Interpret a CSV/CSA Maturity Profile: A Worked Example
Follow one synthetic profile from 18 answers to seven area scores and an overall 54.8. Then see what changes when three answers are unknown, and why a 92.9 can still carry a review prompt.
Written by CSV/CSA Maturity Editorial TeamPublished 11 min read
On this page13 sections
To interpret a CSV/CSA maturity profile, read the seven area scores first, then the overall score, then the two things that qualify both: numeric coverage and any review prompts. This tutorial reproduces the published example step by step. Eighteen synthetic answers produce area scores between 44.4 and 66.7 and an overall score of 54.8, in the Defined band. Making three of those answers unknown removes the overall score, and a separate profile that scores 92.9 still carries a priority review prompt.
Everything here applies to model version csa-maturity-2.0.0 of the CSV/CSA Maturity Assessment, a self-assessment of computer system validation (CSV, which here does not mean comma-separated values) and computer software assurance (CSA) practices for one team, site or system portfolio. Every profile on this page is a synthetic sample written to explain the arithmetic. None describes a real organization, an industry average or a target.
What the model counts
The assessment has 18 questions in seven practice areas. Each answer is one of four practice statements, valued 0 to 3, or "I do not know yet". Three rules produce every number below.
Area score = 100 × (sum of answer values) / (3 × questions in the area)
Overall = (sum of the seven unrounded area scores) / 7
Coverage = 100 × (numeric answers) / 18
The published methodology holds the rubric and the rules for rounding, bands and prompts. A calculator is enough to follow along.
Step 1: List the 18 synthetic answers
Synthetic sample: the complete worked profile. The level is the value of the practice statement selected for each question.
| ID | Practice | Practice area | Level |
|---|---|---|---|
| Q01 | Identifying systems in scope | System inventory | 2 |
| Q02 | Regulated-work relevance decisions | System inventory | 2 |
| Q03 | Intended-use descriptions | Intended use | 2 |
| Q04 | Needs and acceptance expectations | Intended use | 1 |
| Q05 | Risk-linked assurance effort | Risk-based assurance | 1 |
| Q06 | Justified assurance activities | Risk-based assurance | 2 |
| Q07 | Unexpected results and issues | Risk-based assurance | 2 |
| Q08 | Needs-to-evidence connections | Traceability | 2 |
| Q09 | Evidence version identification | Traceability | 2 |
| Q10 | Evidence links after change | Traceability | 1 |
| Q11 | Change impact assessment | Change management | 1 |
| Q12 | Supplier and external changes | Change management | 1 |
| Q13 | Ongoing suitability review | Change management | 2 |
| Q14 | Review and approval responsibility | Approvals and record controls | 2 |
| Q15 | Record protection and trust | Approvals and record controls | 2 |
| Q16 | Evidence package retrieval | Evidence and review readiness | 1 |
| Q17 | Responsibilities and competence | Evidence and review readiness | 2 |
| Q18 | Findings follow-through | Evidence and review readiness | 1 |
To make the levels concrete, take Q11. Level 1 is the statement "Impact reviews occur inconsistently or after implementation." Level 2 is "Changes have an appropriate pre-implementation impact review, responsible owner and approval."
The profile has eleven answers at level 2 and seven at level 1. There is no 0, no 3 and no unknown answer. The 18 values add up to 29.
Step 2: Calculate each area score
Apply the area formula to each practice area. Keep the unrounded value for the next step and round only what you display.
| Practice area | Answers | Arithmetic | Score | Band |
|---|---|---|---|---|
| System inventory | 2, 2 | 100 × (2 + 2) / (3 × 2) = 66.666… | 66.7 | Defined |
| Intended use | 2, 1 | 100 × (2 + 1) / (3 × 2) = 50 | 50.0 | Defined |
| Risk-based assurance | 1, 2, 2 | 100 × (1 + 2 + 2) / (3 × 3) = 55.555… | 55.6 | Defined |
| Traceability | 2, 2, 1 | 100 × (2 + 2 + 1) / (3 × 3) = 55.555… | 55.6 | Defined |
| Change management | 1, 1, 2 | 100 × (1 + 1 + 2) / (3 × 3) = 44.444… | 44.4 | Developing |
| Approvals and record controls | 2, 2 | 100 × (2 + 2) / (3 × 2) = 66.666… | 66.7 | Defined |
| Evidence and review readiness | 1, 2, 1 | 100 × (1 + 2 + 1) / (3 × 3) = 44.444… | 44.4 | Developing |
Intended use lands exactly on 50.0. Each threshold belongs to the higher band, so 50.0 is Defined and not Developing.
Step 3: Average the seven unrounded values
(66.666… + 50 + 55.555… + 55.555… + 44.444… + 66.666… + 44.444…) / 7
= 383.333… / 7
= 54.76190476
As a fraction the result is 1150 / 21. Rounded half up to one decimal place it is 54.8. The band comes from the displayed value, and 54.8 sits in Defined (50.0 to below 75.0).
Use the unrounded values. Averaging the rounded scores from the table gives 383.4 / 7 = 54.77, which also displays as 54.8 here. That agreement is a coincidence of this profile, and the model never rounds twice.
Why 29 / 54 = 53.7 is not this model
The obvious shortcut is to add up all 18 answer values and divide by the maximum: 29 / 54 = 0.537, or 53.7. The model does not do this. Pooling gives every question the same weight, which lets a three-question area count half as much again as a two-question area. The equal-area mean gives every area one seventh of the result, so a question in a two-question area carries 1/14 of the overall score and a question in a three-question area carries 1/21.
The two methods disagree here because the stronger answers are unevenly spread. The three two-question areas average 61.1 (11 of a possible 18 points). The four three-question areas average 50.0 (18 of a possible 36 points). Pooling weights those two groups 18 to 36. The equal-area mean weights them 3 to 4. If you see 53.7 for this profile, the pooled formula was used, and the number is wrong for this model.
Step 4: Check numeric coverage
Coverage is 100 × 18 / 18 = 100.0%. All 18 answers are numeric, so the profile is complete and may show an overall score and band.
Coverage tells you how much of the profile is scored. It does not tell you how sure anyone is, and it does not tell you that a single record was examined. A profile with 100.0% coverage can still rest on answers nobody has checked.
Step 5: Read the profile in layers
The tool summarizes this profile in one sentence: "Your self-reported maturity score is 54.8/100: Defined. This summarizes your selected practice descriptions across seven equally weighted areas." Its description of the band reads: "Your answers describe a documented baseline across several areas. Check whether it is applied consistently and how changes are handled."
Those two sentences are the least informative part of the result. Read below them:
- Areas. Five areas are in the Defined band, from 50.0 to 66.7. Change management and evidence and review readiness are at 44.4, in Developing. The overall band does not describe those two.
- Answers. Seven answers are at level 1: Q04, Q05, Q10, Q11, Q12, Q16 and Q18. Each one is a statement that a practice exists but is informal, inconsistent or incomplete.
- Prompts. There are none. A prompt needs a 0 (priority review) or an unknown answer (verification) on Q02, Q05 or Q15, and this profile has 2, 1 and 2 there. Q05 at level 1 raises no prompt, although it does appear in the list of improvement points.
- Coverage. At 100.0%, nothing is hidden behind an unknown answer.
Step 6: See which three actions the tool selects
With no prompts and no unknown answers, the tool ranks the seven level 1 answers by one rule: the lower answer value first, then the lower-scoring area, then area order and question ID. All seven share the same value, so the area score decides.
- Change management (44.4): Q11 and Q12.
- Evidence and review readiness (44.4): Q16 and Q18. This area ties with change management on score and comes later in area order.
- Intended use (50.0): Q04.
- Risk-based assurance (55.6): Q05.
- Traceability (55.6): Q10. This area ties with risk-based assurance and comes later in area order.
The tool shows at most three primary actions and merges entries from the same area. Q12 is folded into the Q11 action and Q18 into the Q16 action. Q05 and Q10 stay in the full ordered list.
| Action and questions | First step | Evidence to check | Suggested owner | Related guide |
|---|---|---|---|---|
| Assess change impact before implementation (Q11, also covers Q12) | Require a pre-implementation impact review with a responsible owner and approval for changes in scope. | Recent change with impact rationale, approval and closure record | Change owner, system owner and quality | Change control evidence |
| Run a small evidence-retrieval exercise (Q16, also covers Q18) | Ask a reviewer who did not build the package to retrieve the current evidence for one system and note any gaps. | Sample evidence package and a documented retrieval exercise | Quality, process owner and training owner | Maturity improvement plan |
| Agree needs and acceptance expectations (Q04) | Resolve ambiguous acceptance expectations for the most important needs with the people who use and oversee the system. | Approved requirements sample with acceptance expectations and change history | Process owner with users and quality | Documenting intended use |
This order is a facilitation sequence. It is not a risk classification, and a reviewer who knows the systems may change it. Completing an action does not raise an answer by itself. To change a score, the question has to be answered again against its four statements.
Step 7: Make the three risk answers unknown
Synthetic sample: the worked profile with Q05, Q06 and Q07 changed to "I do not know yet", as if the respondent could not support any statement for the three risk questions. Fifteen of the 18 answers are now numeric, so numeric coverage is 100 × 15 / 18 = 83.3%. The risk area has no score, only a possible range of 0.0 to 100.0. The profile is partial: it has no overall score and no overall band.
In place of a score the model reports a possible range. The lower bound sets every unknown answer to 0 and the upper bound sets it to 3, and both use the same equal-area mean. The lower bound is rounded down and the upper bound is rounded up, so the displayed range never understates the real one.
Lower bound (unknown answers as 0)
(66.666… + 50 + 0 + 55.555… + 44.444… + 66.666… + 44.444…) / 7
= 327.777… / 7 = 46.825… rounded down: 46.8
Upper bound (unknown answers as 3)
(66.666… + 50 + 100 + 55.555… + 44.444… + 66.666… + 44.444…) / 7
= 427.777… / 7 = 61.111… rounded up: 61.2
The possible range is 46.8 to 61.2. It crosses the line between Developing and Defined, which shows plainly why no band is given.
The actions change as well. An unknown answer on a priority question outranks every improvement step, so the first action becomes "Check the function-level risk rationale" (Q05, also covering Q06 and Q07), with the first step "Ask the responsible owner for one current example." The change management and evidence retrieval actions follow, and the Q04 action drops out of the first three. The profile also shows a verification prompt for Q05: "Verify the evidence for this practice before relying on the profile."
The partial example used in the methodology
The methodology's own partial example is a second synthetic sample: every answer is 2 except Q05 to Q07, which are unknown.
| Practice area | Answers | Score or possible range | Numeric coverage |
|---|---|---|---|
| System inventory | 2, 2 | 66.7 | 100.0% |
| Intended use | 2, 2 | 66.7 | 100.0% |
| Risk-based assurance | unknown, unknown, unknown | 0.0 to 100.0 | 0.0% |
| Traceability | 2, 2, 2 | 66.7 | 100.0% |
| Change management | 2, 2, 2 | 66.7 | 100.0% |
| Approvals and record controls | 2, 2 | 66.7 | 100.0% |
| Evidence and review readiness | 2, 2, 2 | 66.7 | 100.0% |
| Overall | 15 of 18 numeric | 57.1 to 71.5, no score, no band | 83.3% |
Lower bound: (6 × 66.666… + 0) / 7 = 400 / 7 = 57.142… rounded down: 57.1
Upper bound: (6 × 66.666… + 100) / 7 = 500 / 7 = 71.428… rounded up: 71.5
Both ends of this range sit inside the Defined band, and the profile still shows no band. The model does not classify a result it cannot compute. The only action is the verification of Q05, which also covers Q06 and Q07, because no other answer is low.
Set the two partial profiles side by side. Both ranges are 14.4 points wide as displayed. One whole area out of seven is open, which is 100 / 7 = 14.29 points, and outward rounding adds the rest. The width comes from which questions are unknown. The position comes from the 15 answers that are known.
Step 8: Read a 92.9 profile that carries a prompt
Synthetic sample: every answer is 3 except Q15, which is 0. Level 0 on Q15 is the statement "The team cannot explain dependable controls against loss, alteration or inappropriate access."
Approvals and record controls: 100 × (3 + 0) / (3 × 2) = 50.0
Each of the other six areas: 100.0
Overall: (6 × 100 + 50) / 7 = 650 / 7 = 92.857… shown as 92.9, Measured
Coverage is 100.0%. Alongside the score, the profile shows a priority review prompt for Q15, headed "Check how assurance records are protected" and introduced with "Review this practice independently of the overall score." The evidence it names is "Applicable access and record controls, retention responsibility and a retrieval or recovery check."
Two points follow from this profile.
First, the prompt does not touch the arithmetic. The formula gives 92.9 and the profile shows 92.9. Nothing is capped or deducted.
Second, an average absorbs a weak answer at every level. The approvals area shows 50.0, Defined, although one of its two practices sits at the lowest statement. The overall score then absorbs that area into a Measured total. The prompt exists so that a reader who looks only at the total still sees the Q15 answer. The article on unknown answers and high scores takes this further.
"Measured" is a name in this rubric. It does not mean that evidence was inspected or that controls are effective.
Load the example, then assess your own scope
The worked example section of the assessment page shows this synthetic profile, and "Try this example" loads its answers into the questionnaire. The tool asks before it replaces anything: "This will replace your current answers. Load the example?" You can choose "Load example" or "Keep my answers".
Starting your own assessment does not load the example. To assess your own scope, choose one team, site or system portfolio, answer the seven areas and review your answers. Before your profile appears, a short form asks for your first name, last name, work email and company. The profile is then shown on screen with a PDF download.
What this example does not establish
- The numbers are arithmetic on invented answers. They show how the model behaves, not how any organization performs.
- Reproducing a score confirms the calculation. It does not confirm that the answers behind it were accurate.
- The three actions are a starting order for discussion. They are not findings and not a risk ranking.
- A possible range is not a forecast, and a band is not a regulatory status.
- The results belong to model version
csa-maturity-2.0.0. A later version may give different numbers for the same answers. - This page makes no regulatory claim. A profile is not a determination of compliance, and applicability belongs with your quality and regulatory specialists.
Related reading
- The assessment methodology in full states every rule used above, including the recommendation rules and the change log.
- Unknown answers, benchmarks and high scores: how to read them sets out what each kind of profile lets you say.
- Turning profile gaps into an owned improvement plan covers the next step: giving actions like these an owner, evidence to check and a review date.
This article describes assessment model csa-maturity-2.0.0.
