Anonymised, illustrative composite. A mechanical subcontractor kept bidding the same mix of work it always had — until a published national release showed that mix was fighting the market, not riding it.
At a glance
A mechanical subcontractor in Manitoba, doing HVAC and plumbing rough-in on mid-rise residential and light commercial jobs, was bidding at a volume its two estimators could sustain but not improve on. Roughly two in three of its tenders were multi-unit residential, because that was the pipeline the firm had always chased.
Across two quarters the firm won 9 of 60 bids submitted — a 15% hit rate, spread thin across a segment where it was competing against a wide field on price alone. Nobody at the firm had actually checked whether that segment was where the growth in the market was.
Statistics Canada’s Daily release on investment in building construction for May 2026 (Table 34-10-0293-01, released 2026-07-20) recorded total investment edging down 0.3% nationally to $23.4 billion, with the residential sector down 0.5% moderated by a 0.2% increase in the non-residential sector; year over year, investment in building construction had grown 5.9%. The same release broke out multi-unit residential growth by province, including a $30.8 million increase in Manitoba specifically.
Two quarters after reweighting its estimating effort toward non-residential and institutional tenders, and away from the softest part of the residential mix, the firm won 14 of 55 bids submitted — a 25% hit rate, ten percentage points higher on a smaller and more deliberately chosen set of submissions.
The decisive step was not a new pricing model. It was reading a public, dated, sourced release before deciding where to point the estimating team, instead of defaulting to the segment the firm had always bid. StatCan’s Daily release carries the release date, the source table number, and the month it covers — which is what makes it usable for a decision like this rather than a stale impression of “the market.”
The firm did not stop bidding residential work; it stopped treating every segment as equally worth an estimator’s time when one national release, updated monthly, was telling it plainly which side of the market was growing and which was softening.
Hit rate rose from 15% to 25% over the following two quarters, on fewer total bids submitted — the firm bid less, but bid where the underlying demand was moving in its favour. It now checks the current month’s release before setting each quarter’s bidding targets, rather than reviewing the market once a year.
For the adjacent question of what to do with the tenders that come in anyway, see screening RFPs with AI before you commit to bidding and how to follow up on construction bids.
Had the firm kept its old segment mix, the same 55 bids at the prior 15% hit rate would have produced roughly 0.15 × 55 ≈ 8 wins, against the 14 it actually won by reweighting — six additional wins traceable to bidding where the release said demand was moving, not to bidding harder. The firm also submitted 5 fewer bids in the process (60 to 55), precisely the volume it dropped from the softening residential segment rather than continuing to chase it.
Watch for a hit rate that is flat or falling while total bid volume stays constant or rises — that pattern usually means the estimating team is treating every tender as equally worth chasing. Before adding volume or headcount to fix it, check whether the current month's StatCan release shows the segment being chased is actually growing; a released, dated, sourced number is cheap to check and expensive to ignore.
A hit rate moving from 15% to 25% is a 66.7% relative improvement (25 ÷ 15, minus one), not merely ten percentage points — a distinction that matters when comparing this firm's result against a different segment's baseline. The firm's estimating lead now applies one discipline every month before setting targets: read the current release, note the release date and table number, and compare only investment-dollar figures against investment-dollar figures — never blending a StatCan Daily dollar figure with an employment or permit-count figure from a different source, which would produce a number that looks precise but measures two different things. The same release also warns against carrying one month's snapshot into a different month's decision; the firm now re-checks the current release every quarter rather than working from memory of what an earlier release said.
A 30-minute call is enough to tell you whether AI pays for itself here.
