A buyer opens three bids for the same trade. The first maps cleanly to the bill of quantities. The second bundles half as many items into its own structure. The third is a functional description with no line items at all. All three quote a total price, and not one of those prices is comparable to the others.
Situations like this are routine in tendering. And that is exactly why one of construction’s biggest productivity levers still sits largely untouched.
Key Takeaways
- In only about 30% of tenders do bids map cleanly to the specification. In the other ~70%, bidders answer with their own structure, so the quoted totals aren’t comparable until the bids are normalized.
- Procurement is 40–70% of a construction company’s spend and one of its least-digitized processes, the largest manual bottleneck hiding in plain sight.
- A lower total often means fewer priced positions, not a cheaper bid. Comparing only the shared positions frequently shrinks the gap to a fraction of what the totals suggest.
- McKinsey puts AI’s procurement efficiency potential at 25–40%. The decision stays human; the busywork of making bids comparable no longer has to take days.
By the numbers
The context: a $22 trillion industry with a productivity problem
According to McKinsey, global construction spending will climb from $13 trillion in 2023 to roughly $22 trillion by 2040. The title of the analysis says the rest: higher productivity is “no longer optional.”
Because while other industries have grown more productive for decades, construction has stood still. Between 2000 and 2022, construction productivity rose by a meager 10% globally. The overall economy grew 50% over the same period, and manufacturing 90%.
In the U.S. the picture is even starker. Total construction spending runs at about $2.19 trillion (annualized, early 2026), yet construction labor productivity has fallen by roughly 30% since 1970, even as economy-wide productivity nearly doubled. Add to that the fact that construction is one of the least-digitized industries in the economy, and the pattern is clear: enormous spend, flat output.
The overlooked lever: tendering and procurement
Ask where the biggest productivity gains in construction are hiding, and procurement is rarely the first answer. Yet it’s one of the largest manual bottlenecks there is.
Procurement typically accounts for 40 to 70% of a construction company’s total spend. It’s the process that moves the most money and decides the most margin. And it’s also among the least digitized: much of it still runs on Excel, email, and PDF. In U.S. construction, non-optimal activity (rework, hunting for data, resolving conflicts) is estimated to cost more than $177 billion a year, with workers spending about 35% of their time, over 14 hours a week, on non-productive tasks. A large share of that waste is information that never gets properly captured or compared.
Lots of volume, lots of margin, and still done by hand. That’s the lever.
Why comparing bids is so hard
The naive assumption goes: you write a specification, bidders fill in prices, you compare. In reality, that clean path holds in only about 30% of cases. In the other roughly 70%, bidders don’t answer the question that was asked: they offer what they have.
A precast manufacturer offers precast, even when the spec calls for cast-in-place work. A supplier prices its own catalog structure, not the bill of quantities. A waste contractor uses its own material classification. The result: procurement receives bids that aren’t directly comparable. The real work isn’t comparing prices, it’s making the bids comparable in the first place.
A real (anonymized) example: a specification asks for a complete shaft “including ventilation, dirt trap, and climbing irons” as a single line item. A material supplier prices the base shaft at €1,655.90, ventilation as a surcharge (+€35.00), the dirt trap as another (+€35.00), and the climbing irons separately (+€15.50). The comparable price doesn’t exist anywhere in the bid; it has to be calculated: €1,741.40. Only then can it sit next to a competitor who quotes everything in one number.
And it gets trickier. A bidder can post a much lower total simply because it prices fewer positions, not because it is genuinely cheaper. Compare only the positions that every bidder quoted, and the gap often shrinks to a fraction of what the totals suggest. Anyone who looks only at the bottom line is comparing apples to oranges, and risks making an expensive wrong call.
What good bid evaluation actually requires
Whether it’s done in spreadsheets or software, a reliable bid evaluation has to do four things that routinely fall through the cracks in a manual process.
- Map, don’t just place side by side. Every bid item has to be matched to the right specification item, even when one item is composed of several components or split across multiple spec lines.
- Calculate comparable prices, transparently. The comparable unit price often only emerges by adding up surcharges. Every one of those calculations has to stay traceable, otherwise the analysis looks wrong the moment someone checks the original.
- Lose nothing. Everything a bidder offers on top (options, extra items, logistics costs, items from other lots) carries information. In a manual process, most of it disappears.
- Make the gaps visible. Which item does which bidder not cover? Where is an alternative that needs a technical call before its price even counts? The price comparison is only half the truth if those questions stay open.
That’s exactly why procurement ties up so much productivity: it’s skilled work that demands domain knowledge, and today it’s still done almost entirely by hand.
Where this is heading
McKinsey puts the potential of AI and automation in procurement at 25 to 40% more efficiency, with roughly half of all procurement activities considered automatable. For bid evaluation, that doesn’t mean a machine makes the award decision. It means the grind of making bids comparable (the mapping, the adding-up, the flagging of gaps) no longer has to take days. It can take minutes. The decision stays with the human; the busywork before it doesn’t.
Sources
- McKinsey, “Delivering on Construction Productivity Is No Longer Optional” (2024)
- McKinsey, “The Strategic Era of Procurement in Construction” (2023)
- McKinsey, “Transforming Procurement Functions for an AI-Driven World” (2025)
- U.S. Census Bureau, Monthly Construction Spending (2026)
- Federal Reserve Bank of Richmond, “Five Decades of Decline: U.S. Construction Sector Productivity,” Economic Brief EB 25-31 (2025)
- NBER Digest, “The Stagnation of U.S. Construction Productivity” (Feb 2025)
- PlanGrid & FMI, “Construction Disconnected” (2018)
- 30/70 rule and shaft example: internal Arctis AI analysis of real construction tenders.
Frequently Asked Questions
See it on your own bids
That’s what we’re building at Arctis: you upload the bids and get a structured comparison of prices, items, and deviations, traceable back to the source. If you evaluate bids and want to see how it works in your context, we’d be glad to show you in a short demo.
Book a demo