instruments, primary mirrors, the telescope structure, the secondary mirror assembly, and the top-level observatory software are all deliverable items. When combined, they become TMT.

As the cost estimate is rolled up, project designers can consider ways to reduce costs or to eliminate items or—perhaps—to add some feature to TMT thought to be out of reach. This is part of designing-to-cost. And it is the opening round of staying on budget.

As project manager, I look forward to that night on the mountain when the first starlight will fall on the giant mirror and the very first inkling of new science will be revealed. But, in the back of my mind, I know that I will also be thinking of how we stayed on budget.