Independent program assurance is a review that reports to the sponsor, not to the program, and reads the source records rather than the status deck. Budget 1 to 2% of the program for it. That is the rule we recommend, and the reasoning runs both ways: under 1% cannot read the records deeply enough to contradict the systems integrator, and over 2% starts to compete with the program for money and attention. On a $12M program that is $120,000 to $240,000. It pays for itself because it catches drift about two reporting quarters before the color on the status report changes, and a rescue by then costs 30 to 50% of the original budget.
Budget 1 to 2% of the program, and treat both bounds as real
The right number for independent assurance is 1 to 2% of the total program budget, set once and paid whether or not it finds a problem. The lower bound is the one sponsors breach most often. A review funded at a few days of one reviewer's time can read the status deck and interview the program manager, but it cannot pull eight weeks of the defect log, reconcile a mock load against the trial balance, and sort the change-request register by the design document each entry cites. Without that depth, the review ends up restating the SI's own account in a different font. It cannot contradict the program, which is the only thing it is there to do.
The upper bound protects the program. Spend more than 2% and the assurance function grows a standing team, a reporting cadence, and a claim on the same scarce people the build needs. At that point it reads as a parallel program office, and the SI starts managing the reviewer instead of the work. External program management office support already runs $5,000 to $20,000 a month by ERP Research's figures. Assurance sits beside that cost, not inside it, and it stays light on purpose.
Consider a $12M tier 1 program over 24 months. At 1 to 2%, assurance costs $120,000 to $240,000, which buys a qualified reviewer for a few days each quarter plus a deeper read at each major gate. On a $4.5M tier 2 program, the same rule gives $45,000 to $90,000, enough for two or three gate reviews rather than a continuous watch. The percentage holds across sizes because the records to read scale with the program.
Assurance reads the source records, not the RAG status
A RAG status is an opinion about a plan. Independent assurance ignores the color and reads the records underneath it, because that is where a program drifts first. The same seven signs that lead the status report by about two quarters are all visible in documents the program already keeps. The review's whole job is to pull those documents and check them against the thresholds, rather than accept the summary built from them.
Six records carry most of the signal. Each one answers a question the status deck rounds off.
| Source record | What the status deck says | What the record reveals |
|---|---|---|
| Defect log, weekly, 8 weeks | "SIT cycle 2 at 68% execution" | Whether open severity 1 and 2 defects are rising and the closed-to-opened ratio has fallen below 1.0 |
| Mock-load reconciliation | "Data migration 80% complete" | Record load success rate and the variance between the legacy and loaded trial balances |
| Change-request register | "14 change requests in review" | How many change requests cite a signed design document, and which process areas they cluster in |
| SI staffing list | "Resourcing: green" | Turnover against the named roles in the statement of work, including any change of program manager, lead architect or data lead |
| Steering minutes and decision log | "Steering committee held, no escalations" | Whether decisions that belong to the committee are being made one level down |
| Workshop attendance sheets | "Business participation strong" | Named process-owner attendance at design and test sessions, and who signed the designs |
The point is not to recompute percent complete. It is to reach the numbers the percentage hides. "Data migration 80% complete" usually means 80% of the extraction objects are built, which is not the same as 80% of records loaded, and neither is the same as a trial balance the controller will sign. Assurance reads to the signed trial balance. The full list of thresholds, and the reasoning for each, sits in the pillar on the warning signs and is not repeated here.
Bring it in at design sign-off, before cutover, or within 30 days of three signs tripping
Independent assurance earns its fee at decision points, so time it to the three that matter most. Two are planned gates. The third is a trigger you hope never fires.
The first is design sign-off. Sign-off is where fixed-fee risk transfers from the SI to you, and anything reopened afterward becomes a change request the SI is right to charge for. Design is only 10 to 20% of implementation hours by ERP Research's breakdown, so the window is short and easy to compress. A review here checks that the design documents carry business process owners' signatures and not just an IT analyst's, before you accept a design you will live with for fifteen years.
The second is before cutover. This is the go-live decision, and it rests on two records: the mock-load reconciliation and the eight-week defect trend. A review in the weeks before the date reads both and gives the sponsor an independent go or no-go, rather than a date defended by the people being paid to hit it.
The third is reactive. If three or more of the seven signs trip, commission the review within 30 days. The pillar explains why three is the threshold and two quarters is the lead time. The discipline for the sponsor is simple: do not wait for the status report to turn amber, because by then the lead time is spent.
| Trigger | What the review reads first | The decision it informs |
|---|---|---|
| Design sign-off | Change-request register, design signatures, workshop attendance | Whether to accept the design and transfer fixed-fee risk |
| Before cutover | Mock-load reconciliation, 8-week defect trend | Go or no-go on the go-live date |
| Three or more signs tripped | All six records against the thresholds | Whether to reset scope, date or SI contract now |
The reviewer cannot report to the program or depend on the date
Independence is the whole product. The reviewer must not report to the program and must not have a bonus that depends on the go-live date. Three sources meet that test, each with a trade-off, and one obvious candidate fails it.
Internal audit is independent by design, with a reporting line to the audit committee that no program manager can lean on. The trade-off is depth: an audit team that has never read a defect log may need an ERP specialist beside it to judge the records. A finance controller brings the domain the mock-load reconciliation needs, since the signed trial balance is the controller's own work, and sits outside the SI's chain. The trade-off is time, because the controller has a day job the program does not backfill. An independent advisor brings ERP depth and has read these records before. The trade-off is incentive: confirm the advisor is paid for the review alone and does not also sell the rescue the review might recommend.
The SI's own quality review, and the software vendor's equivalent, is useful but not independent. It catches technical defects early and it is worth having. It cannot contradict the date, because the reviewer's firm is paid on the same contract and measured against the same go-live. A quality review inside the SI reports, in the end, to the people whose fee depends on the program looking well. Treat it as a second line of defense, never as the independent one.
| Who runs it | Independent of the date? | ERP depth | Watch for |
|---|---|---|---|
| Internal audit | Yes, reports to the audit committee | Often thin on ERP specifics | Pair with a specialist to read the technical records |
| Finance controller | Yes, outside the SI chain | Strong on the trial balance and benefits | A day job the program will not backfill |
| Independent advisor | Yes, if scoped to review only | Strong, has read these records before | An incentive to sell the rescue it recommends |
| SI or vendor quality review | No, paid on the same contract | Strong | Use as a second line, not the independent one |
The math: 1 to 2% now against 30 to 50% later
Independent assurance is cheap insurance against the most expensive outcome in the program. A rescue costs 30 to 50% of the original budget, by Deliberate Directions' NetSuite guidance. It is a single advisory firm's figure, so treat it as a planning range, but the shape is not in doubt. On the $12M program above, a rescue is $3.6M to $6M. Assurance at 1 to 2% is $120,000 to $240,000. The review pays for itself if it moves even one of those outcomes.
The return is in the lead time. The seven signs appear about two reporting quarters before the status color changes, so a review that reads the records in month 14 can surface a slipping mock load while there are still months to run data cleanup, rather than in the week before cutover when the only choices left are a delay or a bad go-live. Gartner predicts that by 2027 more than 70% of recently implemented ERP initiatives will fall short of their original business case. Most of those programs will go live. Assurance does not stop a program going live. It stops a program going live on a case it can no longer meet, by catching the scope that quietly moved to a phase 2 nobody funded.