Blog · Procurement and supply chain · Procurement
Separate like-for-like procurement KPI movement from definition changes, evidence corrections and denominator exclusions using an explicit bridge.
Spend under management rises from 50% to 83.3%. That sounds like a major control improvement until the reviewer learns that the team also changed what counts as managed and removed a category from the denominator. A bridge should show how much movement survives under comparable rules.
The SUM guide retains the chosen measure. This page owns change attribution, not a new SUM definition or a savings claim. A rising procurement coverage rate does not by itself prove financial savings; that question belongs to the claimed-versus-realized savings guide.
Keep the original period, numerator, denominator, rate, policy version, source extract and approved exclusions. Store subsequent corrections with a reason and predecessor version rather than overwriting the old result.
Separate evidence repairs from policy changes. Finding a previously missing contract is different from deciding that approved suppliers now count as managed. Reclassifying a supplier alias is different from routing new purchases through a contract. All can move the percentage; they call for different management conclusions.
Document what changed before calculating the bridge. A table of version IDs, effective dates and changed predicates is more useful than a note saying methodology updated.
Recalculate both periods under a comparable policy, supplier/category mapping and denominator. Retain the observed purchases for each period while holding these rules constant. If a company was acquired, a category disappeared or source coverage changed, show that as a perimeter/evidence issue rather than silently dropping the unmatched rows.
A common-perimeter movement still reflects the purchases and their mix. It may result from different categories being bought, not from improved routing inside every category. Inspect category and unit results before attributing the movement to a procurement initiative.
The unknown-status method handles incomplete evidence. Unknown dollars should not move into a confirmed bucket merely to balance the bridge.
Both periods initially contain $100,000 of comparable positive spend. Under the old policy, Period 1 has $50,000 managed. Period 2 has $60,000 managed under the same policy. The newly approved policy then counts an additional $15,000 in Period 2 as managed. Finally, a revised perimeter excludes $10,000 that was unmanaged under both policy versions.
| Calculation step | Managed | Denominator | Rate |
|---|---|---|---|
| Period 1 as originally reported | $50,000 | $100,000 | 50.0% |
| Period 2 on old policy and common perimeter | $60,000 | $100,000 | 60.0% |
| Period 2 on new policy, same perimeter | $75,000 | $100,000 | 75.0% |
| Period 2 on new policy and revised perimeter | $75,000 | $90,000 | 83.3% |
The rate bridge is 10.0 percentage points of like-for-like purchasing movement, 15.0 points of classification-policy change and approximately 8.3 points of denominator change. Together they explain the 33.3-point reported increase, subject to displayed rounding.
Only the first ten points survive this example's common-rule comparison. Even those ten points are not proof that a particular intervention caused the change. The reviewer needs the line-level movements and category mix to make that claim.
Hold the denominator constant while showing classification effects, then apply the revised perimeter as a separate step. If an exclusion contains both managed and unmanaged spend, remove both relevant amounts; do not subtract only the denominator.
For ratios, attribution depends on step order when several inputs change together. Record the order, or use a stated decomposition method and reconcile its residual. Do not present a sequential bridge as the unique mathematical attribution of every interaction.
An evidence-only update deserves its own line. If $5,000 changes from Unknown to Managed because a missing agreement is supplied, call it evidence resolution. Preserve the previously issued uncertainty range alongside the corrected result.
For each source line retain old/new policy status, old/new perimeter inclusion, evidence version, source period and change reason. Group the line differences only after checking unique keys and candidate matches.
Power Query documents multi-column joins and grouping. A movement ledger still needs reason codes from reviewed policy and evidence. Merge queries, Group rows.
Issue the original reported rates, common-rule rates, sequential bridge, changed-line list and unresolved evidence together. Check numerator and denominator totals at every step. A bridge that does not reconcile should not be issued as an explanation.
Use the procurement pilot specification to agree these controls. Bring two authorized extracts and both policy versions to discuss a scoped review. This analysis can explain reported movement; it does not establish that an automated restatement service is available or that the rate increase delivers cash savings.
No. Keep evidence resolution as its own bridge reason when no invoice or purchasing route changed.
Not necessarily. Record the order or use an explicit decomposition method, and reconcile any interaction or residual.