A lot-level relevance shortlist is one row per official lot, tied to the procedure and notice version from which it came. Each row preserves the lot identifier, scope, place, term, classifications, criteria, documents and exact capability evidence. It also carries procedure-wide limits, group rules and missing information that affect interpretation. The row ends in one of five states: direct match, conditional match, adjacent, no match or unresolved. It answers where a narrow capability appears. It does not decide whether the supplier should bid, can satisfy every condition or can deliver the whole contract.
A broad notice can hide a small, precise opportunity. A national facilities contract may contain one access-control lot. A digital transformation programme may contain a data-migration lot. Yet the reverse error is just as common: a search result mentions the capability in the procedure description, while no lot actually buys it as a separable deliverable. Titles and classification codes compress meaning. Attachments hold boundaries that search indexes omit. Some rules sit on the lot, some on a group and others on the procedure. If a researcher collapses those layers, a good specialist lot disappears into a broad record or an attractive phrase becomes a false lead.
Resolve the procurement hierarchy before judging relevance. Enumerate every official lot, retain its stable identifier and read its own scope before comparing it with a written capability statement. Then apply shared conditions from the procedure and any group of lots. Treat a title or code as a route to evidence, not as proof. A direct match requires explicit support for the deliverable and no known scope blocker. A conditional match names the unresolved dependency. If the source is incomplete or contradictory, the honest state is unresolved. An agent can extract, compare and cite the evidence, but should return the lot documents for qualification rather than turn lexical similarity into a bid recommendation.
Capability
Turn the specialist offer into a test, not a keyword
Start with one narrow capability. “Cybersecurity” is a market category, not a useful test. “Perform annual penetration testing of externally facing web applications and deliver retest evidence in the United Kingdom” is testable. Record the deliverable, the asset or process affected, the buyer outcome, delivery location, scale boundary and proof the supplier can show. Add what the supplier does not provide. If it can test applications but cannot run a 24-hour security operations centre, that exclusion is as important as the positive description.
Create a small evidence vocabulary from the capability, including buyer phrases, standards, credible classification codes and concrete outputs. Use it to locate passages, not to score fit on word frequency. “Testing” in a risk-management lot may describe an incidental control. “Security operations” in a procedure title may cover several lots, only one of which contains penetration testing. The decisive question is whether the buyer purchases the deliverable inside this lot and whether the supplier can assume the complete boundary assigned to that lot.
| Field | Example value | How it is tested |
|---|---|---|
| Deliverable | Web application penetration test and retest | Explicitly required in scope or schedule |
| Object | Externally facing production applications | Asset class appears in the lot documents |
| Outcome | Prioritised findings and closure evidence | Outputs and acceptance terms align |
| Boundary | Assessment, reporting and one retest | No inseparable managed-operation duty |
| Geography | Remote delivery with UK workshop access | Performance place and attendance fit |
| Proof | Named testing method and qualified personnel | Required evidence can be supplied |
| Exclusion | No continuous monitoring service | Reject bundled operations scope |
Hierarchy
Map the procedure before interpreting an individual lot
Open the official notice and record its procedure identifier, notice identifier, publication date, version and status. Then enumerate the objects that the source calls lots. TED eForms assigns technical identifiers such as LOT-0001 and keeps the same lot identifier through the procedure. It also represents a procedure without commercial lots using LOT-0000. Preserve that identifier for joins, but describe the commercial reality accurately: “procedure not divided into lots, represented by technical lot LOT-0000.” Do not invent “Lot 0” as a choice available to bidders.
Keep groups and parts distinct. A group of lots combines named lots for shared information such as group-specific award criteria or value. A part can appear in planning information before the final procurement structure exists. Neither is interchangeable with a tenderable lot. Build the hierarchy from explicit source relationships. If the rendered portal and machine-readable notice disagree, retain both observations, label the conflict and check the authoritative documents or a later corrected notice before using the row.
- Retain official identifiers exactly, including leading zeroes.
- Record the source version on which the lot map depends.
- Link each lot to its group only when the source states the membership.
- Describe a technical single lot separately from commercial lot division.
- Do not promote a planning part into an open tender lot.
Evidence
Read lot scope beside group and procedure rules
For every lot, capture its title, description, main and additional classifications, place of performance, planned term, estimated value wording, options, selection criteria, award criteria, execution requirements, milestones, document links and communication route where published. Read the attached specification, schedule and pricing file when the notice points to them. A portal card is useful for discovery, but it rarely defines the complete work package. Save a page, section, field or stable fragment for every decisive statement so another person or agent can inspect it.
Next attach inherited conditions without pretending they were repeated at lot level. Procedure-wide rules can say how many lots a supplier may tender for or win, whether a tender must cover all lots, and which combinations the buyer may award. A group can have its own criteria. Those rules do not erase the lot-specific evidence, but they can change the next action. A supplier that matches LOT-0003 directly may still face an all-lots requirement or an award combination that bundles LOT-0003 with work outside its boundary. Mark the lot match and the procedural blocker as separate facts.
| Layer | Typical facts | Failure when omitted |
|---|---|---|
| Procedure | Identity, overall purpose, maximum lots, all-lots rule | A relevant lot appears independently tenderable when it is not |
| Group of lots | Membership, group criteria, combined value | Shared evaluation logic disappears |
| Lot | Scope, place, term, criteria, requirements, documents | The headline substitutes for the purchased work |
| Document | Detailed tasks, interfaces, volumes, pricing lines | A partial capability looks like whole-lot coverage |
| Source history | Version, correction, last check, status | An obsolete structure is treated as current |
Judgement
Name the match, the blocker and the unknown separately
Use a small set of states with written thresholds. “Direct match” means the lot explicitly purchases the capability, the supplier covers the known delivery boundary and no known fact contradicts the match. It does not mean eligible or worth bidding. “Conditional match” means the core deliverable fits but a named dependency remains, such as a required location, missing schedule, consortium need or uncertain whole-lot obligation. “Adjacent” means the capability could support the work but is not itself the lot outcome. “No match” requires a cited conflict. “Unresolved” means the available official evidence cannot support a responsible judgement.
Keep positive and negative evidence in separate columns. A code, title phrase and one matching task may support further reading. A mandatory bundled service, incompatible location or whole-lot obligation may block fit. Do not average them into a numeric similarity score that hides the reason. When two documents conflict, cite both and let the controlling-document rules or an official clarification settle the issue. The agent should not choose the version that creates the more attractive opportunity.
| State | Minimum evidence | Next action |
|---|---|---|
| Direct match | Explicit deliverable, covered boundary, no known blocker | Retrieve the full pack and qualify |
| Conditional match | Core fit plus a named unresolved dependency | Resolve the dependency |
| Adjacent | Useful supporting role, not the purchased lot outcome | Consider partner route or monitor |
| No match | Cited scope, location or obligation conflict | Exclude with reason |
| Unresolved | Missing, inaccessible or contradictory evidence | Obtain the source or clarification |
Work product
Return a shortlist that another agent can challenge
Create one row for every official lot, not only the promising ones. Include procedure and notice IDs, source version, lot ID and group, title, concise scope, place, term, value basis, key criteria, document anchors, capability evidence, blocker, unknowns, relevance state, last checked time and next action. Keeping rejected rows prevents the same lot from being rediscovered and shows that the procedure was fully inspected. If there are many lots, publish the full evidence table and a filtered view rather than deleting the negative cases.
Rank direct and conditional matches only after state assignment, using declared factors such as evidence completeness or closeness of the purchased deliverable. Never rank by commercial attractiveness unless a later qualification step has the necessary facts. Preserve official wording in the evidence field and keep the analyst interpretation beside it. A safe handoff says, for example, “LOT-0003 is a conditional capability match; the pricing schedule is unavailable, so whole-lot coverage is unresolved.” It does not say “best tender to bid.”
- Produce exactly one row for each official lot.
- Expose source fact, interpretation and uncertainty in separate fields.
- Use canonical notice and document links, not search-result URLs.
- Keep the last-checked timestamp and notice version together.
- Route shortlisted lots into qualification as a new decision stage.
Agent contract
Give an agent explicit rules for evidence and abstention
An agent is well suited to enumerate lot objects, normalize identifiers, follow public document links, locate capability phrases and assemble comparable fields. Require it to retain the source for every extraction and to distinguish absent from not yet checked. Its response contract should reject a lot-level conclusion when the lot map is incomplete, the latest notice is uncertain, decisive documents are inaccessible or the capability boundary has not been supplied. Those are input failures, not low scores.
The agent must never copy procedure-level text into every lot as though the buyer published it there. It should label inherited rules and preserve their origin. It should not infer that silence means a supplier may bid for unlimited lots, that every lot shares a deadline, or that a group is an award bundle. It may report what a combination rule says but should not predict the buyer’s eventual award. The useful answer is compact and inspectable: candidate lot, exact reason, conflicting fact, missing evidence, source and safe next step.
- Abstain when the official lot inventory cannot be completed.
- Never fill an absent criterion, deadline or location from a sibling lot.
- Preserve inherited conditions with their procedure or group provenance.
- Return contradictions instead of selecting the convenient source.
- Keep discovery, qualification and bid decisions as separate operations.
What good looks like
Useful outcomes from find relevant tender lots
- Every candidate points to one procedure, notice version and official lot identifier.
- Lot-specific facts remain separate from procedure-wide and group-wide conditions.
- The supplier capability is expressed as testable deliverables, boundaries and proof.
- Direct, conditional, adjacent, rejected and unresolved matches are distinguishable.
- Each judgement cites the words or field that support it and records the last check.
- Missing documents and conflicting descriptions stay visible instead of being guessed.
- The shortlist routes promising lots into full qualification without claiming bid readiness.
Operating model
How to run the work
- 01
Define the narrow capability
Write the deliverable, buyer outcome, delivery boundary, geography, mandatory proof and explicit exclusions in terms that can be tested against procurement evidence.
- 02
Resolve the official hierarchy
Capture the procedure and notice identifiers, publication version, every lot ID, any group membership and whether an apparent single lot is only a technical representation.
- 03
Read every lot in context
Inspect each lot description, classifications, location, term, criteria, requirements and linked documents, then bring down the shared procedure and group rules that still apply.
- 04
Test fit and blockers separately
Record the exact capability evidence, whole-lot obligations, exclusions, dependencies and unknowns before assigning a controlled relevance state.
- 05
Return an inspectable shortlist
Rank only by declared relevance and evidence quality, attach source anchors and dates, and send direct or conditional matches to document retrieval and bid qualification.
Evaluation
Questions that change the decision
- Is the record divided into commercial lots, parts used for planning, or one technical lot?
- Which official lot contains the capability as a purchased deliverable rather than background language?
- Does the supplier cover the whole lot or only one inseparable component?
- Which conditions belong to this lot, its group or the procedure as a whole?
- Can suppliers tender for this lot alone, and are award or combination limits stated?
- Do the notice and procurement documents agree about scope and identifiers?
- What fact makes a match direct, conditional, adjacent, rejected or unresolved?
- Which document or official clarification is needed before qualification can begin?
Failure modes
Where teams lose control
A broad procedure title may hide the only relevant lot.
A promising lot title may omit bundled services that the specialist cannot deliver.
A capability mentioned in background text may not be an awarded deliverable.
Procedure-wide selection rules may disqualify an otherwise relevant lot.
Group award criteria or permitted combinations may change the meaning of a single-lot match.
A portal view may omit a lot attachment or display an older notice version.
A planning part may be mistaken for a lot that suppliers can tender for now.
The technical identifier LOT-0000 may be described as commercial lotting when the procedure is not divided.
Search ranking may reward repeated keywords rather than complete scope fit.
An agent may silently fill missing eligibility, geography or document facts.
Measurement
Measure the finished job
Measure the completed workflow, including review effort and exceptions. Output volume on its own is not evidence of a better process.
- procedures with every official lot enumerated
- candidate rows with stable lot and notice identifiers
- rows with lot-specific and shared conditions separated
- matches supported by a quoted field or document anchor
- conditional matches with a named dependency and owner
- unresolved rows preserved rather than forced into yes or no
- promising lots later rejected for hidden whole-lot scope
- shortlisted lots traceable to the latest official source version
Questions
Common questions
Is a matching lot title enough to put the tender on our shortlist?
No. Read the lot description and documents, then apply procedure-wide and group rules. The title is a locator, not proof of the complete work package.
What does LOT-0000 mean in a TED eForms notice?
TED represents a procedure without commercial lots using one technical lot identified as LOT-0000. Preserve the identifier, but do not describe it as a supplier choice among lots.
Can a supplier bid for only the matching lot?
Only the notice and tender documents can answer that for the procedure. Check maximum lots allowed, any all-lots requirement, group rules and permitted combinations.
Should classification codes decide lot relevance?
No. They help locate and compare records, but scope, deliverables, requirements and documents decide whether the capability is actually purchased in the lot.
Can an agent recommend the best lot to bid?
This shortlist can identify evidence-backed capability matches. A bid recommendation requires a separate review of eligibility, economics, capacity, competition, risk and strategic value.
Sources
Primary references
- Guidance: Lots UK Cabinet Office
- Procedure, Group of Lots, Lot and Part related information Publications Office of the European Union
- Directive 2014/24/EU, Article 46, consolidated text EUR-Lex
About the author
Tony Kim
Tony writes about applied AI, dependable product engineering and the systems that turn complex response work into controlled delivery.
Zelius
Managed tender intelligence and bid execution for teams that want the commercial outcome.
Suppliers, founders and commercial teams pursuing public or private opportunities. Start with the workflow, constraints and evidence you already have.
