On a heavy-civil package you usually get between five and fifteen sub bids, and most teams handle them ad hoc: bids land in an inbox over three days, someone eyeballs the totals, the apparent low gets a phone call, and the award rationale lives in nobody's head in particular. That works until it doesn't — until a scope gap surfaces as a change order, or an owner asks why you passed on the low number and the only answer is "we just knew." A repeatable lifecycle fixes that. Treat the sub bids as a managed pipeline with five stages — intake, leveling, scoring, decision, and a documented award — and every package runs the same way regardless of who is at the desk. Subcontractor bid analysis is the discipline that turns a pile of PDFs into a defensible award; the stages below are how you run it without dropping anything.

1. Intake: log and date-stamp every bid

The first stage is the most boring and the one teams skip, which is exactly why it bites them. Every bid gets logged the moment it arrives: bidder name, scope package, the file, and a date-and-time stamp. The stamp matters because subs revise. A sub who emails a corrected number at 4:52 on bid day has created two versions of their bid, and if you do not record which one is current you will level the wrong one. So intake is not "collect the PDFs" — it is establishing, for each bidder, exactly one authoritative bid and the timestamp that proves it.

By hand, intake breaks down at the version problem. A shared folder of files named Bid_Final.pdf, Bid_Final_v2.pdf, and Bid_REAL_final.pdf is not a log; it is a guess — and nobody writes down the time, so when two numbers conflict three weeks later there is no record of which arrived last. The fix is a single intake list filled in before you do anything else.

2. Track the field and the no-quotes

Once bids are logged, the next stage is knowing your field — which means tracking who didn't bid as deliberately as who did. You invited nine subs for the structural concrete package and five quoted; the four no-quotes are information. A sub who routinely bids your work and declined may have seen something in the plans — a sequencing problem, a quantity that does not pencil, a spec they can't meet. Two no-quotes for the same stated reason is a signal worth reading before you award.

The thinness of the field also changes how much weight any single number deserves. Five real bidders give you a defensible peer group; two behind a wall of no-quotes give you almost no benchmark, and the "low" number deserves far more suspicion. Manually, no-quotes are invisible — they never generate a PDF, so they never enter the folder. Record the invite list and the declines on purpose, or the field looks deeper than it is.

3. Level every bid to your scope

Now the real work. Leveling reconciles differently-formatted bids so they measure the same scope. No two subs format alike: one itemizes forty lines, another sends a one-page lump sum, a third bundles dewatering into "site prep." Before any total means anything, every bid has to be mapped onto your scope line by line, with units converted so a per-cubic-yard price and a lump sum for the same work sit in the same row.

The output you are building toward is an apples-to-apples comparison, and the heart of it is two-way scope-coverage gap analysis: for each bidder, the scope items they didn't price (gaps) and the items they carried that you didn't ask for (extras). A missing item is a future change order with that sub's name on it, and it is precisely how the apparent low bidder stops being low — the "low" total looks $150K under the next until you find it excluded traffic control and dewatering the others carried.

This is the stage that eats a precon analyst's afternoon. Normalizing one forty-line bid runs twenty to forty minutes by hand; multiply across the field and every package and leveling becomes the bottleneck that pushes awards to the last day. Worse, the gap that costs you is the one you miss while eyeballing eleven columns at 6 p.m. For a one-page version of the leveling sequence, we keep a free bid leveling checklist you can download and mark up.

4. Score the six dimensions

A leveled total is necessary but not sufficient. Price alone rarely decides a heavy-civil award: the cheapest bid with a $180K scope gap is not actually cheapest. So this stage scores each bidder across six dimensions:

  • Price — the leveled total after gaps and extras are reconciled, not the cover-sheet number.
  • Scope — how completely the bid covers your SOW; every coverage gap is priced risk.
  • Schedule — durations, sequencing, and whether the sub's plan fits the project's critical path.
  • Compliance — bonding, insurance, DBE/MWBE commitments, prequalification, required forms.
  • Performance — past results on similar work: crews, equipment, how the last few jobs went.
  • Risk — the structural warning signs in the pricing itself.

The risk dimension is where deterministic rules earn their keep, because they fire on math and catch the same thing every time. Four rules run across the leveled field: an unbalanced unit price at or below $1.00 (cost shifted off the item to game quantity overruns); a peer outlier more than 2× or less than 0.5× the peer median for that item; a total-bid outlier more than 20% off the field; and a front-loaded mobilization exceeding 10% of the total. None automatically disqualifies a bidder — a front-loaded mobilization can be legitimate on a job with heavy early equipment moves — but each forces a written reason before you move on, instead of surfacing in month three as an RFI. By hand, the soft dimensions get skipped under time pressure: the team scores price, glances at schedule, and never writes down compliance or performance, so the "score" is really just price wearing a jacket.

5. Decide under a decision mode

With six dimensions scored, the decision stage is about being explicit about how you weigh them — and committing to that weighting before you see the winner. That is what a decision mode is: a named rule for how the dimensions combine. Choosing it up front keeps the analysis honest, because it stops you reverse-engineering the weights to justify a bidder you already liked.

There are seven modes to pick from depending on the package and the job's pressures: Lowest Responsible Bid for the cheapest qualified number on well-defined scope; Best Value when quality and price both matter; Lowest Risk when a blowup would be catastrophic; Schedule Priority when the critical path runs through this package; Scope Completeness when coverage gaps are what you fear most; Budget-Constrained when you have a hard ceiling; and Custom Weighting when none of the presets fits. The point is not that one mode is right — it is that you name it, the same mode produces the same ranking from the same inputs, and the choice is on the record. Manually, this stage barely exists: most teams never declare a mode at all, which is why two analysts looking at the same bids reach different recommendations and neither can say why.

6. Document the award

The last stage is the one that protects you. Six months into a $40M job, the question is "why did we pass on the low bidder?" — and a spreadsheet does not remember. The award stage produces a written record while the reasoning is fresh: the leveled tab, the scope-coverage gaps per bidder, the flags that fired and why each was accepted or dismissed, the decision mode, and the ranked recommendation that followed. If a human overrides the recommendation — and sometimes judgment should overrule the math — the override is logged with the reason, so the record shows not just what you decided but that you decided it deliberately.

The evidence has to be traceable, which means page-cited quotes: when the memo says a bidder excluded dewatering, it points to the page of their bid where the exclusion appears — the difference between a recommendation an owner trusts and an assertion they take on faith. By hand, this is where it all leaks away: the reasoning lived in cell comments and a few emails, and reconstructing it under a dispute is guesswork. A documented award is not bureaucracy; it is what lets you defend the decision to an owner, a partner, or a dispute panel without re-doing the analysis from memory.

Running the lifecycle without the afternoon

Bid Reasoner runs all five stages on the sub bids you receive. It reads each bid PDF, normalizes the line items, and maps them to your scope via peer-median normalization, which works in any US state without government data. (For a head start, built-in state-DOT pay-item baselines exist for select states — currently NY and NJ — but they are never a requirement.) It runs the four risk rules deterministically, produces the two-way scope-coverage gap analysis, and scores all six dimensions. You pick a decision mode, it returns a ranked recommendation with the page-cited evidence behind every flag, and any override is logged with your reason. The award stage is seven generated Word documents — leveled tab, award memo, evidence report, and the rest — ready to hand to the owner. It complements HCSS HeavyBid and B2W rather than replacing them. If your process is a folder of PDFs and a tired spreadsheet, our subcontractor bid analysis runs exactly this lifecycle.