Pre-qualification criteria in tender evaluation: a practical guide

Most conversations about slow tendering focus on the wrong stage. People talk about technical scoring, or negotiation, or approvals. In our experience deploying evaluation systems inside large Indian buyers, the stage that quietly consumes the most calendar time is the one that is supposed to be mechanical: pre-qualification criteria screening.

PQC is a pass-or-fail gate. There is no judgement to exercise and nothing to score. And yet it routinely takes weeks, sometimes months. This piece explains why, and what actually helps.

What pre-qualification criteria are

Pre-qualification criteria are the minimum eligibility conditions a bidder must meet before anyone looks at their technical or financial proposal. They answer one question: is this organisation allowed to compete for this work at all?

Across most public and large private tenders in India, PQC clusters into four families:

• Financial capacity — average annual turnover over a stated look-back period, net worth, solvency certificates, and sometimes working-capital thresholds.
• Past experience — completed works of a similar nature, above a value threshold, within a defined window. Usually evidenced by work orders plus completion certificates.
• Technical and manpower capability — plant and equipment, qualified personnel, and in-house capacity for the scope.
• Statutory and compliance standing — GST and PAN registrations, relevant certifications such as ISO, and declarations that the bidder is not blacklisted or debarred.

The defining feature is that all of it is objective. The thresholds are written in the tender document before bids open. Nobody is exercising discretion; they are checking arithmetic and matching documents against stated rules.

Why an objective stage takes so long

The difficulty is not the decision. It is the evidence hunt.

A single serious bidder on an infrastructure tender may submit several hundred pages: audited financials for three or five years, a dozen work orders with matching completion certificates, registration copies, affidavits, power-of-attorney documents, and organisation charts. Multiply by fifteen to twenty bidders and the committee is reading three to six thousand pages to answer perhaps forty yes-or-no questions.

Three properties of that corpus make it worse than the page count suggests:

• It is scanned, not digital. Much of it arrives as photographs or scans of stamped and signed paper, at varying resolution and skew. Text selection does not work.
• The evidence is scattered. The turnover figure needed for one criterion sits on page 214 of one bidder's submission and page 61 of another's. There is no consistent ordering.
• Every finding must be defensible. A rejection at PQC stage is the most-challenged decision in procurement. Evaluators are not just finding the number; they are building a record that survives scrutiny.

Six failure modes we see repeatedly

1. Criteria written so they cannot be evaluated cleanly

Phrases like "works of a similar nature" or "adequate experience" invite interpretation at a stage designed to exclude it. If two reasonable evaluators can disagree, the criterion belongs in technical evaluation, not PQC.

2. Turnover periods that do not line up with audited years

A criterion referencing the last three financial years collides with bidders whose audits are not yet complete for the most recent one. Without an explicit rule for that case, each bidder gets handled slightly differently, which is exactly what creates challenges.

3. Work orders without matching completion evidence

Bidders frequently submit the order but not the certificate, or a certificate for a partially completed scope. Deciding whether that counts is a policy question that should be settled before evaluation opens, not negotiated bidder by bidder.

4. Currency, units and scale mismatches

Lakhs against crores, rupees against dollars for international bidders, and figures stated in different units within the same document. Simple, and a persistent source of arithmetic error under time pressure.

5. Compliance documents that have expired

A certification valid at the time of issue but expired at bid submission. Catching this requires checking dates on every document, which is precisely the sort of task that degrades as human attention fatigues on page 2,800.

6. Inconsistent treatment across bidders

The most serious one. When evaluation runs over several weeks, the standard applied in week one drifts from the standard applied in week four. This is rarely deliberate and almost always indefensible when challenged.

What automation should and should not do

The useful split is narrower than most vendors claim. Automation is genuinely good at the evidence hunt and genuinely inappropriate for the determination.

Well suited to automation: locating candidate evidence across thousands of pages, reading scanned and skewed documents, normalising figures into consistent units, cross-referencing work orders against completion certificates, flagging date expiries, and applying the same numeric threshold identically to every bidder.

Not suited to automation: deciding whether a bidder qualifies. That determination carries legal accountability, and accountability cannot be delegated to a system that cannot be summoned to explain itself.

This is why we treat source traceability as non-negotiable rather than as a feature. In our own RFP evaluation system, every extracted value is anchored to a bounding box on the page it came from, so a reviewer verifies by looking at the original document, not by trusting the extraction. An unverifiable answer in procurement is worse than no answer, because it looks like progress.

A checklist before you evaluate

1. Every criterion has a stated numeric or binary threshold with no interpretive language.
2. The look-back period is defined against a fixed date, with an explicit rule for incomplete audit years.
3. The required evidence for each criterion is named in the tender document, not left to bidder discretion.
4. Unit and currency conventions are stated, including the conversion basis for foreign bidders.
5. Validity is assessed as at the bid submission date, and that is written down.
6. One evaluator or system applies each criterion across all bidders, rather than each evaluator taking a subset of bidders.
7. Every finding, pass or fail, cites the page it came from.

Point six is worth dwelling on. Splitting work by bidder feels efficient and is the main structural cause of inconsistent treatment. Splitting by criterion instead — one person or process applying the turnover test to all twenty bidders — costs nothing extra and removes an entire class of challenge.

The short version

PQC is slow not because the decisions are hard but because the evidence is buried, and because every finding has to survive being questioned. Fixing it is mostly about removing interpretation from the criteria, applying each rule uniformly across bidders, and making the evidence hunt cheap without letting the machine make the call.