Who pays: what public ERP implementation contracts commit the buyer to

Every document in this set is a public record, retrieved from a published source and linked from the dataset. There are no client or prospect documents here, because the firm has none to use. Each document is read against the same fixed list of eleven checks, in the same order, and each check produces one of five states: it fires; it does not fire because an identifiable protective clause stops it, which we also quote; it does not fire because the topic is simply absent; it is not assessable, meaning we tried and the governing section could not be retrieved; or it was never run against that document at all. No finding is recorded without the clause it rests on. Where our record captured a close paraphrase or a summary of the clause rather than the sentence itself, it is shown below without quotation marks and described as our record of the clause.

The eleven checks are: C1 data migration listed as a buyer responsibility with no supplier conversion hours; C2 integrations counted but not designed; C3 testing not budgeted by the implementer; C4 change management with no deliverable; C5 hypercare ending before the buyer’s first month-end close; C6 assumptions that shift risk silently to the buyer; C7 re-baselining or re-pricing rights for the supplier; C8 a post-design re-estimation right; C9 a no-fit-gap method with test-stage change orders; C10 a quoted fee that is contractually only an estimate billed on time and materials; C11 a single go-live with deferred functionality contractually dropped.

The standing score printed against each document counts C1 to C7 only, so it runs 0 to 7. C8 to C11 are recorded where the document addresses them but never counted. C3 is the only two-tier check: it is counted in the standing score only where it fires at HIGH — where the document assigns test-script writing and execution to the buyer and states no implementer test hours or fee. Where test stages are named but ownership is shared or unquoted and no implementer effort is stated, C3 is recorded at MEDIUM and is not counted. That rule is why two scores moved on 2 October 2026; the severity section below says so plainly.

The denominator convention, which every figure on this page obeys. Seventeen documents were read. Sixteen are supplier-written — nine vendor-direct, written by the software publisher that also implements its own product, and seven integrator-led, written by a third-party systems integrator. The seventeenth is a buyer-drafted model contract, the University of Arkansas System’s own template issued to bidders. It is named separately as a comparator, excluded from every supplier or vendor split and from the firing table, and carries no score at all — a null, not a zero. No count about supplier paper on this page is stated against a denominator of seventeen. Each finding below also carries its own assessable denominator, because a check cannot be assessed against a section nobody could retrieve.

We say “contracts” throughout rather than “statements of work”, because one document in the set is an executed agreement body whose statement-of-work exhibit was never in the file we retrieved.

The median is 4 of 7, across the 16 supplier-written contracts, as of 2 October 2026. Earlier published versions of this analysis printed the median as “between 4 and 5”, against a distribution that still carried the Downers Grove and Sutter Creek documents at 5. Both moved to 4 on 2 October, which is what makes the median a clean 4 rather than a range.

Two scores at the bottom of the table are floors rather than readings. The Citizens Property Insurance contract scores 2 on a partial record: its statement-of-work exhibit, payment schedule and third-party-products exhibit are referenced but were not in the retrieved file, so scope, hours, conversion, integration and testing ownership are all unassessed. The Warnaco contract scores 2 with four of the seven standing checks not assessable, because four exhibits of a very large filing could not be retrieved. Neither number should be read as a clean document. A low score means unread, not clean.

What changed on 2 October 2026. Two documents that this analysis previously counted as C3 HIGH were re-examined and downgraded. The Village of Downers Grove contract (Tyler Technologies) moves to MEDIUM, recorded but not counted in the standing score; its score falls from 5 of 7 to 4 of 7. The City of Sutter Creek contract (Tyler Technologies) moves to not assessable; its score also falls from 5 of 7 to 4 of 7.

Why. Both findings rested on the same ambiguous Tyler data-conversion-verification sentence — the buyer “will be responsible for running reports and testing critical processes to verify the returned Data” — which had already forced the San Rafael contract to not assessable on 30 September. Reading that sentence as proof that the buyer also owned test-script writing and execution outright is more than it supports. Under the testing-ownership test adopted on 22 September 2026, HIGH requires the document to assign test-script writing and execution to the buyer and to state no implementer test hours or fee. Neither of these two documents does the first half. Downers Grove at least names a test database and a UAT line on its project timeline, which is why it reads MEDIUM rather than not assessable. Sutter Creek’s statement-of-work exhibit was readable only as far as its table of contents, so no test stage is named in that record at all. Both were re-examined from the stored record without a new fetch, because both exhibits remain unreadable.

What that does to the numbers. Earlier published versions of this analysis carried a split of 5 HIGH, 8 MEDIUM, 1 non-firing and 2 not assessable. The split is now 3 HIGH, 9 MEDIUM, 1 non-firing and 3 not assessable. On vendor-direct paper the HIGH count halves, from four documents to two. The severity picture is softer than we previously published.

What that does not do to the finding. The direction is unchanged: no contract we can assess states the implementer’s test hours by stage. That underlying finding has read 14 of 15, then 13 of 14, and now 12 of 13, moving each time only because the assessable denominator shrank by one — never because a document was found to state implementer hours after all.

We publish our own corrections because an analysis nobody can audit is a claim, not evidence. Correcting a number in public is the cost of the position, and we would rather pay it here than have a buyer rely on a HIGH we could not support.

This is a sample, not an industry benchmark. Sixteen supplier-written contracts, plus one buyer-drafted comparator. Twelve of the sixteen buyers are public bodies and four are private companies, whose contracts are public only because they were filed as exhibits with the US Securities and Exchange Commission — not because those buyers are public bodies. The documents span January 2001 to October 2025. Nothing here supports a statement about the ERP services market as a whole.

Five of the sixteen supplier-written documents come from one publisher, Tyler Technologies. That is the largest single skew in the set, and it is why the Tyler-specific data-conversion-verification sentence shows up in three separate records above. Read every figure on this page with that concentration in mind.

A low score means unread, not clean. Documents arrive with exhibits missing, truncated by the source, or redacted by the filer before anyone outside saw them: a missing statement-of-work exhibit, a fee schedule cut off mid-section, a confidential-treatment redaction over every hour and dollar figure. In those cases the check is recorded as not assessable, with the reason named against that document, rather than folded into a lower score that looks the same as a document we read in full and found protective. The two lowest-scoring documents in the table above are both partial records.

No trend over time. The two most recently added documents both score below the median, which argues against any story about contracts improving. We do not claim one.

No ranking of vendors, and no cost attached to a flag. This analysis says what a document commits the buyer to. It does not say what any supplier is like to work with, and it cannot tell you what a given flag will cost you, because that depends on your staff, your data and your calendar, none of which are in these documents.

Severity and negotiation calls are the buyer’s. We record what the clause says and what we would ask for. What you accept is a commercial judgement that belongs to you.

Corrections. If a finding here does not match the underlying document, that is a correction we want and will publish. The documents are all linked, one row each, in the open dataset.