Pre Bid Risk Assessment: A Register Read From the Document

The bid or no bid decision, in most custom manufacturers I have watched, is not a decision. It is a feeling, formed in an email thread, usually after somebody has already spent a day reading the package. The feeling is about the customer, the competition and how busy the shop is. It is almost never about the sentence on page 9 that doubles the engineering, or the foam system on page 79 that nobody ordered, because nobody has read page 79 yet. A pre bid risk assessment is the thing that should replace the feeling, and the version most guides describe would not have caught either sentence.

This piece is about the version that would: a register read from the customer's document, page by page, before anybody builds a cost. Every example in it comes from a public bid package I have read for this site, and each one links to the article where I read it in full. I build software that does the reading, so take the last section with that in mind. The register itself needs no software at all.

In brief

• A pre bid risk assessment for an engineered product is not a score of location, competition and client. It is a list of the sentences in the customer's document that can cost you money, each with its page, its exposure and who decides.
• The generic five factor lists miss the money because the money is in sentences. A nuclear crane standard in a routine paragraph. A warranty that changes length eighty pages later. A delivery date a quarter of the normal lead time. A daily penalty against milestones the contract never names.
• Seven classes cover almost everything we have found: scope silence, internal contradiction, standards and approvals load, commercial terms, amendments, substitution language, and pricing rules you do not have written down.
• Score each row into three buckets, price it, exclude it or walk, not a one to five matrix. Contingency is the sum of the bracketed exposures, never a flat percentage.
• The register does not end at submission. On won jobs it is the only honest way to learn whether your risk pricing was right, and it becomes the company's risk memory.

What a pre bid risk assessment is, and what it is not

The phrase gets used for three different things, and only one of them is this article. Construction owners run a pre bid risk assessment on a project before they let it, to decide how to package and contract the work. Proposal teams run one on an opportunity, to decide whether the win probability justifies the cost of bidding; that is the bid or no bid decision as the proposal software vendors describe it, and it is mostly about competition, relationship and capacity. Both are real, and neither is what a custom manufacturer needs when a 99 page request for a fire truck or a 356 page tank package lands in the inbox.

What the manufacturer needs is narrower and harder. The customer has written a document. Somewhere in it are the sentences that decide whether the job makes margin, loses it, or cannot be done at the price anyone will pay. The pre bid risk assessment is the act of finding those sentences before the cost build, writing them down with their page numbers, putting a bracket on each one, and deciding what to do about it: price it, exclude it, or walk away. The output is a register. It is not a number, not a colour, and not a meeting.

It belongs at station two of the nine stations in the manufacturing quoting process, and it fails there for the same reason the whole process fails: it lives inside one person. The estimator who has done thirty of these knows where to look. The estimator who has done three does not, and neither does the sales manager making the bid decision from the cover letter.

Why the five factor lists miss the money

Search for bid or no bid guidance and you will find the same list in different clothes: project location, duration, size and scope, competition, client, and whether you have the capacity. All of it matters. None of it is where the expensive mistakes were in the packages I have read.

In a federal bridge crane specification for a flood control dam in Missouri, 27 pages long, the most expensive sentence sat in the middle of a routine safety paragraph on page 9: nuclear certification, testing and rules of construction shall be in accordance with a lock out regulation and ASME NOG-1. NOG-1 is the standard for cranes in nuclear facilities. A crane built to it is a different and far more expensive machine than the Class A crane the rest of the document describes, and I am fairly sure the sentence is a guide specification option nobody deleted. The location factor would have scored this job as easy. The sentence makes it either a clarification request or a machine nobody wants to build.

In a fire district's request for a pumper truck, page 16 lists the warranties that must be submitted with the bid, on forms that may not be altered on pain of disqualification. Pages 97 and 98, in the specification's own warranty section, ask for different lengths: the stainless body goes from 15 years to 20 years or 100,000 miles, the frame corrosion warranty from 20 years to 25. The same document also wants the truck delivered 120 days after contract signing, when the normal wait for such a truck was 12 to 18 months, and names roughly 60 components by brand with no exceptions allowed. The client factor would have scored this as a friendly municipal buyer. The document is a compliance matrix of about three hundred paragraphs, each with a YES and a NO printed beside it, and one blank for the price.

The pattern holds across every package on this site. A 668 page municipal wastewater book carried its requirements for one piece of equipment in 14 sections spread through it. A 179 page marine package stated the datum for freeboard in one place and the deck load in another and did not say which governed. A 356 page elevated tank bid set liquidated damages of $1,000 a day past substantial completion, again past final completion, and then a third rate against milestones the contract never named. None of these is a factor. Each is a sentence with a page number, and the only assessment that catches it is one that reads.

Seven risk classes for engineered bids

After enough packages the risky sentences sort themselves into a small number of classes, and knowing the classes tells a reader where to look. These are the seven we use, with where each one hides and a public example of it.

The seventh class surprises people, because it is not in the customer's document. It is in yours. A register that lists every risk the customer created and none of the ones your own process creates is half a register. The plate price that came from an email fourteen months ago is a pre bid risk in exactly the way the nuclear standard is, and it is the one you can fix without sending a question to anybody. The tribal knowledge piece covers where those rules live and how to get them out.

How to do the first pass in one sitting

The order you read a package in decides what you find, and the natural order, drawings first, then the section about your product, then whatever is left, is the order that misses the most. This is the order that works, and it fits in one sitting for anything under a few hundred pages.

Start at the bid form and the agreement, not the specification. The bid form tells you whether exceptions are allowed at all, which decides whether bucket two exists for this job. The agreement carries the damages, the bond, the validity and the delivery date, and those four rows are usually the largest brackets in the register and the fastest to find. Then the amendments and the answered questions, read before the original text, so that you read the original already knowing what changed. Then the reference list, every standard it cites, with a mark against each one you do not know by heart. Then the section for your product, slowly, with the reference list beside you, which is how the nuclear standard gets caught on the first read rather than at the pre construction meeting. Then the sections around yours: the electrical section that specifies your motor's starter, the civil section that fixes the footprint, the testing section that makes you supply the weights. Then the general conditions, quickly, for anything that names the contractor's obligations. The drawings last, because by then you know what the words asked for and can see where the drawings disagree.

One reader, one pass, one register. If the package is the 668 page kind, the pass is a day and the register is the reason the day was worth spending. If it is the 27 page kind, the pass is an hour, and the sentence on page 9 is still there.

The register: columns and eight rows

The register is a table, and its value is in the columns, not in the software that holds it. A spreadsheet is fine. What matters is that every row carries the same six things, so that two people reading the same package would produce registers that can be compared.

The page and the sentence, quoted, not paraphrased, because paraphrase is where the meaning drifts. The class, from the seven above. The exposure, as a bracket in dollars, wide if it has to be, never blank. Who decides, which is a name, not a department. The action: price, exclude, clarify or walk. And the assumption you will write on the quote if you bid without an answer, so that the customer sees it and so that the estimator who builds the job can find it later.

Here is what a register looks like, built from the public packages above. The exposure brackets are illustrative; the sentences and pages are real.

Notice what the register does that a matrix cannot. It tells the person building the cost exactly which lines carry a decision somebody else has already made, and it tells the person signing the bid exactly which sentences they are signing against. When the job is won, it tells the project manager where the estimator thought the trouble would be. That last use is the one nobody plans for and the one that pays.

Three buckets, not a matrix

The standard risk matrix, likelihood one to five against impact one to five, was built for project risk, where the events are in the future and the numbers are guesses. Pre bid risk is different. The sentence is already on the page. It is not likely or unlikely; it is there. What is uncertain is what it will cost you and whether you can move it. So the register uses three buckets instead of a score, and the rule for each is simple.

Price it. The exposure has a bracket you believe, the customer will not remove the clause, and the bracket is a small share of the job. The foam plumbing is priced. The test weights are priced. The realistic duration is priced. The bracket goes into the number as its own line, not smeared into a percentage, so that it can be seen and argued with.

Exclude it. The bid form allows exceptions or assumptions, and a written exclusion costs you less in win probability than the bracket costs you in margin. The foam system is excluded by name. The unnamed milestone rate is excepted. The nuclear standard, if unanswered, is excluded by name with the standard you priced to stated in its place. An exclusion the customer accepts is worth more than a clarification they never answer.

Walk. The bracket is a large share of the job, the customer will not move, and the form does not allow exceptions. A 120 day delivery on a machine with a year's lead time, no exceptions permitted, with damages attached, is a walk unless the chassis is on your floor. Walking early is the most valuable output a register can produce, because a day of reading is cheap and a month of costing is not. The fixed price contract risk piece has thirty two public companies explaining to their investors what happens when this call goes the other way.

The bid no bid checklist for custom manufacturers

If you want the bid or no bid decision to be a decision, ask these ten questions in this order, and notice that seven of them are about the document rather than about the market. They take one reader and one pass through the package, and they produce the register as a side effect.

1. Have we read the whole thing, or the cover letter and the drawings? Until the answer is the whole thing, every later answer is a guess. On the wastewater book, the requirements were in 14 of the sections and the cover letter mentioned none of them.
2. Is there a sentence that changes what the product is? A standard, a certification, a material clause. The nuclear standard on page 9 of the crane specification is the archetype. Find it or confirm it is absent.
3. Does the document disagree with itself, and have we sent the question? If a contradiction is found and no clarification goes out, you have chosen to price both readings and charge for one.
4. What is silent that the product needs? Anything plumbed for, wired for, referred to or implied and never specified. Foam systems. Test weights. A rope diameter nobody has measured.
5. Can we meet the commercial terms, and what do they cost? Delivery against lead time, validity against material volatility, damages against the realistic schedule, bond against the bid including alternates. Each as a bracket.
6. What did the amendments change, and what should they have changed? Read the amendment against every section it touches and every section that depends on the sections it touches.
7. Which named items can we supply, and which will we have to prove equal? Sixty named brands with no exceptions is a supplier list, not a specification, and it may not be yours.
8. How many numbers in our draft cost have a written source? This is the seventh class, and it is the only question on the list about you. The share is usually lower than the owner expects.
9. Who decides each open row, and are they available before the deadline? A register with owner decisions on it is useless in a week the owner is travelling. The fire district gave bidders about three weeks.
10. Given all of that, do we want this job? Now the market questions. Competition, relationship, capacity, strategic value. They matter, and they come last, because they only matter for a job you can actually deliver at a price you can actually name.

How much contingency to carry

The most common question about bid risk is also the one with the least honest answer in circulation. The circulating answer is a percentage: five, ten, fifteen, depending on how the estimator feels about the customer. The honest answer is that contingency is the sum of the register's bracketed exposures, weighted by how likely each is to land, and it should be visible as lines, not hidden as a rate.

A flat percentage does two things badly. It charges the customer for risks that are not in the document, which loses bids, and it under charges for the ones that are, which loses money on the bids you win. And it is the mechanism behind a pattern anyone who has opened a public bid tabulation will recognise. On a Florida transportation letting we studied, six contractors priced the same 4,000 square foot floating dock between $110 and $300 a square foot, on the same drawings and the same specification. Some of that spread is capability. Most of it is six different feelings about risk, expressed as six different percentages, on a document only some of them had read to the end.

Two commercial rows deserve their own treatment because they are risks you can write out of the quote rather than price into it. Validity: a quote asked to hold for 180 days on a $252,000 piece of equipment was still being held to its price on day 260, because the quote did not say what happened after day 180. A validity clause with a stated rule for the day after costs nothing to write. Escalation: a materials basis date and a named index on the quote turn a steel price move from your problem into a formula. Both are register rows with the action "write it on the quote", and both are cheaper than any percentage.

The register after the bid

Most companies that build a register at all file it with the bid and never open it again. That throws away the only part of the exercise that compounds. On every job you win, the register is a list of predictions: this sentence will cost about this much, this assumption will hold, this clause will bite. The job then happens, and each prediction is either right or wrong, and almost nobody checks.

Checking is not complicated. When the actual cost comes in, put the register beside it. Which bracketed rows landed inside the bracket. Which excluded items the customer later insisted on anyway. Which walk decisions, in hindsight, you would make again. Ten won jobs, done this way, teach a company more about its own risk pricing than any amount of industry benchmarking, because the risks are the ones in its own documents, in its own market, priced by its own people. This is the feedback check from the quoting system audit, run on the register instead of on the total, and it is the instrument that turns a register from a form into a memory.

It also answers the question the auditors ask after signature. In the annual reports we read, the estimate of cost to complete each fixed price contract becomes a critical audit matter once the contract exists, tested by outside auditors against actuals. Before the contract exists, the same estimate is a spreadsheet nobody has tested. A register that is scored against actuals on won jobs is the pre signature version of that test, and it is the only one the company controls.

Where software helps, and where it does not

Everything above can be done with a printed package, a highlighter and a spreadsheet, and for a company that bids ten engineered jobs a year it should be. The register's judgment, which bucket, which bracket, whose decision, is human work and stays human work. No system should make a scope decision on your behalf, and any system that resolves a contradiction silently has done exactly that.

What software changes is the first pass. Finding every sentence that changes the price in 99 pages, or 356, or 668, and tying each to its page, is reading, and a reading engine does it in the time it takes to find a pen, on every package rather than on the ones somebody had time for. It finds the contradictions because it has both pages in front of it at once. It finds the silence because it knows what the product needs. What it hands the estimator is the raw material for the register: the list of candidate rows, cited, before any costing begins. The estimator still decides each one. On the packages above, that first pass is where the day went, and it is the day the bid or no bid feeling was formed without.

What we do, and where the audit fits

Mavlon builds reading engines for custom manufacturers, and the register is the first thing our engine's output is good for, before it prices anything. But the register does not need us, and most companies should build one by hand on their next three bids before deciding whether any software is worth having. If you would rather someone ran the instruments with you, the quoting system audit starts with your own archive and your own estimators, and one of its three honest answers is that your process is fine and here are the rules we wrote down. Bring the package you least want to read again. That is the one with the sentence on page 9.