The lowest bid is frequently not the cheapest. The levelled bid is.
Last updated
Zepth Vector module
0.5%
of project value spent on every bid submitted — across £8bn of tendered work studied
Constructing Excellence (UK)
Roughly. Winners spend more than losers, and the published figures sit either side of this — we quote the order of magnitude rather than a derived multiple.
4–6
prequalified bidders — the evidence-backed sweet spot. Beyond it, good firms start declining to bid
Constructing Excellence (UK)
Tender management is the controlled process of inviting, equalising and evaluating competitive bids — from prequalification through award. Done well, it buys the right scope at a defensible price, with an audit trail.
Done by email and spreadsheet, it produces awards to incomplete bids, information-leak disputes, and variations that eat the “savings” within six months.
Bidding is expensive, and it is expensive for the people you most want to hear from. UK research across £8 billion of tendered work found bidders spend on the order of 0.5% of project value on each bid they submit — real money, spent before anyone has won anything.
Which is why an over-long tender list quietly buys you worse pricing, not better. At a one-in-five win rate, a good contractor does the arithmetic and declines. The firms that remain are the ones with spare estimating capacity, and that is not the same thing as the firms you wanted. The evidence points to a sweet spot of four to six selective bidders: procurement costs keep falling a little beyond that, but the quality of participation drops faster than the savings arrive.
The stakes on the evaluation side are larger still. Misunderstanding of contract obligations and scope change are perennial top causes of construction disputes, and the average North American dispute now runs to tens of millions and over a year to resolve. Most of those disputes trace back to something that did or did not happen during the tender: an exclusion missed in comparison, unequal information between bidders, a scope gap discovered after award.
Bid levelling is where evaluation is won or lost. Raw bid totals are close to meaningless. A proper BOQ comparison matrix normalises scope exclusions, provisional sums, alternates, blank rates — the “included elsewhere” that turns out not to be included anywhere — and quantity deviations, before anyone compares a single number. Skip it and the chain is predictable: award to the “lowest” incomplete bid, post-award variation claims, dispute. The lowest bid is frequently not the cheapest. The levelled bid is.
Abnormally low and unbalanced bids are different diseases. An abnormally low tender is a total price too low to perform: an insolvency and claims risk. The disciplined response, mirrored in EU and multilateral-development-bank rules, is to ask for an explanation first and reject only if the explanation fails. An unbalanced bid is different — a plausible total with a gamed distribution: items front-loaded to capture early cash flow, or rates inflated on items the bidder believes are under-measured. Different tests, different remedies. Conflating them is how bad awards get justified.
Addenda control is equal information. Every clarification answered privately to one bidder is simultaneously a protest vector and a mispricing machine. The discipline: vendor questions anonymised, answered by broadcast as numbered addenda, acknowledged in the bid form, with a Q&A cutoff before submission. And late is late. The strictness of late-bid rules exists to protect the integrity of the process rather than to be cruel — and timestamped electronic submission is what makes the rule enforceable and auditable rather than merely stated.
Bid validity is a clock, and it is running. The ninety-day convention only works if evaluation and approvals actually fit inside it. In volatile markets an expired validity means bidders reprice or walk away, and you have spent the tender period to arrive back where you started. Set validity against a realistic evaluation programme rather than against habit.
Appointing a preferred contractor early — on qualitative criteria and limited pricing — buys real things: buildability input while the design can still absorb it, and an earlier start on site.
It also has a known failure mode, and it is worth naming rather than glossing: stage-two price creep, once the competitive tension is gone. The contractor you appointed on quality is now the only contractor you have, and both of you know it.
The mitigations are unglamorous and they work: an open-book stage two, rates benchmarked against the market, and a retained right to re-tender that you are genuinely willing to use. Two-stage tendering is a good tool. It is not a free one.
Twenty quotes arrive as inconsistent PDFs. Someone re-keys them into a comparison sheet under deadline pressure. An exclusion is missed, and it becomes a change order at three times the money it would have cost to price at tender.
An addendum reaches seven of the eight bidders, and the eighth has a claim. Nobody can prove who saw what, or when.
The process ends where it was always going to end: with an award nobody can defend, and a comparison nobody can reconstruct.
The full lifecycle lives in one workflow. Tender packages are built with hierarchical BOQs. Vendors are invited against defined submission requirements. Q&A threads run centrally with broadcast answers, and addenda are issued and acknowledged on the record. Submissions stay sealed until the deadline.
Comparison matrices are generated from the vendors’ own line-item pricing — not re-keyed from PDFs by someone at eleven at night. And award converts directly into the purchase order and contract, so the audit trail from enquiry to commitment is continuous, and nothing drifts between what was bid and what was bought.
Bids are compared at BOQ line level, so the comparison is real — exclusions priced in, blank rates filled, provisional sums aligned.
Abnormally low and unbalanced bids are screened separately, because they are different problems with different remedies.
Every bidder had the same information, and you can prove it — anonymised questions, broadcast addenda, acknowledgements on the record.
The award converts straight into the contract and purchase order, so nothing drifts between what was bid and what was bought.
Tender packages built with structured bills of quantities, so vendors price against the same line items rather than their own interpretation of them.
Generated from the vendors’ own pricing — not re-keyed from twenty inconsistent PDFs by someone under deadline pressure.
Questions anonymised, answers issued to every bidder, acknowledgement captured. Equal information, provable.
Bids stay sealed until the deadline. The timestamp is what makes “late is late” enforceable and auditable.
Abnormally low totals and gamed distributions checked separately against the estimate — different tests, because they are different diseases.
The award converts directly, so the trail from enquiry to commitment is continuous and nothing is re-typed.
Capacity, financials, licences, HSE history — narrowed to four to six genuine candidates. A longer list is not more competition; it is less.
Drawings, specifications, BOQ and submission requirements. Two-envelope — technical and commercial separated — where quality genuinely carries weight in the award.
Questions anonymised, answers broadcast to every bidder as numbered addenda, acknowledgement captured, deadline extensions handled formally. A private answer to one bidder is a dispute you have not had yet.
Submissions stay sealed until the deadline, and late is late. A timestamp is what makes that rule defensible instead of merely stated.
Technical scoring where the award is weighted, then commercial levelling at BOQ line level, with abnormally-low and unbalanced-bid checks against the estimate.
The recommendation is documented, and the award converts directly into the purchase order and contract — no re-keying, no drift between what was bid and what was bought.
Every vendor’s BOQ compared item by item — pricing anomalies, blank rates, exclusions and front-loading patterns flagged, and the evaluation recommendation drafted. Weeks of levelling become a day, and the committee still makes the call.
Bids screened against the estimate and against each other, with the specific line items driving the anomaly cited. A flag without the items behind it is just an accusation.
Duplicate vendor questions detected and answered consistently, with draft broadcast responses prepared for the tender manager. Consistency across answers is itself an equal-information control.
Negotiation opportunities and risk flags per bidder, with estimated impact — evidence supporting the award recommendation, not a substitute for making it.
The engineer’s judgment stays in charge; the AI removes the latency and the blind spots.
Tender dashboards with the line-item comparison matrix, technical and commercial scores, compliance flags, and the addenda and acknowledgement trail. Abnormally-low and unbalanced-bid screens report the items behind the flag, not just the flag. Exportable for commercial governance, audit, and — if it comes to it — the challenge.
Prequalification, invitation to tender, the clarification period with its Q&A and addenda, submission, technical then commercial evaluation, and award into contract. The discipline in the middle stages is what determines whether the award survives scrutiny — the ends are the easy part.
Normalising every bid to a common scope before comparison — pricing in exclusions, aligning provisional sums, filling blank rates, reconciling quantity deviations — so that the comparison is between prices rather than between opinions about scope. The raw lowest number is frequently not the cheapest offer.
Read the full answerBest practice, and the rule under EU and multilateral-development-bank procurement regimes, is no: request an explanation first, and reject only if it does not stand up. Automatic rejection invites a challenge. Accepting without scrutiny invites a default halfway through the project, which is considerably worse.
Read the full answerSingle-stage means full competition on complete documents, which gives maximum price tension. Two-stage appoints a contractor early on qualitative criteria and firms the price up collaboratively — better buildability and programme, with stage-two price creep as the risk you have to manage once the competition is gone.
Centrally, anonymised, and broadcast to all bidders as numbered addenda that bidders acknowledge in their bid form. Any private answer to one bidder is unequal information: a protest vector, and a guarantee that one of your bidders is pricing a different job from the rest.
Four to six prequalified bidders is the evidence-backed range. Beyond that, the cost of bidding — on the order of 0.5% of project value per bid — drives good firms to decline, and the quality of your pricing falls even as the number of bids rises.
Read the full answerRelated modules
Related answers
In the guides
Terms defined here
Zepth is the construction project delivery platform — it runs construction, procurement and asset management on one record, and does the work: reading the drawings, reviewing the submittals, matching the invoices and flagging the risks, with a human sign-off on anything consequential.
A short, tailored walkthrough on your real workflow — no generic demo.