The estimating departments that leak the most money rarely look undisciplined. They have a process, a bid board, a Monday meeting. The leaks live in the exceptions — each one defensible in isolation, each one compounding. Here are the nine we find most often when we audit a precon desk, in rough order of annual cost.
1. Scoring after the takeoff starts
The most expensive mistake, and the most common. Once forty hours are in the quantity survey, no room full of humans scores the pursuit honestly — sunk cost adds a point to every answer, and "we've come this far" finishes the job. Fix: the scorecard is a gate, not a review. No score, no takeoff. The tool takes five minutes; there is no schedule so tight it can't afford them.
2. Blending negotiated and hard-bid stats
Negotiated work wins at 40-60%; hard-bid at 10-30% by trade. Average them and you get a comfortable number that hides a failing pipeline. One drywall client reported a "healthy 31%" — decomposed, it was 58% negotiated propping up 7% hard-bid, and the hard-bid desk was burning $360k a year. Fix: two pipelines, two hit rates, two sets of benchmarks. Always.
3. Serving coverage bids
Some invitations exist so the GC can show three numbers next to the one they've already chosen. The tells: the invitation arrives late with a short fuse, the spec reads like a competitor's catalog, and you have never once won with this GC despite "being close." Fix: track win rate by inviter in your pursuit log. Two years of zero with one GC is not a relationship — it's a subsidy. Decline fast, with a courteous letter, and say what work you would price seriously.
4. No written thresholds
"We're selective" without numbers means the loudest voice in the room is the threshold. The same pursuit gets bid in March and declined in July depending on mood. Fix: written bands — 45/32/20 by default — and written decision rights about who can override them, per the go/no-go meeting format.
5. Ignoring cash in the bid decision
Estimating optimizes for margin; nobody in the room owns the question "can our balance sheet carry this job's receivables?" Then the 8%-margin job with 10% retention and a 75-day owner becomes a liquidity crisis in month four. Fix: the AR model in the Excel workbook — peak AR over 40% of working capital is a hard kill, no matter the margin.
6. The silent no-bid
Declining by ghosting. The GC's coordinator chases, re-plans coverage, and quietly moves you to the backup tier — the one that only sees invitations when the A-list is busy. You never find out. Fix: the 48-hour no-bid letter, every time. It costs five sentences and preserves the deal flow your selectivity depends on.
7. Frozen thresholds in a moving market
The bands that were right at two months of backlog are wrong at seven. Shops that never recalibrate either starve (too picky in a drought) or choke (too loose in a boom, staffing peaks with strangers). Fix: quarterly calibration against backlog — 50/38/25 when full, 40/28/17 when thin. The scorecard tool does this automatically from your backlog input.
8. Strategic bids without a written strategy
"Strategic" is the word that launders every bad pursuit. A real strategic bid names its objective — enter the healthcare market, keep the drywall crews whole through Q1, earn the second look from a target GC — and gets audited against it. Fix: a strategic tag in the pursuit log with a one-line objective. Count them quarterly; if more than 15% of your bids are "strategic," none of them are.
9. Never mining the log
The pursuit log is where most shops' best data goes to be ignored. Score bands vs. outcomes tells you if the rubric works. Win rate by owner tells you who to invest in. Kill reasons tell you what your market is really like. Fix: two quarters of honest logging, then one hour a quarter reading it. The hit-rate-by-band stats in the workbook's pursuit log compute themselves — someone just has to look.
These leaks multiply. A shop that scores late (1), blends its stats (2), and serves two coverage bids a quarter (3) can show a plausible-looking pipeline while burning a third of its estimating budget on pursuits that were never winnable. The fixes above are cheap; every one of them is a process rule, not a hire.
The order to fix them in
Start with the gate (mistake 1) — it makes every other fix enforceable. Then split the stats (2), because you cannot manage what you've averaged away. Then the letter habit (6), because it protects deal flow while your selectivity tightens. The rest follow inside a quarter, and the compounding runs the other way: better selection frees hours, freed hours deepen the surviving bids, deeper bids raise the hit rate that justifies the selection.
PILARS accelerates the last leg: with takeoff at $100 per trade per plan, the hours you stop wasting on unwinnable pursuits go into winning the scored ones. See pricing here.