Keep every RFP draft on one verified fact baseline

Meaning and scope

The shared value needs a shared definition

Asterford’s fictional tender concerns support for an existing terminal estate. The accepted inventory observation for the stated reference date identifies 84 installed terminals and 16 separate spare terminals. Together they make 100 owned terminals in the defined estate. The example assumes that these source facts have already been checked by their competent owner. It does not use the baseline process to settle a dispute about the inventory.

Create separate propositions for installed count, spare count and owned total. Each has a stable identifier, a definition, unit, population, reference date, source version and accountable owner. The total also records its derivation: 84 plus 16, with the categories confirmed as disjoint and exhaustive for this estate. Without those conditions, even a correct addition would not establish the total.

The Government Data Quality Framework distinguishes consistency from accuracy and completeness, and relates quality to intended use. It is government data guidance. The bid application is to test whether repeated values describe the same thing, while keeping separate evidence for whether those values are correct and whether the inspected set is complete.

Do not create a generic field called terminal count and expect authors to infer its meaning. A method statement about installed equipment should bind to the installed proposition. A sentence about all equipment owned can bind to the total if it preserves the population and date. The price model needs the buyer’s actual charging basis, which must be identified before choosing either number.

Authority and uncertainty

Acceptance for drafting does not turn assumptions into facts

The working set can contain more than observed facts. It may include the selected service design, an approved interpretation, a conditional planning assumption or an unresolved question. Give these different types and statuses. A decision to model two training sessions is a supplier choice; an assumption that the buyer will provide a room remains conditional until the required evidence or agreement exists.

For each item, name who owns the evidence, who can interpret its business meaning, who can approve its use in this bid and who maintains the record. The coordinator may administer all identifiers without holding any of those substantive powers. Copying a value from an approved slide does not establish that its original approval covered this customer, period or claim.

The 2025 AQuA Book describes separate data, assumption and decision logs, documented validity conditions and version control. Its scope is analytical assurance. Adapted here, that separation keeps a working hypothesis from being displayed as an established fact and makes the reason, authority and limits of its use inspectable.

Hold a materially disputed proposition outside the permitted fact set, or carry it as explicitly unresolved with its prohibited uses. Route it to the source-conflict owner. The baseline must not manufacture agreement by selecting the most convenient source or by taking a majority vote among drafts. Ten copied sentences can all descend from one mistake.

Baseline identity

Keep the current lookup and the cited snapshot separate

R6 identifies the selected revisions of the facts and decisions, their source references, scope and permitted drafting use. The item revisions need not share the number six. Preserve a recoverable description of that combination and record who established it, when, and subject to which open conditions. Avoid a baseline that depends on whatever a dashboard happens to display when someone opens it.

W3C Data on the Web Best Practices recommends version indicators, version history and separate identifiers for a dataset series and its individual versions. It concerns published data. The limited internal adaptation is to distinguish a convenient current-baseline lookup from the exact snapshot an author or reviewer used; it does not require publishing protected bid records.

An author can use a current lookup to find R6, then record R6 and the item revisions used. If the lookup later points to R7, the old draft does not acquire R7’s facts by association. Conversely, a file saved after R7 exists may still contain R6 material. File timestamps, baseline identifiers and factual conformance answer different questions.

Keep the source and its permitted capture distinct. A live database query, a dated extract, a screenshot and a manually transcribed sentence are different representations. Record parameters and locators needed to recover the evidence within its access rules. A matching hash can support byte identity; it does not prove authority, semantic equivalence or permission to disclose the source.

Inspection population

The active-copy inventory sets the limit of the claim

List the objects that can still feed the response: editable answers, calculation workbooks, diagrams, attachments, partner drafts, local working copies and rendered exports. Record their owner, location, artifact version, declared baseline, return route and current status. A document does not leave this population because it is inconvenient to retrieve. Remove it only when its owner and the integration process confirm that it cannot be used as a current input.

NASA’s configuration status accounting records current and historical documentation, identifiers, proposed-change status and discrepancy dispositions. In its systems-engineering context, this makes controlled items visible. The bid adaptation is an accountable inventory of the artifacts being checked, including pending and superseded copies, rather than an assumption that one folder contains every active input.

Use proportionate discovery within authorized locations and ask contributors to identify offline copies. Do not search private devices, unrelated projects or inaccessible partner systems without permission. Record those boundaries. If an expected partner draft has not arrived, its state is not received, not consistent. A delivery acknowledgement proves receipt, not that its contents conform.

The declared inspection scope may be narrower than the whole bid. State that directly, such as the terminal-estate claims in six named artifacts. A check of those claims cannot certify every other assertion in the same files. Sampling can reveal defects and support a risk-based review; it cannot substantiate a claim that every material occurrence was inspected.

Worked inspection

Different values can conform, while matching values can fail

At the fictional 10:30 review, Asterford declares six active artifacts for the terminal-estate check against R6. The method and diagram each describe 84 installed terminals. The executive summary says 100 owned terminals, including 16 spares. All three can conform because they use the appropriate scoped facts. The point is preservation of meaning, not replacement of every occurrence with 84.

The price model labels its line installed terminals but uses 100. For the case, D-CHARGE has already established that this line requires the installed count. The conformance failure is therefore specific: the model uses the owned total for an installed-only field. Correcting the input still requires the authorized model owner and subsequent pricing checks; the baseline report does not itself approve a new price.

The PDF candidate was exported from R5 and still describes 78 installed terminals. It is stale against R6 and must not inherit the method’s successful check. The sixth artifact, the partner’s offline contribution, has not been received. Its factual content is unknown. The resulting report records three conforming checks, one scope mismatch, one stale contradictory export and one unavailable artifact.

Five of six artifacts were available for inspection, about 83.3 percent. Three of six have a conforming result for the declared claims, 50 percent. Neither number is a bid-readiness score. In particular, the unavailable partner contribution might contain a consequential error, and unrelated claims in the other five artifacts remain outside this narrow check.

NIST SP 800-128 section 3.4 links configuration monitoring to comparison with approved baselines and identification of unrecorded components. It addresses information-system security. Here it supplies a bounded inspection analogy: check observed artifacts against the intended set and look for omitted consumers. This is not a claim of NIST compliance or a security assessment.

Evidence of conformance

A baseline stamp cannot replace inspection of the claim

Work from each controlled fact to its declared consumers, checking that the right proposition survived in prose, cells, diagrams and exports. Then work back from material draft claims to their baseline bindings. The reverse pass finds an unsupported claim that no owner listed as a consumer. Search can locate exact numbers and distinctive phrases, but manual review must cover synonymous wording, image text, rounded values and implicit scope.

Record the artifact version, location, baseline item revision, check performed, result, reviewer and time. For derived values, preserve the input bindings and formula or documented method. A unit conversion can conform without displaying the source’s original number; a rounded percentage can fail if its rounding changes the asserted threshold. Mark an uncertain semantic match for review rather than treating a similarity score as proof.

A draft need not be rewritten solely because the baseline identifier advanced. If R7 changes only a fact the draft does not use, inspect the dependency boundary and record compatibility with the relevant R7 items. Do not relabel the old inspection as a new one. If the changed item is used, the former result becomes insufficient until the affected claims and decisions have been rechecked.

Changes to the draft itself can also invalidate a result. A shortened summary may remove including spares while retaining 100. A re-export may include an older diagram. A source correction can invalidate a previously supported baseline without changing any draft bytes. Define these triggers and the latest required source check so that an apparently quiet workspace is not mistaken for permanent validity.

Operational handoff

Hand over the checked set, the gaps and the next decision

The deliverable contains the baseline manifest, fact and decision records, active-copy inventory and claim-conformance results. Identify the current working snapshot, the exact inspected artifact versions, excluded historical copies, missing items and unresolved authority questions. The recipient should be able to reproduce the terminal-count check without reading the team’s chat history.

Keep the internal report within its permitted audience. Authors may receive a scoped view that exposes the approved proposition and usable evidence without revealing protected costs, personal data or partner material. A shared baseline is a shared set of rules and references, not universal access to every underlying source. Required customer disclosures still need their own accurate, authorized treatment.

When an addendum changes the factual basis, hand the event to the addendum-propagation process. When two teams propose interacting changes, use the joint-revision process. When the underlying facts disagree, use source reconciliation. This standing baseline record receives their controlled outcomes; it does not replace their decisions or the final release manifest.

An assistant may inventory authorized artifacts, extract candidate bindings, perform reproducible comparisons and draft the discrepancy report. It must not settle source authority, convert a hypothesis into a fact, edit an operational source to fit the bid or certify inaccessible drafts. Changing approved positions, contacting outsiders, disclosing protected records and submitting the offer require separate authorization.