PPP Contract Management Toolkit

Performance Monitoring Maturity Assessment

Not whether the monitoring regime exists, but how well the authority runs it: whether the report arrives and is validated before the invoice is paid, how much of the performance data is the contractor's word and how much the authority has checked, whether the audit and inspection rights have ever been used, what happened to the deductions and the exclusions, whether rectification closes, and whether the records would stand up. A maturity level per dimension, the overall level at the lowest, never an average.

v1.0 — September 2026

Methodology

Eight dimensions of monitoring practice, each at one of four levels: Reactive, Informal, Systematic, Assured. For each dimension the team states its level against a written description, and the register figures for the periods reviewed cap it: a level the figures contradict is not credited, and a dimension with no figures at all is held at Level 2. The overall level is the lowest dimension. The dimensions, the register fields and the statuses (reported against verified, calculated against waived, required against completed) follow the product's Build 2 performance, payment and evidence registers; the rights and periods are the precedent's monitoring, payment, records and default clauses. The Contract Health Check asks whether the regime exists; this asks how it is run, and the regime entries here are context, not score.

Contract identity
Required once an amount is entered.
On a user-pay contract the invoice, disputed-amount and set-off lines apply only to a stream the authority pays; leave them blank where there is none.
Blank means monthly. Used to convert the periods reviewed into months (for the audit interval and the trigger windows).
A whole number; blank means one. Used to count the reports due and the reports a trigger window can hold (for example 3 where reports are monthly and payment quarterly).
The regime, as the executed contract sets it (context, not score)

What the agreement provides. The Contract Health Check tests whether these exist and hold together; here they are recorded so that the practice can be measured against them and so that a right the contract does not give is not scored as a right unused. A regime gap is reported as a note. The paste-text step proposes the provisions present with the phrases found; it proposes the weakest option the phrase supports and never sets a number.

Shown in the results as context; not scored.
Shown in the results as context; not scored.
Kept as context only; the limbs entered below are what is scored.
Each limb is a number of erroneous reports in a window of months, payment periods or days, as the executed contract sets it; the trigger is met where any limb is met. Enter the figures from the clause; leave blank where the contract has no trigger.
Shown in the results as context; not scored.
Shown in the results as context; not scored.
Read from the contract text (optional)

Paste the agreement, schedules included. The tool looks for the periodic report and its due day, the validation window, self-monitoring and the authority's own monitoring, the increased-monitoring trigger, the quality audit right, records and helpdesk data, audit access, retention, service failure points, warning notices, the annual plan and report, the satisfaction survey, the disputed-amount window, the no-set-off clause, benchmarking and a blank payment schedule, and proposes the provisions present with the phrases found. It proposes the weakest option the phrase supports; nothing is set until you accept it; no number is ever set from the text. The text stays in this browser tab.

Reporting and validation

Counts from the authority's own registers for the periods reviewed. Where the register does not hold a figure, leave it blank: a blank is reported as not evidenced, and where a rule would test it (its base is above zero, or it is the base of a figure entered) the dimension is held at Level 2 until it is entered; the field is marked after the run. A dimension whose figures are all blank is held at Level 2. A zero is a fact and counts.

Twelve monthly periods is the usual window.
Of the reports that were due; a period not yet due does not count.
Authority-paid stream only.
Not clarifications resolved in the validation window.
For a limb of 2 in 3 months: the most erroneous reports in any three consecutive months. Where the window is longer than the periods reviewed, count the earlier periods too. Tested only where the contract has the trigger.
KPIs, data and verification
Of those due in the periods reviewed; a quarterly KPI measured each quarter counts.
Helpdesk, sensors, an asset system, the authority's own inspections; not the contractor's summary.
Monthly KPIs times periods, plus the less frequent ones as they fell due.
Verified here means re-measured, inspected or reconciled to a source the authority holds. In the product, "verified" is the status a validator sets after that check; a reported figure never becomes verified by being reported.
Tested against the number of exercises: a count the date contradicts is not evidenced.
Audit and access
Leave blank where the contract gives no audit right.
Used where the interval is longer than the window reviewed.
Deductions, exclusions and payment
Failures the authority accepts occurred, whoever reported them.
These go to the dispute procedure and are not applied until resolved.
Of those neither waived nor disputed.
The notice under the payment clause is the authority's; a deduction the contractor disputes is not one.
Tested only where the contract has a no-set-off clause, as entered on tab 1.
Rectification and escalation
Records, lifecycle, plans and people
Report, raw measurement, rectification record, payment calculation, correspondence on disputed lines.
Staff, inspections, systems, audits; compared with the amount deducted, for context.
Practice, by dimension

For each dimension, choose the description that fits the way the work is done today, not the way the procedure says it should be. The figures on tab 2 cap the level: a level the figures contradict is not credited, and a dimension with no figures is held at Level 2.

Run the assessment from tab 3 to see the level, the dimensions, the indicators and the findings.

Purpose

The Performance Monitoring Maturity Assessment asks how well a contracting authority runs the monitoring regime its contract gives it. It is screening-level. It takes the counts the team can read from its own registers for the periods reviewed, the team's own description of its practice, and the provisions of the executed contract, and returns a level per dimension and an overall level with the reasons. It does not measure the private party's performance, decide a deduction, characterise any conduct, or certify anything; it measures the authority's grip on the numbers it pays against.

The Contract Health Check in this series asks whether the regime exists and holds together: whether each KPI has a target, a method and a data source, whether the monitoring system is installed and accessible, whether deductions are calculated and applied. This tool records those provisions as context so that a right the contract does not give is not scored as a right unused, reports a regime gap as a note, and scores practice only.

The four levels

LevelNameWhat it means in practice
1ReactiveThe authority reacts to what the contractor reports or to complaints; no check of its own, no register that separates reported from verified, deductions and rectification handled case by case.
2InformalChecks happen, records exist, but not on a cadence and not to a standard: some reports validated, some results verified, some deductions applied, rights used occasionally, evidence files patchy. A dimension with no figures entered is held here.
3SystematicThe cadence is kept: reports validated before payment, a register with reported and verified kept apart, the authority's own checks on a plan, deductions calculated and applied in the period, rectification tracked with due dates, evidence filed per period, waivers logged as decisions.
4AssuredSystematic, and tested: verification covers a stated share of results by design, audit and access rights used on a schedule with outcomes notified, the recorder and the validator are different people, the tally against default thresholds is reconciled with the contractor's, the annual review, the survey and the lifecycle check feed back into the plan.

The eight dimensions and the caps, as coded

The product this series draws on, PPP Contract Review & Operating Map v1.0.0, runs the operating phase through registers whose statuses this tool borrows: a KPI record carries a source (authority, contractor, site record, system import, independent certifier and so on) and a verification status (unverified, under verification, verified, rejected) that a person with the validator role sets, and the product refuses to conclude a breach from a figure that is not verified; a deduction carries a status (not calculable, potential, calculated, agreed, disputed, waived), and the register note says a deduction is computed only where the formula is unambiguous and the inputs are verified; rectification carries required, in progress, completed, failed; evidence carries required, received, verified, missing, disputed. The dimensions follow those registers. The table lists every cap the code applies. A blank figure a rule would test holds its dimension at Level 2 as not evidenced (a numerator blank against a base above zero; a base blank under a numerator entered; a count above a base of zero, which is inconsistent and gives no share), and so does a dimension whose base is a zero the tool cannot test (no report due, no result recorded, no failure established, no rectification expired); a practice select answered "not applicable" is not evidence.

Dimension (primary figures)Cap at 1Cap at 2Cap at 3
1. Reporting cadence and validation (reports on time, late, missing; invoices paid unvalidated)Half or more of the reports due not received; invoices paid unvalidated in half or more of the periods reviewed.Any report not received; any invoice paid unvalidated; on-time share of the reports due below 75 per cent.On-time share below 95 per cent.
2. KPI measurement and data quality (KPIs measured at the contract's frequency; KPIs with a readable source; log reconciliation; erroneous reports)None.Measured share below 70 per cent; no KPI with a readable source; the log not reconciled; the most erroneous reports in any window of a limb of the trigger at or above that limb's figure, where the contract has the trigger (the limbs entered as fields).Measured share below 100 per cent; readable share below half; the log reconciled only sometimes.
3. Verification against self-reporting (results recorded and verified; exercises; outcomes notified; fraud; independent monitor)No KPI result recorded; no result verified and no exercise; misreporting notified as fraudulent with the increased-monitoring notice answered "not used", where the contract gives the right.Verified share below a quarter.Verified share below three quarters; results verified by reconciliation only, with no exercise; outcomes not notified, or outcomes notified only sometimes; an independent monitor present and not used.
4. Audit and access rights exercised (audits against the interval, or the last audit where the interval implies one or none in the window; records access; the contractor's QA reports)No audit and no records access exercised, where the contract gives both rights.Audits below half of those the interval implies (where it implies more than one); no audit in the window and no last audit date; records access never exercised where the right exists; a records right as entered with no access in practice.Audits below those the interval implies; the last audit older than the interval; the contractor's QA reports received and not read, or not received.
5. Deductions, exclusions and payment (failures, calculated, applied of those neither waived nor disputed, waived, exclusions checked, notices, set-off, caps, traceability)Failures with no deduction calculated; set-off taken while disputed where the contract has a no-set-off clause.Deductions calculated for fewer than three quarters of failures; waived and disputed together at half or more of those calculated (not evidenced); applied share below three quarters; waivers not logged; waivers at half or more of the deductions calculated; exclusions checked below half; a disputed-amount notice late; deductions not traceable.Deductions calculated for fewer than all failures; waived and disputed together above a quarter of those calculated; applied share below 100 per cent; waivers above a quarter of deductions (all logged); some waivers unlogged; exclusions not all checked; caps not checked every period; most but not all deductions traceable.
6. Rectification and escalation tracking (rectifications expired and completed; overdue; repeats counted; points tallied; warning notices; cure)None.On-time completion below three quarters; repeats not counted; no tally where the mechanism exists; breaches at the warning threshold with no notice where the ladder exists.Rectifications completed on time below 100 per cent; any rectification overdue; tally kept with nobody named; repeats counted only sometimes; cure not tracked, or cure tracked only partly.
7. Records and evidence (complete evidence files)No period with a complete file.Complete files below half.Complete files below 100 per cent.
8. Review, lifecycle, plans and people (lifecycle check; annual documents; survey; benchmarking; staff)None.Lifecycle not checked; annual documents not received where required; no monitoring staff.Lifecycle checked sometimes; annual documents late or not reviewed; survey not run when due or results not used, where required; benchmarking not run when due, where required.
Dimension level = min(stated level, every cap the figures set); a dimension whose primary figures are all blank is capped at Level 2 (not evidenced); a stated level of "Don't know" is shown at the lower of Level 2 and the cap, and sets the overall level like any other
Before any rule runs, every count it reads is classified as valid, blank, negative, above its base (a base of zero included) or contradicted by another entry (on time, late and missing that do not account for the periods, with at most one report, the latest, not yet due; a window count above the total; waivers and disputes above the deductions calculated; an amount waived against no waiver; warning notices with no breach at the threshold, or more than a warning and a final warning for each; overdue rectifications above those not completed on time; an audit count, or an exercise count, the date of the last one contradicts; a window count of erroneous reports that cannot hold the total over the periods reviewed; a fractional count). Anything but valid, where the rule applies, is one not-evidenced cap at Level 2 naming and marking the fields, and the figure earns nothing
A clean year (no failure established, or no rectification expired) cannot test that dimension and holds it at Level 2 whatever else is entered; so does a year in which every deduction calculated, or half or more of them, is waived or unresolved in dispute
The increased-monitoring trigger is entered as limbs, one row each: a number of erroneous reports in a window of months, payment periods or days, as the executed contract sets it; the trigger is met where any limb is met, each limb tested on its own count of the most notified in any window of that limb. Nothing in the contract wording is read for the score; the wording is kept as context. Windows are counted in reports, at the performance reports per payment period entered (one where blank): a window in months holds the reports evenly spaced within it, one in payment periods that many periods' reports, and one in days the reports the calendar allows (for reports on the same day of each month, the shortest and longest spans of the calendar itself; otherwise 28 to 31 days a month). Where the tool concludes against the entries (a limb met, a count too low, counts on several limbs that cannot hold together, a window that holds every report reviewed) it uses the fewest reports a window certainly holds; where it must not (a limb it cannot meet, a count above what a window can hold) it uses the most a window could hold. Erroneous reports are placed only among the reports received (on time or late), over the reports due; more erroneous reports than reports received is inconsistent. At most one report, the latest, may be not yet due at the review date; it holds no erroneous report, and a larger shortfall is not evidenced. Every limb's count is tested against its figure whatever the number notified in the periods reviewed: a window that can reach back counts earlier periods, a count at or above the figure meets the trigger, and a blank count where a count could meet the limb (always, for a window that can reach back) is not evidenced. Day windows with reports on the same day of each month are tested in every alignment of the calendar (start month and day, four years), and a conclusion against the entries must hold in all of them. A window count below the least the total forces, above the total (unless the window can reach back before the reports reviewed) or above what the window can hold, or a count that the other limbs' counts make unreachable, is inconsistent (not evidenced); where the total forces a window at or above a limb's figure, the limb is met. A limb blank beside others, in any position, or part-entered, is not evidenced where a count could meet it, and a regime gap otherwise; removing a limb moves the ones below it up; a limb its window cannot hold is a regime gap to check, and not evidenced where the reports notified could meet it; a window that can reach back before the reports reviewed, with its count not entered, is not evidenced; more erroneous reports than reports due are inconsistent; the notice answered used with no fraud notified and no limb met is inconsistent, whatever the trigger is answered
A blank figure a rule would test is not evidence either way: the dimension is held at the lower of Level 2 and the level the worst figure would give, and never scores above an entered figure
A yes/no practice question left blank where it applies (or answered "Not checked", or "not applicable" where the other entries rule that out) is not evidence either: the dimension is held at the lower of its level and Level 2, and the field is marked
A contract provision answered "Not read", or left blank, is not an answer: where the practice would fire a cap under that provision, the cap fires as not evidenced at the lower of its own level and Level 2, so a provision not read never scores above one present; the independent-monitor select, which carries the practice too, is held at Level 3 until it is answered
Overall level = the lowest dimension level (never an average); a dimension the contract makes inapplicable (no periodic report) is shown as not applicable and left out, and any figure or level entered against it is kept and reported as a finding; where that entry is below the overall level, the headline level carries a qualifier saying so
Months between two dates are calendar months (a year to the day is 12); a date after the review date, or after today, is not used, and a review date after today is noted and marked
"Every dimension is assured" is printed only where there is no finding or indicator flagged, no note, no field marked for entry, no regime gap and no contract provision not read, and the headline reads "Level 4: Assured" only then; otherwise it reads "Level 4 — not assured" with the reasons
Profile, the required regime entries, the periods and KPI counts, or any dimension statement missing → Not concluded
Four or more dimensions stated "Don't know" → Not concluded
Audits the interval implies = floor(periods reviewed × period length in months ÷ audit interval in months); where that is zero the last audit date is compared with the interval instead
Shares are clamped to the range 0 to 100 per cent; a count above its base is noted

The overall level is a suggestion that a named person accepts or overrides with a reason; both appear in the export; no override is accepted while no level has been suggested.

The precedent's monitoring machinery

The precedent has the contractor monitor its own performance under the payment mechanism (Clause 25.1), lets the authority monitor at any time for any purpose with the contractor's assistance and notify the outcome (25.2(a)), and lets it increase both its own and the contractor's monitoring where a report is fraudulent or a bracketed number of reports in a bracketed number of months are erroneous, until the contractor has shown it can report properly (25.2(b) and (c)); the contractor bears its own increased-monitoring cost on both limbs, and the authority's own cost is recoverable only on the fraud limb (25.2(d)). The monthly report is due on a bracketed Business Day after each period with the draft invoice; the authority reviews and validates it during a bracketed window before the final invoice (32.2). The authority may audit the contractor's quality system at approximately bracketed intervals and carry out other monitoring, spot checks and audits (29(e)), and the contractor's own quality manager audits the quality system and reports to both parties (29(a), 29(e)). The contractor keeps records of incidents, maintenance and staff matters, the first two open to the authority's inspection and all of them to its auditors (72.3, 72.4), with helpdesk data by direct access, for a bracketed number of years (72.5), and gives the authority's representative all records for any purpose in connection with the agreement (64); a quarterly list of information given to the lenders is provided (72.7). A disputed amount is notified by the authority within a bracketed window with evidence and is not set off until finally determined, save under the retention-fund provision (32.5, 32.6, 40). Service failure points count towards default on three bracketed windows (Contractor Default limbs (i), (j) and (k)); breaches that carry no deduction go through a warning notice and a final warning notice (36.1, 36.2). Deductions continue during a relief event but are disregarded for the default count (48.9). An annual service delivery plan and an annual executive report are due on fixed dates (31); a customer satisfaction survey, optional in the precedent, runs on stated dates (32.10); market testing has its own optional provision (32.11), and benchmarking is where the executed contract provides it. The precedent's payment mechanism schedule is blank, so the KPI list, the points, the caps, the ratchets and the rectification periods come from the executed contract. There is no operations-phase role for the independent certifier; verification in operations is the authority's own.

The product's rules cited in the findings: PO11 (a KPI measured with no evidence or data source stated), PO13 and PO14 (a deduction with no rectification step; rectification with no period), PO16 to PO18 (persistent-breach machinery not connected to KPI failure; failure straight to default; counts that differ between provisions), PO19 (measurement depends on a system nobody is obliged to maintain), PM4 and PM5 (invoice process incomplete, including submission timing; certification with no certifier or period), PM7 (no treatment of disputed amounts), PM12, PM16 and PM17 (a cap on a different basis; a KPI with no payment effect; a deduction triggered at a different figure). The product itself carries no alert for a late monthly report and files reporting dates as administrative; that gap is the reason the first dimension is here.

What to do with a low level

  • Reactive on verification: choose the KPIs that move the payment, re-measure a sample every period from a source the authority holds, and record the result as verified against the contractor's figure. A reported number is not a measured one.
  • Informal on reporting: put the report day and the validation window in the calendar with an owner, and stop paying an invoice whose report has not been validated. The right to adjust a period that has closed is hard to recover.
  • Rights never exercised: use each of them once, on notice, and notify the outcome. A right unused for years is harder to use when it is needed.
  • Deductions calculated and not applied, or waived without a decision: log every waiver as a decision by a named person with a reason, and reconcile the applied figure to the payment record each period. Check every exclusion the contractor claims against the authority's own record of the cause.
  • Rectification not tracked: give every failure a rectification status and a due date, count repeats per KPI, keep the tally against the default thresholds, and name the person who decides whether a potential default is one.
  • Evidence files patchy: hold the report, the raw measurement, the rectification record, the payment calculation and the correspondence on any disputed line per period, and mark each item received, verified, missing or disputed. The condition data the lifecycle check produces is the handback file.

Reading the contract text

The optional step on the first tab is the same pattern engine as the other tools in the series. Here the targets are the periodic report and its due day, the validation window, self-monitoring and the authority's own monitoring, the increased-monitoring trigger, the quality audit right and its interval, records and helpdesk data, audit access, retention, service failure points and the default limbs, the warning-notice ladder, the annual plan and report, the satisfaction survey, the disputed-amount window, the no-set-off clause, benchmarking and a blank payment schedule. It proposes the weakest option the phrase supports (an audit right "with no interval stated" unless a finalised interval is found; records "partly" unless helpdesk data is named; annual documents "one of them" unless both are named outside the definitions; a survey always "optional", to be confirmed; warning notices only where both notices are found outside the definitions; points "bracketed" unless a finalised threshold is found), shows the option text on the proposal, flags bracketed figures as not finalised, and never sets a number.

Limits

  • The counts are the authority's own and are taken as entered; the tool does not read the registers.
  • The caps are the tool's conventions, listed above; a contract with a different cadence or a different verification standard needs them read with that in mind.
  • Periods are converted to months by the period length entered; the audit interval arithmetic assumes it.
  • The precedent's payment mechanism is blank; the KPI list, points, caps, ratchets and rectification periods are the executed contract's.
  • On a user-pay contract the invoice, disputed-amount and set-off lines apply to an authority-paid stream only, and the deduction lines mean the performance penalties or abatements the contract sets.
  • Nothing here measures the private party's performance, characterises its conduct, or gives a view on the merits of any deduction, exclusion or waiver.

References

  • PPP Contract Review & Operating Map v1.0.0, Benchmark Baseline (2026): the Build 2 performance, payment and evidence registers (KPI record fields; the source, verification, deduction, rectification, breach-stage, cure and evidence statuses; the role capabilities record, validate, approve; the human-only decisions and the prohibited actions, including certifying a disputed payment); the alerts engine (levels, priorities, the reporting gap); the mechanism elements for KPI deductions, persistent breach and the service payment; rules PO1 to PO21 and PM1 to PM17.
  • National Center for Privatization & PPP (Saudi Arabia), Project Agreement Precedent Template, as read by the product: Clauses 21.1, 25, 29, 31, 32.2, 32.5, 32.6, 32.10, 32.11, 36, 40, 48.9, 64, 72; the Contractor Default limbs on service failure points; Schedule 6 (blank).
  • World Bank, PPP Reference Guide, Version 3 (2017), Section 3.6 on managing PPP contracts: monitoring, reporting and the authority's role in the operating phase.
  • HM Treasury, Standardisation of PF2 Contracts (2012), the payment mechanism and monitoring provisions: self-monitoring and the authority's monitoring rights.
  • Infrastructure Pre-Feasibility Toolkit, the readiness pattern used across this series: one status per dimension, worst wins, a named person confirms.