What Stackmark's dataset covers today: 55 systems, 880 module-depth cells, 218 sources, 352 documented deployments. This edition is provisional: its figures follow the dataset until it is published, and then they stop moving for good.
This edition has not been published. Its figures are regenerated on every build and will change as the dataset does, so nothing here should be cited yet. An edition is frozen at publication, not at generation: a snapshot taken while the record is still being assembled would state coverage gaps that were never true of the site, and the gaps section is the last place to be confidently wrong. Once published, the figures stop moving and stay online unchanged.
This edition covers software a finance function runs the business on: enterprise resource planning suites, the accounting systems that sit below them, and the finance-stack products that take over one process: accounts payable, spend, tax compliance. 10 of Stackmark’s categories carry at least one system in this edition.
A system is in the dataset when three conditions hold: it is generally available, not announced; its own documentation states what it does, at a level of detail that lets a module be recorded as core, moderate or partner-delivered; and at least three sources could be retrieved and dated, one of them the vendor’s own. A product that fails the third condition is left out and named in the note that follows the edition, rather than admitted on a thinner basis. Inclusion is not an endorsement and exclusion is not a judgement: both are statements about what could be sourced.
The systems in this edition state a fit from 1 to 100,000 employees and fall across 4 of the four cost bands. The full method is on the methodology page.
Every system in this edition against every module, in the five states the record allows. Rows are alphabetical. There is no total row and no total column, because a count of modules is not a measure of a system: a suite that lists sixteen areas shallowly is not better than a specialist that documents four deeply.
| State | Cells | Share | What it means |
|---|---|---|---|
| core | 297 | 34% | Core: native, documented as a primary module |
| moderate | 270 | 31% | Moderate: native, documented as secondary or lighter scope |
| partner | 27 | 3% | Partner: delivered through a certified add-on or partner product |
| none | 122 | 14% | None: the vendor does not offer it |
| unverified | 164 | 19% | Unverified: not yet checked against a source |
One bar per system, from the lower to the upper bound of the company size its vendor states a fit for, on a logarithmic headcount axis. Tint is the cost band; the glyph marks a system offered on-premise. Rows are sorted by the lower bound, which is a stated fact, and nothing here is ranked.
Nothing, because this is the first edition. There is no previous snapshot to compare with, and a diff against an empty dataset would read as growth rather than as a starting point. From the next edition this section carries the change in every count and names the systems added and removed.
Changes to the underlying data between editions are visible now, per commit, in the dataset changelog.
Stated because a reader is entitled to know what has not been done. Each of these is a gap in our record, not a finding about the market.
164 of 880 depth cells are unverified
That is 19% of the map. An unverified cell means we have not checked that module for that system, which is a different statement from the vendor not offering it. We record the question rather than assert an absence we cannot source.
34 of 55 systems have no documented deployment
Some publish no customer stories at all; several put their customer lists behind a search interface a citation cannot reach, so the companies are named on screen but not at any URL we could cite. Where a vendor’s index credits a customer to a different product in its range, the row was not recorded against this system.
0 of 7 partner directories carry no vendor tier
Where a vendor publishes its own partner tier we record it and sort alphabetically within it. Where we have not yet read the vendor’s partner finder, every firm reads as unranked, an honest blank rather than a tier we assigned.
The localization matrix covers 0 of 55 systems
Country-level statutory support (e-invoicing schemes, statutory reporting, payroll) is the thinnest part of the record. The field exists on every system and is populated for almost none of them, and this edition says so rather than omitting the dimension.
Four things a reader might expect from a document shaped like this one, and why they are absent.
A ranking would need weights, and the weights would be ours rather than yours. A finance team consolidating twelve entities and a job shop scheduling a shop floor are not shopping in the same order of merit. The Coverage Map and the Range give the evidence; the ordering is the reader’s.
We hold no ratings, and a number computed from our own module counts would dress an arbitrary weighting as a measurement. Depth is recorded per module in five states, each traceable to a source and a date.
We have no data on what companies run, and the documented-deployment count is not a proxy for it: a system with more rows has published more case studies, which is a fact about its marketing. Coverage here means coverage of this dataset.
A published list price appears only where the vendor prints one, with the URL and the date it was read. Everything else is a cost band. There are no modelled totals of ownership, because we would be inventing the inputs.
This edition is provisional, so its window is not open yet. It runs for the 21 days before publication, and the exact dates appear here once that date is set. Corrections are made through the public queue and logged, so a change to the record is visible rather than silent.
Vendors correct facts, never placement. There is no placement to correct: rows are alphabetical, no system is highlighted, and the only emphasised element in any figure is the reader’s own headcount. A vendor may tell us a module is core rather than moderate, that a price has changed, or that a size range is misstated, and we will check it against a source and log the outcome either way.
To raise one, use the corrections queue. Anything affecting a figure in a published edition is corrected in the next edition, not in this one: an edition already cited must not change under the reader who cited it.
Every table behind this edition downloads as CSV and JSON from the dataset page, under Creative Commons Attribution 4.0 International (CC BY 4.0). No form, no email, no gate. Reuse the figures with attribution and without cropping the legend or the edition label.
Stackmark (2026). Stackmark ERP and finance software dataset. Retrieved 2026-09-09, from https://stackmark.co/reports/2026-q3/