Bid Evaluation Matrix

Purpose and model boundary

This model compares exactly three bidders across twelve fixed evaluation criteria. It calculates weighted scores, ranks and outright criterion leads, then repeats the winner calculation under equal weights and with the single heaviest criterion removed. The three views are a sensitivity check on the stated weighting, not proof that the selected bidder is suitable or that a procurement process is compliant.

The spreadsheet remains the calculation authority. The page sends the two fixed grids to the workbook and presents the returned winners, scores, chart and status without reproducing the scoring logic in browser code.

Inputs and units

The criterion wording and the number of criteria are fixed by this version of the workbook. Price is not entered as currency: if price is evaluated, it must first be converted into a rating on the same 0-to-5 scale using a method chosen and documented outside this model.

Governing relationships

For criterion i and bidder j, let w_i be the criterion weight and s_ij the numeric rating. With maximum rating s_max = 5, the stated-weight score is:

Score_j = sum(w_i × s_ij) / (s_max × sum(w_i))

The equal-weight comparison is:

EqualScore_j = sum(s_ij) / (12 × s_max)

Let h be the first criterion carrying the maximum entered weight. The heaviest-criterion comparison is:

WithoutHeaviest_j = sum over i != h (w_i × s_ij) / (s_max × sum over i != h w_i)

Scores are ranked from greatest to least. Equal scores receive equal rank, and a winner output lists every tied bidder. The stated-weight margin is:

Margin = greatest Score_j - second-greatest Score_j

The result is close when 100 × Margin < 2, using the workbook's two-percentage-point threshold. A bidder leads a criterion only when it alone has the greatest rating on that row; tied criteria are counted separately.

Calculation sequence

1. Validate all 36 rating labels, the twelve numeric weights and the three bidder names.
2. Map each rating label to its 0-to-5 value and calculate stated-weight points and normalized scores.
3. Rank the stated-weight scores, identify the winner or tied winners, and calculate the runner-up margin.
4. Recalculate all three bidders with equal criterion weights.
5. Identify the first maximum-weight criterion, remove it, and recalculate when positive weight remains.
6. Compare the three winner strings. Robustness is consistent only when all three agree, sensitive when they differ, and not assessable when the heaviest criterion carries all positive weight.
7. Count each bidder's outright criterion leads and evaluate Model_Status in the precedence below.

Outputs and interpretation

The three winner outputs answer three different questions: who wins under the entered weights, under equal weights, and without the first heaviest criterion. Robustness summarizes agreement among those views. It is not a confidence interval or statistical significance test. Margin_Over_The_Runnerup is expressed as the gap between normalized scores; the close-result rule reads that gap in percentage points.

The chart compares every bidder under all three weightings. Criteria_Led_* counts outright row wins, while the score and rank still reflect all weighted ratings. A bidder can therefore win the total while leading relatively few criteria.

Validation and status logic

The workbook evaluates status in this order:

Earlier conditions take precedence. For example, an exact tie returns the close-result CHECK before the workbook considers weighting sensitivity.

Assumptions and limitations

• Ratings are subjective judgements on a coarse ordinal scale. The model treats the distance between adjacent labels as equal and does not reconcile multiple evaluators.
• The twelve criteria, three-bidder capacity, rating scale and two-point close threshold are fixed on the published page. They may not match a particular procurement policy.
• If several criteria share the greatest weight, the removal test excludes only the first such row, because the workbook uses the first maximum match.
• Equal weighting and one-criterion removal are limited sensitivity tests. They do not test every plausible weighting, rating error, scenario or dependency among criteria.
• The model has no pass/fail gateway, minimum technical score, price formula, abnormally-low-bid test, conflict-of-interest workflow, moderation record, qualitative evidence store or multi-stage evaluation.
• Weights should be approved and fixed before bid scores are known. The workbook cannot detect whether weights were changed after bids were opened.
• Regulated procurement rules vary. Some regimes restrict numeric weighting or require disclosed factors and procedures that this calculator does not implement.

Restrictions and non-computing states

Both inputs are fixed-size grids: twelve criteria by three rating columns plus weight, and one row of three bidder names. The page refuses an incorrect grid shape, an unlisted rating, a blank required cell, or a weight outside the declared integer range before the calculation runs. The workbook independently returns the exact NOT VALID messages above for the download path. A zero total weight makes every normalized score undefined. When one criterion carries all positive weight, the stated and equal views remain computable but the without-heaviest view is explicitly not assessable.

Errors and warnings

A rejected entry means the submitted grid did not meet the published input rules and no workbook decision was made. Workbook NOT VALID means a score cannot be trusted. Workbook CHECK retains a computed comparison but flags ambiguous names, a close outcome, an unavailable sensitivity view or a winner that depends on weighting. A connection or calculation-service failure is an availability problem, not a procurement finding.

The workbook implements a general normalized weighted-scoring method and states that it reproduces no procurement framework or official evaluation scheme. Its release audit also identifies the jurisdiction-specific limits illustrated by FAR 15.304 and AFARS 5115.304. Those sources do not define this workbook's criteria, weights or ratings and must not be read as endorsing the model.