A lot-selection decision chooses the exact set of lots for which one bidder configuration will prepare tenders in one procurement. It starts with the complete official lot universe and the buyer’s current rules for submissions, awards, groups, combinations and shared obligations. It then imports each lot’s eligibility, capability, delivery, contract, economics and response-effort findings. The work product is a lot_bid_portfolio with one row per lot, a register of buyer rules, every material award set the supplier could receive, shared and incremental costs, delivery exposure, exclusions, authority, expiry and a controlled final state. The decision does not reinterpret lot scope, set prices, reserve people, approve exceptions or authorize submission.
Multi-lot procurements invite a misleading shortcut: score each lot, sort the scores and bid the top few. That can select an impossible portfolio. A supplier may be allowed to tender for three lots but be able to deliver only one of the pairs the buyer could award. One attractive lot may depend on a shared depot, transition team or certification whose cost has been counted nowhere. Two weak stand-alone lots may become viable together because a permitted combination removes duplicated setup work. The reverse also happens when a volume discount, common manager or reused response is assumed without support in the documents. A neat ranking conceals these interactions. The decision must survive the buyer’s allocation rules and every award outcome created by the chosen bids.
Select sets, not isolated rows. First remove any lot with an unresolved hard gate. Then construct only the combinations the buyer permits and the supplier could be awarded. Test each combination against peak delivery demand, aggregate qualification, contract exposure, remaining bid effort and contribution range without double-counting shared costs. A cap on awards does not automatically let the bidder choose which wins to keep; only the published allocation method or an accepted preference mechanism can control that outcome. Keep a complete rejected-lot record so an agent cannot silently narrow the universe. The final recommendation names one approved set, the other feasible sets considered, the facts that would change it and the next action that remains within authority.
Buyer rules
The buyer defines the possible portfolios
A bidder cannot choose a lot set responsibly from the lot descriptions alone. The current tender notice and documents may limit the number of lots tendered, impose a separate cap on awards, require all lots, define groups, permit combined contracts or prescribe how the buyer chooses between excess wins. Those rules create the set of outcomes the supplier must be ready to accept. Preserve the source wording and scope. A number shown at procedure level should not be copied into a lot row as though it were a lot condition.
The UK Cabinet Office’s current lots guidance distinguishes the maximum number a supplier may tender from the maximum it may be awarded. It also says the buyer must publish the criteria used when a supplier would otherwise win above the award cap and disclose an option to combine lots. Article 46 of Directive 2014/24/EU contains the same structural distinctions for procurements in its scope. French Code de la commande publique article R2113-1 requires the consultation documents to state whether suppliers may bid for one, several or all lots and, where relevant, the allocation rule above an award maximum. These sources show why a generic instruction such as “pick your best three” is unsafe.
TED eForms exposes the procedure-level fields behind part of this logic: BT-31 for the maximum number of lots submitted, BT-33 for the maximum awarded, BT-763 where tenders must cover all lots, and identifiers for groups of lots. Structured fields are useful entry points. The tender documents may still carry preference mechanics, aggregate qualification tests, discount rules and consequences that the notice summary omits. The record therefore links each rule to its authoritative location, version and observation time.
| Rule | Evidence to preserve | Selection consequence |
|---|---|---|
| Lot universe | Official lot IDs, groups and current version | Defines every row that must be decided |
| Tender limit | Maximum lots submitted and any all-lots rule | Constrains candidate bid sets |
| Award limit | Maximum lots awarded to one supplier | Constrains but does not choose the winning set |
| Excess-win method | Published objective allocation criteria | Determines which awards survive the cap |
| Combination option | Named lots or groups that may be combined | Changes contract and delivery exposure |
| Bidder preference | Permitted form and stated legal effect | Controls outcomes only to the extent published |
| Shared condition | Procedure or group clause and affected lots | Adds a gate or cost to each covered set |
Hard gates
Remove failed lots before comparing value
A portfolio model is not a substitute for lot qualification. Import the accepted findings that already answer whether the lot is open, relevant, eligible, contractually tolerable and deliverable. Keep the evidence version beside the import. A red condition of participation, an expired deadline or a forbidden deviation removes the lot. No commercial weighting should offset it. An unresolved material fact produces a conditional candidate, not a discounted score.
The scope level matters. One insurance condition may apply to the whole procedure, a turnover threshold may aggregate across a proposed combination, and a specialist credential may belong only to one lot. Apply each condition at its stated level. Do not replicate a procedure-wide pass into every row and then count it as independent proof. Equally, a failure on Lot 4 should not eliminate Lot 2 unless the rule links them.
Give each row a reason that another reviewer can challenge. `gate_passed` means the necessary inputs are accepted for portfolio work. `conditional_candidate` names the missing fact and its decision date. `hard_gate_failed` cites the requirement and failure. `portfolio_excluded` is different: the lot is individually viable but omitted because another permitted set uses resources or budget better. `authority_missing` halts the affected comparison.
| Field | Question answered | Failure treatment |
|---|---|---|
| Current status and deadline | Can a tender still be made? | Stop the lot if closed or unresolved at the gate |
| Participation and eligibility | May this bidder configuration compete? | A failed mandatory condition excludes the lot |
| Whole-lot capability | Can the complete stated scope be covered? | Route a precise gap to the partner decision |
| Contract boundary | Is a known term outside authority or tolerance? | Escalate or exclude without hiding the term |
| Delivery fit | Can one award be performed with evidenced resources? | Keep conditional supply out of confirmed capacity |
| Input currency | Are the imported records still valid? | Recheck before using stale evidence |
Set logic
Test the awkward win, not only the preferred win
Suppose a supplier tenders for Lots A, B and C and the buyer may award any two. The relevant exposure is not simply “up to two lots.” Unless the published mechanism removes one pair, the supplier should test A+B, A+C and B+C as well as each single award. If B and C need the same installation crew during the same fortnight, the three-lot bid set is unsafe even when A+B is the team’s preferred result.
Build award sets from the buyer’s mechanism. Respect mandatory groups, award caps, combination options, ranking rules and any bidder preference the buyer undertakes to follow. Do not invent a right to refuse one lot after award. A choice that depends on post-award withdrawal is a contract and procurement risk, not a capacity treatment. If the rules are unclear, preserve the conflict and seek authorized clarification before the set becomes the working plan.
Large procedures can create many mathematical subsets. The record need not print combinations that have no material difference. It must cover every distinct resource, qualification, commercial or contractual exposure. Group equivalent sets only after stating the equivalence rule, such as identical schedules and separate crews. Keep the full machine-readable enumeration behind any summarized view so an agent can show what it excluded.
| Proposed bids | Possible award | Evidence required | Disposition |
|---|---|---|---|
| A, B, C | A only, B only or C only | Single-lot qualification and delivery | Test each |
| A, B, C | A+B | Combined peak, aggregate tests and shared terms | Test |
| A, B, C | A+C | Combined peak, aggregate tests and shared terms | Test |
| A, B, C | B+C | Combined peak, aggregate tests and shared terms | Blocking if one crew is double-booked |
| A, B, C | A+B+C | Published award cap | Excluded only when the cap is verified |
Delivery exposure
A set passes only when its possible awards can be delivered
Import the accepted delivery record for each lot, then overlay the lots that may be awarded together. The binding point is often mobilization rather than steady-state operations. Three lots may each need one service manager after month three yet all require the same implementation lead, legal reviewer and training team during the first six weeks. Annual headcount does not reveal that collision.
Use dates, places, skills, equipment, dependencies and current client commitments. Keep named, approved supply apart from recruitment plans, unsigned partners and hoped-for productivity. If the buyer can combine lots into one contract, read whether reporting, security, insurance, service credits or performance measures also combine. A combined document can create exposure beyond the sum of task volumes.
When one possible pair fails, the team has four honest routes: remove a lot, prove additional capacity, use a buyer-recognized preference or allocation rule that prevents the pair, or stop pending clarification. Forecasting that the awkward pair is unlikely does not make it performable. Win probability informs commercial judgment after feasibility; it cannot authorize a promise the supplier could not keep.
| Dimension | Required comparison | Do not count as confirmed |
|---|---|---|
| Mobilization | Concurrent milestones and scarce start-up roles | A person named in several plans |
| Operations | Peak shifts, geography and response times | Average annual availability |
| Equipment | Quantity, location and commissioning dates | Unapproved purchase intent |
| Partners | Authorized allocation and binding evidence | A positive introductory call |
| Existing work | Signed and protected client commitments | An incomplete resource calendar |
| Combined terms | Aggregate KPIs, caps and reporting duties | A sum of isolated lot summaries |
Decision record
Publish the selection without turning it into a bid authorization
The lot_bid_portfolio names the selected set and keeps every other lot visible. `bid_set_approved` means the specified set has current gate evidence, all material award outcomes pass and the authorized remaining pursuit spend is bounded. `conditional_set` names evidence that must arrive before work depending on that lot begins. `reduce_set` identifies a viable smaller set after an interaction failure. `no_lot_selected` records that no permitted set clears the gates. `source_conflict` and `authority_missing` stop the affected decision.
Record the decision owner, evidence versions, selected lots, excluded lots, possible awards, shared assumptions, budget, conditions, next review and expiry triggers. A new amendment, bidder member, partner state, deadline, award rule, delivery commitment, cost range or contract red flag reopens the decision. Passing the gate releases only the next approved planning tranche. It does not release prices, assign staff, communicate with the buyer or submit a tender.
AN-010 owns the discovery of a narrow capability inside the lot universe. AN-026 owns the correct deadline for each lot. AN-044, AN-047, AN-048 and AN-050 supply capability, delivery, investment and partner findings. AN-079 will explain what belongs inside each lot and which obligations are shared. AN-051 consumes those accepted facts and owns the final choice of pursuit set. Response planning, pricing approval and submission retain their separate authorities.
| State | Meaning | Next permitted action |
|---|---|---|
| bid_set_approved | The named set passes current gates and award-set tests | Release the approved planning tranche |
| conditional_set | One named input can still change inclusion | Resolve only that condition before dependent work |
| reduce_set | A smaller documented set remains viable | Seek authority for the reduced selection |
| no_lot_selected | No permitted set clears the current gates | Close or escalate the pursuit decision |
| source_conflict | Buyer rules or controlling versions disagree | Use the authorized clarification path |
| authority_missing | The selection or spend owner is absent | Stop dependent work and obtain authority |
Worked example
Three attractive regions create one undeliverable pair
A fictional public estate tenders heat-pump maintenance in four regional lots. Suppliers may tender for three and receive at most two. The documents evaluate each lot separately, ask bidders to rank their preferences and say that the published preference will break an excess-win outcome after scoring. No lots are combined into one contract. A maintenance company passes the hard gates for North, Central and East. West fails the required four-hour response time because the nearest approved depot is too far away.
North and Central each have an accepted delivery fit. East also passes alone, but Central and East mobilize on 1 February and both require the company’s only authorized refrigerant-training lead for twelve on-site days in the same three-week window. A second trainer is merely a recruitment request. The buyer’s allocation method would respect the bidder’s order only if all three lots otherwise won above the cap; it does not prevent Central and East from being the only two wins. The proposed three-lot set therefore contains an infeasible award pair.
The remaining pursuit cost is £38,000 for common assurance work, £24,000 for North, £19,000 for Central and £21,000 for East. Contribution ranges are positive for all three stand-alone awards, so an isolated ranking would keep them. The set test changes the answer. The approved portfolio is North plus Central. East receives `portfolio_excluded`, with a recheck only if an authorized second trainer becomes confirmed before the internal freeze. The record does not reserve the existing trainer or approve either price. It releases the bounded response plan for two lots.
| Lot | Individual state | Set interaction | Decision |
|---|---|---|---|
| North | Gate passed | Compatible with Central | Include |
| Central | Gate passed | Compatible with North; conflicts with East | Include |
| East | Gate passed alone | Central+East lacks a confirmed trainer | Portfolio exclude |
| West | Hard gate failed | Four-hour response cannot be met | Exclude before comparison |
What good looks like
Useful outcomes from which tender lots should we bid
- Every official lot appears once, including lots screened out before commercial comparison.
- Tender limits, award limits, required-all-lots rules, lot groups and combination rules retain their source anchors.
- Hard failures remain separate from portfolio exclusions and preference decisions.
- Every possible award set created by the proposed bids is tested unless the buyer’s rules exclude it.
- Shared bid cost, lot-specific effort and sunk cost are visible without being counted twice.
- Delivery fit is tested at the busiest common period rather than against annual average capacity.
- Conditional lots stay outside the approved set until their named evidence arrives.
- The recommendation remains valid only for the recorded bidder structure, tender version and decision date.
- An agent can explain each inclusion and exclusion from cited facts rather than an opaque score.
- The result stops before commitment, pricing release, staffing changes or submission.
Operating model
How to run the work
- 01
Freeze the lot universe
Record the procedure, notice, current documents, bidder configuration and every official lot identifier. Import accepted lot boundaries rather than rebuilding them during selection.
- 02
Read the buyer’s allocation rules
Capture how many lots may be tendered and awarded, whether all lots are required, which groups or combinations exist, how excess wins are allocated and whether bidder preferences have legal effect.
- 03
Apply hard gates lot by lot
Import current eligibility, capability, partner, contract, deadline and evidence decisions. Remove failed lots before weighting opportunity value or strategic interest.
- 04
Build candidate bid sets
Enumerate permitted combinations that respect the tender limit. Include conditional sets separately and preserve every excluded lot with its reason.
- 05
Model award exposure
For each bid set, list every material subset the buyer could award under the published mechanism. Do not assume the supplier can decline an inconvenient combination later.
- 06
Test economics and effort
Keep shared pursuit cost, incremental lot cost, contribution ranges, working capital and approved package effects separate. Use current inputs and show unresolved ranges.
- 07
Stress delivery across the set
Overlay mobilization, peak operations, scarce roles, equipment and existing commitments for every possible award set. A feasible preferred outcome cannot cure an infeasible permitted outcome.
- 08
Issue and expire the decision
State the selected lots, conditions, rejected alternatives, authority, spend boundary, evidence versions and triggers that force the portfolio to be decided again.
Evaluation
Questions that change the decision
- Which procedure, version, lots and bidder configuration does this decision govern?
- How many lots may one bidder tender for, and how many may it be awarded?
- Does the buyer require a tender for every lot or permit named combinations?
- Which objective rule resolves an outcome above the award cap?
- Can the bidder state lot preferences, and does the buyer promise to use them?
- Which lots fail a mandatory gate before portfolio comparison begins?
- Which costs and resources are shared, incremental or already sunk?
- What combinations could the buyer award if the proposed tenders succeed?
- Can the supplier deliver every permitted outcome under current commitments?
- Who may approve the pursuit set and the next tranche of bid spend?
Failure modes
Where teams lose control
A high-scoring lot remains in the set despite a failed eligibility or contract gate.
The team confuses the maximum number of tenders with the maximum number of awards.
An informal bidder preference is treated as a binding buyer allocation rule.
Only the hoped-for award combination is tested for delivery capacity.
One shared mobilization cost is allocated to no lot or charged to every lot in full.
A package discount is offered before its economics and approval are established.
Aggregate turnover, experience, insurance or resource thresholds are checked only per lot.
Reused prose makes two lot responses contradict their different scopes or evidence.
An agent drops rejected lots and makes the selected set look like the complete universe.
A later amendment changes a lot, cap or combination without reopening the decision.
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.
- Official lot coverage: represented lots divided by the current official lot count.
- Rule provenance: material allocation rules with current source anchors divided by all such rules used.
- Hard-gate closure: candidate lots with accepted gate records divided by candidate lots considered.
- Award-set coverage: tested material award sets divided by possible material award sets.
- Peak delivery coverage: confirmed supply divided by demand at each set’s binding period.
- Cost reconciliation: shared plus incremental remaining effort compared with the approved pursuit budget.
- Condition age: time since the oldest material input was last accepted.
- Unauthorized action count: contacts, commitments, price releases, staffing changes or submissions without separate approval; target zero.
Questions
Common questions
Should we simply bid the lots with the largest estimated values?
No. Estimated value is neither confirmed revenue nor contribution. First apply mandatory gates, then test remaining cost, margin range, cash exposure, delivery and the combinations the buyer may award.
If the buyer caps awards at two, can we safely bid four lots?
Only if four tenders are permitted and every two-lot outcome that the allocation method can produce is acceptable and deliverable. An award cap limits quantity; it does not necessarily give the bidder control over the pair.
Can the same response be reused across several lots?
Common facts and approved evidence may be reused. Each answer still needs to match that lot’s scope, criteria, limits and delivery case. Reuse should reduce drafting effort without flattening distinct obligations.
What if one lot is viable only when another lot is also awarded?
Record the dependency and verify that the buyer accepts that combination, conditional pricing or package structure. If the buyer may award the dependent lot alone, its stand-alone failure remains material.
Does selecting a lot authorize the team to price or submit it?
No. The record authorizes only the stated next planning tranche and spend boundary. Pricing release, commitments, staffing and submission need their own approvals.
Can an AI agent choose the lot portfolio?
An agent can enumerate rules, import approved records, calculate permitted sets and expose conflicts. It should abstain when a controlling rule, hard-gate input or authority is missing, and it cannot approve spend, capacity, price, partner commitments or submission.
Sources
Primary references
- Guidance: Lots UK Cabinet Office
- Directive 2014/24/EU, consolidated text, Article 46 EUR-Lex
- TED eForms procedure, group, lot and part information Publications Office of the European Union
- Section 30 VgV: division into lots German Federal Ministry of Justice and Consumer Protection
- Article R2113-1, Code de la commande publique Légifrance
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.