Wiki:Packs/Vendor Maturity Assessment Analysis
From WFM Labs
| Pack | |
|---|---|
| ID | CP-WFM-010
|
| Name | Vendor Maturity Assessment Analysis |
| Domain | WFM |
| Blocks | 1 instruction + 4 reference |
| Version | 1.0 |
| Updated | 2026-08-31 |
| Source | Vendor Operating Model Maturity · Interpreting WFM Maturity Assessments · Vendor Governance Placement |
A pack is a deployable Claude Desktop project setup published as a wiki page: one instruction block pasted into a project's custom instructions, plus reference blocks saved as .md files and uploaded as project knowledge. This pack supports analyzing completed Vendor Operating Model Maturity assessment results — the client-side instrument that grades an organization's capability to source, govern, and get value from a multi-tier labour portfolio. It is the vendor-estate sibling of Wiki:Packs/Maturity Assessment Analysis (CP-WFM-009): the same interpretation doctrine (staged gates, evidence-demonstrable scoring, no flattering averages), specialized for the sourcing instrument's mechanics — floor-versus-coverage readings, the spread-and-gap diagnostic, tier and engagement lenses, and the engagement register.
When to use it
Deploy when a Vendor Operating Model Maturity assessment has been completed — evidence items scored, register built, coverage recorded — and the work shifts to interpretation: what maturity position can the estate defend, how uneven is practice across engagements, which dimension caps the estate, and what is the cheapest credible improvement path. The instrument assesses the client organization, not its providers; this pack inherits that orientation throughout.
It is not for administering the assessment (the instrument itself carries the items and facilitation protocol), nor for grading providers — provider performance management is a different discipline with different tooling.
How to deploy
- Create a Claude Desktop project (e.g. Vendor Maturity — [Organization]).
- Paste Block 1 into the project's custom instructions.
- Save Blocks 2–5 each as a
.mdfile with the stated filename and upload all four as project knowledge. - Upload the completed scorecards and engagement register, and begin with: "Run the analysis procedure in 30-procedure.md against the uploaded results."
Block 1 — Project instructions
# Vendor Maturity Assessment Analysis — Project Instructions
## Context
This project analyzes completed Vendor Operating Model Maturity assessment
results for one organization: scored evidence items across seven dimensions
and five staged levels, an engagement register, and coverage notes. The
instrument grades the CLIENT organization's sourcing capability — never the
providers. The deliverable is a defensible maturity position, the
spread-and-gap diagnostic, and an improvement sequence.
## Routing
| Task | Open |
|---|---|
| Check a level, dimension, tier, or the model's design choices | 10-model.md |
| Apply scoring arithmetic, coverage, spread/gap, materiality | 20-scoring.md |
| Run the end-to-end analysis on returned results | 30-procedure.md |
| Draft the findings report | 40-report-spec.md |
## Disciplines
- The model is STAGED and severe by design: a level counts only if every
block beneath it is satisfied (mean ≥ 0.75). Blocks of [1, 0.5, 1, 1, 1]
yield Level 1, not Level 4. Never soften this.
- The overall score is computed from the FLOOR (weakest material
engagement), deliberately — report the number the estate can defend,
not the flattering one. Coverage is reported beside it, never blended.
- Read PRACTICE SPREAD first; use MATURITY GAP only to size the prize —
the gap steps at block thresholds and misleads standalone.
- Tier and engagement are different decompositions; never collapse them.
- Commercial is scored on the contracted estate only; a captive centre's
internal agreement belongs under Processes.
- Blank items are not-assessed: excluded from means, never counted as
zero — and "cannot answer once because providers differ" IS the
Level 2 finding, not a reason to split the assessment.
- Evidence must be demonstrable; "we intend to" scores 0. Any statistic
offered as evidence is held to the same standard — sourced and
traceable, or it does not count.
- Provider-reported figures are treated as claims (self-reported
frontline attrition has been found understated ~3x against measured).
- Level 5 is unavailable without Level 4 demonstrated across two
assessments spanning two years, whatever the evidence shows.
- Organization and provider names stay inside this project; generic
outputs carry none.
## Output
Reports follow 40-report-spec.md: floor position with the capping
dimension named; coverage beside floor; spread verdict
(consistent/uneven/fragmented); hot spots with fix class; the
generalization argument where spread is wide; improvement sequence.
Refuse to produce: unstaged averages, floor/coverage blends, commercial
scores for captives, provider league tables, or L5 claims without the
two-assessment history.
Source: Wiki:Packs/Vendor Maturity Assessment Analysis (CP-WFM-010) v1.0
Block 2 — 10-model.md
# The Vendor Operating Model — Working Reference
## The five levels — defined by unit of management, not adjectives
| Level | Name | Unit of management |
|---|---|---|
| 1 | Transactional | The transaction |
| 2 | Managed engagement | The vendor |
| 3 | Managed portfolio | The estate |
| 4 | Value managed | Business impact vs external reality |
| 5 | Workforce orchestration | All labour |
- L1: work is handed over and becomes the provider's problem; capability
is personal and leaves with the individual.
- L2: each relationship run properly (SLAs, reviews, named owners) but in
isolation; the estate has no shape. Most functions sit here.
- L3: the unit shifts from the deal to the organization — common assets,
one repository, aggregated measurement, written strategy, deliberate
placement. The most consistent inflection point in every published
framework.
- L4: capability baselines supporting statistically valid statements;
external benchmarking sets targets; portfolio analyzed for business
impact; demand and supply accountability separately assigned.
- L5: all labour — in-country, captive, contracted, contingent, digital —
governed as one adaptive system. Introduces no new capability: it is
L4 sustained across two assessments spanning two years.
The spine adapts eSCM-CL (the client-side sourcing capability model);
the staged departure is deliberate and must be stated when presenting:
eSCM-CL is a capability model, this is a staged ladder, and staging
conceals the legitimate pattern of being strong on delivery and weak on
exit.
## The seven dimensions
| Dimension | Weight | Scope | Coverage basis |
|---|---|---|---|
| Goals | 10% | Estate | — |
| Roles | 15% | Estate | — |
| Processes | 15% | Engagement | FTE |
| Relationships | 10% | Engagement | FTE |
| Technology | 10% | Engagement | FTE |
| Commercial | 20% | Engagement | Spend |
| Portfolio | 20% | Estate | — |
Commercial and Portfolio are heaviest because they change what is
POSSIBLE, not merely how well something is done. Commercial weights by
spend because FTE-weighting understates concentrated commercial exposure
(a provider with 14% of headcount can hold 26% of spend).
Three things this model grades that no published framework does: demand
management (routing-decision quality as an accountable output),
commercial construct as a graded ladder, and multi-tier labour design.
## The sharpest test in the model (Roles)
A mature function either holds the routing decision or is not held to
its consequences. Vendor management measured on provider performance
while routing decisions are made elsewhere on cost grounds is the
structural defect the model exists to surface.
## The three tiers — a separate lens from engagements
| Tier | Cost | Tenure | Flexibility |
|---|---|---|---|
| In-country | Highest | Long | Low |
| Captive LCC | Middle | Long | Low–moderate |
| Third-party BPO | Lowest unit | Short by design | High — the actual product |
Engagements nest INSIDE tiers. The captive middle tier is the one most
often missing from the conversation: it buys much of the cost advantage
while keeping tenure — and does not buy flexibility. The design error the
model catches: treating the tiers as one cost ladder and routing on unit
price.
Block 3 — 20-scoring.md
# Scoring Arithmetic and the Consistency Diagnostic
## Per item
| Score | Meaning |
|---|---|
| 1 | Present and demonstrable — the artefact or instance can be produced |
| 0.5 | Partial — incomplete, patchy across the estate, or written but untested |
| 0 | Absent — including "we intend to" and "understood informally" |
| blank | Not assessed — excluded from the mean, never counted as zero |
## Staged computation
- A level block is satisfied at mean ≥ 0.75 across its items.
- Level achieved = highest level where that block AND every block beneath
it is satisfied. [1, 0.5, 1, 1, 1] across L1–L5 = Level 1, not 4.
- Dimension score = level achieved + next level's block score (continuous
1–5). No level achieved → the L1 block score alone.
- Overall = weighted mean of dimension scores, re-weighted across
dimensions actually assessed. Overall level = integer part; below 1.0
reports as "below L1". The weakest dimension usually caps the estate —
name it.
## Estate-level vs engagement-level
Goals, Roles, Portfolio (45%): ONE answer for the whole organization.
Processes, Relationships, Technology, Commercial (55%): score the weakest
material engagement, then record the share of the estate where the item
DOES hold. One assessment per estate, never per provider — needing one
per provider IS the Level 2 finding, and splitting hides it.
## Floor and coverage
- FLOOR = the weakest material engagement — what the estate can defend.
- COVERAGE = the share-weighted reading — what is true of most of the
population. Record coverage only where practice splits; name the
exceptions so the score is auditable.
- The overall maturity score is computed from the FLOOR, deliberately.
## Spread and gap — the most useful output in the instrument
- PRACTICE SPREAD = mean block difference (coverage − floor), 0–1:
< 0.05 consistent · 0.05–0.20 uneven · > 0.20 fragmented (a set of
engagements, not an estate).
- MATURITY GAP = coverage dimension score − floor dimension score. Steps
at the 0.75 thresholds, so a tiny spread on a threshold can produce a
gap above 3. Read the spread first; use the gap to size the prize.
- Interpretation: wide spread means the capability exists somewhere but
has not been generalized — the definition of Level 2. Closing it
requires adopting nothing new, only applying what already works in one
engagement to the rest. That is the argument to make to a sponsor.
## Materiality and tiers
- Materiality threshold before assessing (default 5% of FTE or spend);
coverage across material engagements only. Proportionality is the
Level 3 test, not uniformity — for sub-threshold engagements ask one
question: is the light touch deliberate and rule-based, or neglect?
- Processes, Relationships, Technology also gather evidence per tier and
score the weakest. Commercial is scored on the CONTRACTED ESTATE ONLY.
## The register-derived concentration checks
From the engagement register (one row per delivery cell), derive before
scoring Portfolio L4: delivery cells, distinct providers, distinct
geographies, material engagements, largest provider share of FTE and of
spend, largest geography share, largest single cell. Provider count is
not diversification: four providers delivering from two countries is
concentrated. Quantify provider AND geographic concentration separately.
## Evidence hygiene
- Any statistic offered in evidence is held to the model's own standard:
sourced and traceable or it does not count. Prefer a modest verified
figure over a strong untraceable one — it survives challenge.
- Provider-reported data is a claim (self-reported frontline attrition
observed understated ~3x against measured figures). Technology L3 is
defined by eliminating that dependency.
- Where the literature has no answer (e.g. captive vs contracted tenure/
quality in contact-centre settings), "we do not know, and neither does
the literature" is a legitimate and useful finding.
Block 4 — 30-procedure.md
# Analysis Procedure — Returned Results to Findings
## 1 · Validate the register first
No analysis without the engagement register (one row per delivery cell:
provider, business line, work type, tier, geography, FTE, spend, pricing
construct). Check: material engagements identified against the stated
threshold; tiers assigned; FTE and spend both present. Gaps in the
register are findings.
## 2 · Validate the scorecards
Per dimension: items scored on the 1/0.5/0/blank convention; blanks
confirmed as not-assessed rather than missed; coverage notes present
wherever an engagement-level item splits; evidence notes present for
scores of 1 ("could someone produce the artefact?" — spot-check a
sample).
## 3 · Compute the staged position
Block means → level achieved per dimension (staged rule) → dimension
scores → overall from the FLOOR, re-weighted over assessed dimensions.
Name the capping dimension explicitly. State the position as "Level N,
capped by <dimension>" — never as an unstaged average.
## 4 · Compute coverage, spread, and gap
Coverage readings for the four engagement-level dimensions; practice
spread (verdict: consistent / uneven / fragmented) and maturity gap per
dimension. Where spread is wide, list the engagements holding the
practice — they are the internal donors for generalization.
## 5 · Run the structural checks
- Roles: does the function hold the routing decision it is accountable
for? (The model's sharpest test.)
- Portfolio: the register-derived concentration numbers — provider and
geographic concentration separately.
- Technology: which figures still depend on provider-reported data?
- Commercial: contracted estate only; note the spend-vs-FTE divergence.
## 6 · Extract hot spots
A hot spot is a gap that changes a decision. Label each by fix class:
- generalization (practice exists in-estate; apply it wider — cheapest)
- practice/configuration (build or change a practice)
- instrumentation (measurement dependency to eliminate)
- structural (decision rights, tier design, commercial construct —
needs sponsorship)
The spread analysis feeds the first class; the structural checks feed
the last.
## 7 · Sequence the improvement path
Order: generalization wins first (adopting nothing new), then
instrumentation debts, then practice builds, then the smallest
structural move that lifts the capping dimension. Tie each step to the
level gate it unlocks.
## 8 · L5 check
If any L5 claim appears: confirm the two-assessment, two-year history
exists. Without it, L5 is unavailable regardless of evidence — record
the claim and the rule.
Block 5 — 40-report-spec.md
# Report Specification
## Structure
1. **Executive summary** — five sentences: floor position with capping
dimension, spread verdict, the credibility of the evidence base, top
three hot spots, first move.
2. **Maturity position** — "Level N, capped by <dimension>"; dimension
score table (floor), coverage beside it where it differs; the basis
(items assessed, blanks, materiality threshold).
3. **Spread and gap** — per engagement-level dimension: spread, verdict,
gap; the donor engagements where practice already holds.
4. **Structural findings** — routing-decision accountability; provider
and geographic concentration (separately); provider-reported-data
dependencies; commercial construct ladder position.
5. **Hot spots** — one row each: gap · evidence · fix class
(generalization / practice / instrumentation / structural) · what it
blocks · indicative effort.
6. **Improvement sequence** — generalization first; each step tied to
the gate it unlocks.
7. **Method notes** — instrument version, staging rule stated, floor
basis stated, materiality threshold, register summary.
## Language rules
- The instrument grades the client organization, never providers; write
every finding as an estate capability statement, not a provider
verdict.
- Non-accusatory: gaps are what the estate cannot yet defend, not what a
team failed to do.
- Every figure carries its basis (FTE or spend) and its evidence status;
provider-reported figures are labeled as claims.
- Ranges where judgment enters; "unverified" and "not assessed" said
plainly; "we do not know, and neither does the literature" is a
reportable finding.
## Refusals
The report never contains:
- An unstaged average presented as a maturity level.
- A floor/coverage blend in a single number.
- A commercial score for a captive centre.
- A provider league table (the instrument does not grade providers).
- A Level 5 position without the two-assessment history.
## The one-slide version
Floor position with capping dimension · spread verdict with the
generalization prize sized · three hot spots by fix class · first move.
The generalization argument ("closing the spread requires adopting
nothing new") is the line a sponsor acts on.
Usage notes
Sizing. The instruction block is a standing per-message cost inside the project; reference files load on retrieval. Total pack size is deliberately under 4,500 words.
Drift. Blocks derive from the source articles, not copied. When Vendor Operating Model Maturity, Interpreting WFM Maturity Assessments, or Vendor Governance Placement changes materially, regenerate the affected block and increment the version.
Scope. This pack interprets results of the client-side instrument. Provider performance management, allocation design, and governance placement are adjacent disciplines — see Performance-Based Vendor Allocation Design and Vendor Governance Placement. The WFM-function sibling instrument is served by Wiki:Packs/Maturity Assessment Analysis (CP-WFM-009); the generic interpretation doctrine shared by both is at Interpreting WFM Maturity Assessments.
Change history
| Version | Date | Change |
|---|---|---|
| 1.0 | 2026-08-31 | Initial publication: instruction block + four reference blocks. |
