A numeric threshold reconciliation record compares two or more tender statements as complete rules rather than isolated digits. It preserves each original statement, source and version, then records the subject, measure, comparison operator, value, unit, period, population, trigger, statistic, tolerance and scope. The result states whether the rules are identical, equivalent after an exact conversion, nested, conditional, overlapping, incompatible or still unknown, and identifies the authority needed for the next decision.
Procurement packs repeat numbers in notices, specifications, schedules, questionnaires, pricing forms, scoring criteria, clarifications and portal fields. Two values that look different may be exact equivalents. Two identical values may govern different clocks or populations. A lower response time may be an evaluated target while a higher value is the contractual minimum. Choosing the strictest figure feels cautious, yet it can alter staffing, price, acceptance tests or the offer itself. Choosing the latest file is equally unsafe when the file does not amend the disputed requirement.
Preserve the words around every number before calculating anything. Compare the full meaning of each rule, including what is measured, for whom, over which period, from which event and under which document authority. Convert only when the conversion is exact and the dimensions already match. Use an explicit amendment, clarification or applicable order-of-precedence clause to decide which incompatible rule controls. If that authority is absent, keep the conflict open, ask a neutral question through the permitted channel and stop the unresolved number from becoming a bid commitment.
Raw statement
A number without its sentence cannot control anything
A threshold is a rule, not a cell containing digits. “P1 incidents acknowledged within 30 minutes after portal logging, 24 hours a day” combines a service class, action, maximum, duration, start event and operating window. Extracting only `30 minutes` destroys most of what must be compared. The record begins with the complete issued wording and a locator that another reviewer can open.
Capture the document name, section, page or table cell, file version, publication date, issuer, lot and source role. Include the heading, footnote, legend and referenced definition when they change the row’s meaning. A value in an award criterion may describe how extra performance earns points. The same value in a service schedule may define a contractual floor. Matching digits do not make those statements interchangeable.
Keep the raw token as well as any parsed value. `0.75 h`, `45 min`, `≥ 98%`, “at least ninety-eight per cent” and `2 to 8 °C` carry notation that can expose a scale, inclusive boundary or locale issue. Store a normalized value later. Never overwrite the original representation, because the conversion itself may need review.
| Field group | What to retain | Why it matters |
|---|---|---|
| Source identity | File, version, issuer, date, section, page or cell, lot and source role | A current schedule and a withdrawn annex cannot be given equal status |
| Raw expression | Complete sentence or row, heading, note, symbol and exact number token | The surrounding words define the comparison |
| Application | Subject, service, site, severity, phase, population, condition and exception | The same measure may have several valid thresholds |
| Measurement | Property, operator, value, unit, period, trigger, statistic and tolerance | Digits are comparable only inside the same measurement definition |
| Authority | Amendment link, clarification scope or applicable precedence provision | Team preference cannot repair contradictory buyer instructions |
Comparability
Compare dimensions before values
Build a dimension signature for each statement. Start with the subject and object: who must perform what? Add the measurable property and operator. `At least`, `no more than`, `exactly`, `between`, `before`, `after` and “triggers when” are different instructions. Preserve whether each boundary is inclusive. A range of 2 °C to 8 °C cannot safely become a midpoint of 5 °C, and a cap of 100 cases is not a target of 100 cases.
Next record scale and basis. A percentage needs its numerator and denominator. A rate needs the event count and exposure period. A duration needs a clock, start event, stop event, paused states and calendar basis. A statistic needs its aggregation rule: mean, median, percentile, maximum, minimum or every observation. “98% of urgent calls answered within 20 seconds each month” is not the same promise as “average answer time of 20 seconds each year.”
Finally, compare scope. Lot, geography, service, site type, severity, operating window, delivery phase, user class and exceptions may split one apparent conflict into two valid rules. Keep unknown dimensions visible. If one clause says “four hours from notice” and another says “four hours from classification”, the common value does not prove equivalence. The start events must first be shown to occur at the same instant under the tender’s definitions.
| Pair | Finding | Reason |
|---|---|---|
| 120 minutes and 2 hours | Potentially equivalent | Only if the clock and every scope dimension match |
| 0.98 and 98% | Potentially equivalent | Only if both use the same numerator and denominator |
| 3 business days and 72 hours | Not established | Working calendars and counting rules differ |
| Mean ≤ 2 seconds and p95 ≤ 2 seconds | Different rules | The statistics constrain different parts of the distribution |
| 99.9% each month and 99.9% annually | Different rules | The aggregation periods permit different outage patterns |
| 5 years of company experience and 5 years for a key person | Different rules | The subjects are not the same |
| 4 hours from notice and 4 hours from classification | Not established | The trigger events may occur at different times |
Normalization
An exact conversion is narrow evidence
Normalize only after the signatures match. Unit conversion can then remove a display difference: 0.75 hours equals 45 minutes, and 2,000 milliseconds equals 2 seconds. Retain the multiplier, formula and result beside both raw values. The BIPM SI Brochure defines decimal prefixes and unit-symbol conventions, which supports the arithmetic. It does not define the buyer’s service clock, allowable outage or acceptance method.
Ratios deserve the same discipline. `0.98`, `98%` and `98 out of 100` can express the same proportion, but only if they count the same successful events over the same eligible population. Exclusions can change the denominator. A monthly figure calculated per site may differ from one pooled across all sites even when both display 98%. Preserve numerator, denominator, exclusions, aggregation level and period.
Do not create precision the source did not provide. “Within approximately one hour” cannot be normalized into `≤ 60:00 minutes` without deciding what “approximately” means. A tolerance, rounding rule or grace interval must come from the documents or an authorized answer. When conversion depends on such a judgment, classify it as `conversion_basis_missing` and keep both original rules open.
Worked example
A service desk has three thresholds and no single answer yet
A hypothetical transport authority’s service specification says that Priority 1 incidents must be acknowledged within 30 minutes after they are logged in the portal, during a 24/7 service window. A response schedule asks the bidder to confirm 15 minutes after severity validation during staffed hours. An award criterion gives extra marks for a target below 20 minutes. The values are 30, 15 and 20, but they do three different jobs.
The specification row is a maximum service time from portal logging. The schedule row also looks like a maximum, yet its start event is later and its operating window may be 12 hours on five weekdays. The criterion is an evaluated target, not necessarily a minimum contractual condition. Before comparing 15 with 30, the team must know when validation occurs, what happens outside staffed hours and whether the schedule is intended to restate or replace the specification.
The first useful output is therefore not one chosen number. It is three complete rows, an exact statement of the unresolved relationships and the consequences under each reading. If validation can take 20 minutes, a 15-minute clock after validation may allow acknowledgement 35 minutes after logging. If both clocks start at logging, 15 minutes is stricter. Neither scenario proves what the buyer intended. It shows why the trigger question changes staffing and compliance.
| Source | Complete threshold | Current finding | Required action |
|---|---|---|---|
| Service specification | P1 acknowledgement ≤ 30 min from portal logging, 24/7 | Contract requirement candidate | Confirm current version and contract destination |
| Response schedule | P1 acknowledgement ≤ 15 min from severity validation, staffed hours | Different trigger and window; relation unresolved | Ask whether the row replaces, supplements or measures another event |
| Award criterion | Extra credit for offered target < 20 min | Evaluation threshold with separate function | Decide any offered enhancement through solution, price and approval review |
Clarification
Quote the conflict and ask for the governing relationship
A useful clarification identifies both sources without arguing for one. For the service-desk example: “Section 4.2 requires P1 acknowledgement within 30 minutes from portal logging on a 24/7 basis. Response Schedule row 18 requests confirmation of 15 minutes from severity validation during staffed hours. Please confirm whether row 18 replaces the Section 4.2 requirement, adds a separate measure, or should use the same start event and service window.” The question exposes the decision the bidder cannot make.
Add the operational consequence only when it helps the buyer locate the ambiguity: the trigger and window affect the proposed staffing model, price and service commitment. Do not disclose the bidder’s preferred position, negotiation floor or internal cost. Submit through the channel and by the deadline stated in the procurement. Preserve the issued answer with its audience and any amendment it invokes.
While waiting, work may continue under labelled branches if the deadline requires it. One branch can model 30 minutes from logging, another 15 minutes from validation, and a boundary case can test delayed validation outside staffed hours. These are planning scenarios, not offers. Keep shared work moving, block any affected commitment and assign a time at which the bid authority must decide whether the unresolved risk permits release.
Consistency
A resolved number must travel with its definition
When an authorized answer arrives, update the whole rule, not the digits alone. If the buyer confirms 15 minutes from portal logging for P1 incidents at all times, carry the service class, maximum operator, duration, trigger and 24/7 window into the requirement, solution, staffing plan, response schedule, evidence plan, price review and proposed service-level schedule. A bare find-and-replace can leave “during staffed hours” in one table and recreate the contradiction.
Record which earlier rows are superseded and why. Do not delete their history. Link each dependent bid object to the accepted rule version and reopen its review. The solution owner checks feasibility, finance checks the cost effect, the writer aligns every claim, and the contract reviewer checks the proposed obligation and remedy. Approval of the wording alone does not approve the operational or commercial change.
Finish with a reverse search. Search for both old and new values, unit variants, spelled-out forms and nearby trigger phrases. A search for `30` will miss “half an hour”; a search for `15 minutes` will miss `0.25 h`. Inspect diagrams, screenshots, formulas and attachments that text search cannot reliably cover. Close the record only when every releasable occurrence either uses the accepted rule or has a documented, valid reason to differ.
Decision record
The final record states what is known and what remains open
Issue one record for each independently resolvable pair or group. Give it a stable threshold identifier and list the source rows, dimension comparison, conversions, relationship, authority, decision, owner, affected bid objects and next review trigger. Use explicit findings such as `identical`, `equivalent_after_exact_conversion`, `different_scope`, `different_function`, `nested`, `conditional`, `conflict_resolved` and `conflict_unresolved`.
A result of `different_scope` still needs care. The two rules may coexist and both require delivery. `Equivalent_after_exact_conversion` means only that the values match under the recorded dimensions. It does not establish compliance, feasibility or authority to offer the threshold. `Conflict_resolved` requires a cited controlling basis and an approved downstream position. Keep these conclusions narrow enough that another reviewer can reproduce them.
Set expiry triggers. A new amendment, authorized clarification, replacement file, portal rule, lot change, service design change or corrected source should reopen the record. The closed record then remains useful history without masquerading as the current decision. That history also explains why a price, response or service model changed when the final pack is reviewed.
What good looks like
Useful outcomes from conflicting tender thresholds
- Every disputed number retains its exact source wording, locator, version, issuer and applicable lot.
- Values are compared with their subject, unit, period, population, trigger, statistic and operator intact.
- Exact conversions are distinguished from assumptions about business days, tolerances, averages or event timing.
- Apparent conflicts that govern different scopes or functions are separated instead of forced into one value.
- An incompatible pair is resolved only by an authority that applies to the procurement and the affected requirement.
- The clarification question quotes both sources and explains the practical consequence without proposing a preferred answer.
- Response text, solution design, evidence, price and proposed contract terms use the same approved threshold.
- Later amendments and answers reopen only the records and bid objects that depend on the changed number.
Operating model
How to run the work
- 01
Freeze the source set
Record the procurement, lot, stage, current files, amendment log, clarification log and portal state before comparing any values.
- 02
Capture each threshold unchanged
Copy the complete sentence, row, heading and note with its source locator, version and source role before extracting the number.
- 03
Build the dimension signature
Identify the subject, measure, operator, value, unit, scale, period, scope, trigger, statistic, tolerance and applicable condition.
- 04
Normalize only exact equivalents
Convert units or ratios only after every other dimension matches and retain both the original and normalized representation.
- 05
Test source authority
Look for an applicable amendment, authorized clarification or explicit hierarchy that resolves the precise inconsistency.
- 06
Clarify or approve a bounded basis
Ask the buyer neutrally when possible; otherwise permit only reversible work under separately labelled scenarios and named internal authority.
- 07
Propagate and verify the result
Update every dependent requirement, answer, design, evidence item, price assumption and contract position, then search for the retired value.
Evaluation
Questions that change the decision
- Are the two statements current, applicable to the same lot and issued to the same supplier audience?
- Do they regulate the same subject and the same measurable property?
- Is each operator a minimum, maximum, exact value, range, trigger, cap or evaluated target?
- Do unit, scale, numerator, denominator and rounding basis match?
- Do both rules use the same period, calendar, aggregation method and statistical measure?
- Does the clock start and stop on the same defined events?
- Are geography, site, service, severity, phase, population and exceptions aligned?
- Is one statement a minimum requirement and another an award criterion, price input or contract remedy?
- Which express amendment, clarification or precedence clause applies to this pair?
- What response, price, resource or contractual consequence changes under each unresolved interpretation?
- Who may approve the internal working basis, buyer question and final release?
Failure modes
Where teams lose control
A parser extracts 30 and 15 but loses that one is minutes after logging and the other after severity validation.
Two hours is converted to 120 minutes although the two clauses use different start events.
Three business days is treated as 72 hours without a calendar, working window or holiday rule.
An average of two seconds is treated as equal to a 95th-percentile limit of two seconds.
A monthly availability percentage is compared with an annual percentage as if the periods were interchangeable.
A threshold for the bidding company is applied to each key person, site or consortium member.
The strictest value is adopted silently and creates unpriced staffing or penalty exposure.
A portal validation value is assumed to amend the specification even though its authority is unstated.
A clarification answer is used outside the lot, question or supplier audience to which it was issued.
The resolved number is updated in the narrative but remains different in the price, plan, evidence or contract schedule.
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.
- disputed thresholds with complete source and version details
- threshold rows with every material comparison dimension populated or explicitly unknown
- normalizations supported by an exact conversion and preserved original value
- apparent conflicts classified as equivalent, scoped or functionally distinct
- incompatible rules resolved by an applicable authority rather than a team preference
- open conflicts with a buyer question, owner, deadline and affected bid objects
- dependent response, solution, evidence, price and contract items reconciled after resolution
- retired threshold values remaining in releasable bid material; target zero
Questions
Common questions
Should we always comply with the stricter tender threshold?
No. A stricter internal design may be prudent, but it does not resolve contradictory buyer instructions and may change price, staffing, evaluation or contractual exposure. Establish whether the rules are comparable, then use an applicable amendment, clarification or precedence provision. Keep the conflict open when no such authority exists.
Does the newest tender document automatically replace an older number?
Only when the procurement gives it that effect. Check whether the document is authentic, current, applicable to the same lot and intended to amend the disputed rule. A later upload can add an unrelated schedule without replacing earlier requirements.
When are two differently written thresholds equivalent?
They are equivalent only when subject, measure, operator, scope, trigger, period, statistic, tolerance and all other material dimensions match, and the value conversion is exact. Forty-five minutes and 0.75 hours can qualify. Three business days and 72 hours do not without further rules.
Can an award target differ from the contractual minimum?
Yes. An award criterion can reward an offered improvement while a specification states the minimum acceptable level. Record their different functions and check how an offered target enters the contract. Do not collapse them into one requirement merely because they measure the same property.
What if the portal accepts only one of the conflicting values?
Preserve the portal rule and determine whether it is an issued submission instruction, a technical validation or evidence of an incomplete update. A field constraint does not automatically amend the specification. Raise the mismatch through the permitted clarification or support route before relying on it.
How should a percentage threshold be compared?
Record the numerator, denominator, exclusions, population, aggregation level, period, operator and rounding method. Ninety-eight per cent per site per month can produce a different obligation from 98 per cent across all sites per year.
May drafting continue while the buyer has not answered?
Reversible work may continue under clearly labelled scenarios. Block any claim, price, design or contract position that would silently choose the unresolved number. Name the owner, clarification deadline, internal decision time and last safe release point.
When is the threshold reconciliation finished?
It is finished when the relationship is supported, any conflict has an applicable controlling basis, the result is approved, every dependent bid object has been reconciled and a reverse search finds no unexplained retired values. The record should also state which future events reopen it.
Sources
Primary references
- Procurement Act 2023, section 31 UK Parliament
- Guidance on modifying a competitive procurement UK Cabinet Office
- Directive 2014/24/EU, consolidated text EUR-Lex
- VgV section 20, tender time limits and material changes German Federal Ministry of Justice
- VgV section 53, form and completeness of tenders German Federal Ministry of Justice
- VgV section 56, examination and supplementation German Federal Ministry of Justice
- VgV section 57, exclusion of non-compliant tenders German Federal Ministry of Justice
- French Public Procurement Code, Article L2152-2 Légifrance
- French Public Procurement Code, Article R2152-1 Légifrance
- French Public Procurement Code, Article R2152-2 Légifrance
- FAR 15.206, amending the solicitation Acquisition.gov
- FAR 52.215-1, instructions to offerors Acquisition.gov
- FAR 52.215-8, order of precedence Acquisition.gov
- UNCITRAL Model Law on Public Procurement, Article 15 UNCITRAL
- The International System of Units, ninth edition International Bureau of Weights and Measures
About the author
Tony Kim
Tony writes about applied AI, dependable product engineering and the systems that turn complex response work into controlled delivery.
Ziva
Proposal software for source-grounded RFP, RFI, DDQ and questionnaire response work.
Bid, proposal, presales, security and compliance teams. Start with the workflow, constraints and evidence you already have.