Form 990 Functional Expense Allocation From Supplier Invoices

Every 501(c)(3) and 501(c)(4) above the filing threshold has to allocate every operating expense across three functional columns on Form 990 Part IX: Program Services, Management & General, and Fundraising. Joint costs that benefit more than one function may be split only when the activity passes the FASB ASC 958-720-45 Purpose, Audience, and Content (PAC) test; if any one of the three criteria fails, the entire cost is reported as fundraising. Schedule M is required when aggregate noncash contributions exceed $25,000 in the tax year, with works of art, historical treasures, qualified conservation contributions, taxidermy property, and archaeological artifacts triggering reporting at any amount.

Most nonprofit finance teams treat all of that as year-end work. The trial balance gets sliced into functional columns during close; joint-cost allocations get rationalized as the audit approaches; Schedule M gets reconstructed from a year of donation receipts and acknowledgments pulled out of email folders. Whatever defensibility exists is built backwards.

The work that determines whether Part IX populates correctly is actually done at the moment a supplier invoice enters payables. Which functional column the bill lands in, whether it is a candidate for joint-cost split, whether the document arriving from a vendor inbox is a true invoice or a noncash-contribution acknowledgment routed to the wrong queue: each of those calls is made at AP entry, by whoever codes the bill. Form 990 functional expense allocation from supplier invoices is the operational discipline that turns those AP-entry decisions into a Part IX-ready trial balance with an audit trail attached.

That framing matters because functional allocation is not a tax-prep number. It is a public donor-trust signal that watchdogs and state AGs read directly. The BBB Wise Giving Alliance Standards for Charity Accountability set Standard 8, that a charity should spend at least 65% of its total expenses on program activities, and Standard 9, that it should spend no more than 35% of related contributions on fundraising, and require charities to accurately report expenses, including any joint cost allocations, in their financial statements. Those ratios only mean what they should mean if the underlying allocation is honest. Joint-cost allocation under ASC 958-720 is compliance with the PAC test, not a lever to lift the program-ratio number.

The rest of this article walks the AP-entry workflow that produces a Part IX-defensible trial balance: the natural × functional cross-tab every invoice has to land in, the PAC test applied per invoice, the allocation-key documentation captured at receipt rather than reconstructed during fieldwork, the Schedule M routing decision an AP team makes daily, the platform-tier fit honestly assessed, and where AI invoice extraction sits in that flow.

---

The Natural × Functional Cross-Tab Every Invoice Has to Land In

The Statement of Functional Expenses is a grid. Down the left side run the natural expense lines from Form 990 Part IX: salaries and related benefits (lines 5 through 10), professional fees (lines 11a through 11g, covering management, legal, accounting, lobbying, professional fundraising, investment management, and other), information technology (line 14), occupancy (line 16), travel (line 17), conferences, conventions, and meetings (line 19), depreciation, depletion, and amortization (line 22), and the catch-all "other expenses" rows on lines 24a through 24e. Across the top sit three columns: Program Services, Management and General, Fundraising. Every supplier invoice your organization pays has to land in one cell, or more than one cell, of that grid.

The clearest way to operate this in AP is to draw the grid literally. Take a sheet of paper, sketch the natural lines down the left, the three functional columns across the top, and treat the cells as routing destinations for the invoices arriving today. Once the AP team is coding to a grid rather than to a single chart-of-accounts segment, the question at receipt becomes specific: which row, which column or columns, what split percentage. Four worked examples make the routing concrete.

A conference venue invoice for an annual training event your program staff delivers to participating clinicians: occupancy row, Program Services column, single cell. The venue cost is direct to program delivery. There is nothing to split.

The annual audit fee invoice from the CPA firm: professional fees row (line 11c, accounting), Management and General column, single cell. Audit work is a stewardship function; it does not get bled into Program Services on the theory that it indirectly supports the mission. The whole fee codes M&G.

The invoice from the direct-mail vendor for printing and lettershop on the year-end appeal letter: Fundraising column, on whichever natural row the chart of accounts assigns to printing, postage, and mail services (commonly the "other expenses" rows 24a through 24e where there is no natural-expense line dedicated to direct mail). If the appeal piece is purely solicitation, no joint-cost question arises. The whole invoice codes Fundraising. If the same piece carries a substantive program message alongside the donation ask, the next section's PAC analysis applies before any split is made.

A shared utility invoice for a building used 70% for program delivery and 30% for back-office administration: occupancy row, splitting across Program Services (70%) and Management and General (30%). One PDF, two coded lines in the accounting system. The 70/30 figures come from a square-footage allocation key with a tagged floor plan supporting them. At AP entry, the coder enters two lines against the same bill, each tagged with its functional category and its share of the total.

That last example makes a mechanical point worth pausing on. A single supplier invoice produces multiple coded lines whenever it touches more than one function. The sum of the coded lines equals the invoice total. The split percentages are stamped on each line at AP entry and stay with the line through close, period reporting, and Form 990 preparation. They are not derived later from a month-end reclassification spreadsheet.

There is a difference between splitting an invoice across functional columns at line level and splitting a single GL account at month-end via journal entry. Both produce an apparently allocated trial balance. Only the first builds an audit trail tied to the source document. Which of those two patterns your accounting platform actually supports is a real question, and one a later section answers honestly per platform tier.

Joint Costs and the PAC Test, Applied Per Invoice

The cross-tab routing only gets interesting when an invoice describes an activity that genuinely serves more than one function at once. Direct mail with both a substantive program message and a donation ask is the canonical example, and it is the case the joint cost allocation 958-720 rules were written to govern. FASB ASC 958-720-45 sets the gating test: a joint activity that combines fundraising with another function may have its costs split across functions only when all three of Purpose, Audience, and Content are satisfied. If any single criterion fails, the entire cost of the activity is reported as fundraising. The PAC test joint cost allocation rule is a compliance gate, not an optimization tool, and that is the disposition the AP coder needs going in.

Three worked examples make the test concrete on a per-invoice basis.

Take a direct-mail piece from a public-health charity that pairs an educational campaign on early-warning signs of a cardiovascular condition with a request for donations to fund a screening program. The invoice from the mail vendor lands in AP. Apply Purpose: would the organization undertake this activity even without the fundraising element, because the educational message is itself a programmatic accomplishment the charity exists to deliver? If yes, Purpose passes. Apply Audience: was the recipient list selected for reasons other than donor capacity, such as ZIP-code-targeted populations with elevated incidence of the condition? If yes, Audience passes. Apply Content: does the educational portion of the piece motivate the recipient to take some action consistent with the program's purpose, such as scheduling a screening or contacting their physician, and does it meet the substantive criteria the program would set if the piece were standalone? If yes, Content passes. With all three satisfied, the cost may be allocated across Program Services and Fundraising on a defensible allocation key. The mailpiece, the audience-selection memo, and the column-inch counts that produce the percentage become the evidence record.

Take a telemarketing campaign where the call script runs roughly 90% solicitation and devotes 30 seconds to a "thank you for your past support, here's what we accomplished this year" program update. The invoice from the call center arrives. Purpose may pass and Audience may pass, but Content fails: the program component does not call the recipient to programmatic action and is incidental to a script whose dominant purpose is to ask for money. The full cost of the campaign is reported as Fundraising. There is no defensible split.

Take a targeted mailing sent only to lapsed major donors, even if the piece itself carries strong programmatic content. The list was selected on the basis of donor capacity, which fails Audience. The full cost is Fundraising regardless of what the piece says. List criteria can move an entire activity into Fundraising on their own, and the AP coder needs to ask the audience-selection question at receipt rather than after the auditor does.

The discipline this implies at AP entry is small but consequential. When an invoice for a joint activity arrives, the coder attaches a brief PAC-evidence packet to the bill record before posting: a copy of the mailpiece or the call script, a one-line note on how the audience list was selected, and the planned allocation key if the activity passes. The PAC determination is made at receipt, not in the close cycle, and the supporting evidence is captured contemporaneously rather than reconstructed when the auditor asks for it.

The compliance bar here is real. The California Attorney General's "Abuse of Joint Cost Allocation" publication treats inflated program-ratio claims that rest on aggressive PAC interpretation as a state-level enforcement matter. State AG charity registries in New York, Massachusetts, and elsewhere are paying similar attention. An organization whose joint-cost split makes its program ratio look better than its peers should be able to walk an examiner through Purpose, Audience, and Content for each joint activity, with evidence, without flinching. If it cannot, the conservative interpretation is the only defensible one: allocate fully to Fundraising, accept the lower program ratio, keep the trust signal honest.

Allocation Keys: Build the Audit Trail at AP Entry, Not at Year-End

The PAC determination governs whether a joint cost can be split. The allocation key governs how it is split, and the disclosure of that methodology is itself a financial-statement obligation. FASB ASC 958-720-50-2 requires the entity to disclose the types of joint activities for which costs have been allocated, the methodology used, the total amount allocated for the period, and the portion attributed to each function. Disclosure means the supporting evidence has to exist in the books at the time the activity occurred, not be back-filled in a working-paper schedule when the auditor's request list arrives.

Four nonprofit shared cost allocation key types cover most of what an AP team encounters.

Column-inch (or square-inch) allocation for printed materials. The supporting evidence is the mailpiece itself, with a layout marked up to identify which content is programmatic, which is fundraising, and which is shared overhead like masthead and address block. The column-inch counts produce a percentage; the percentage produces the split. Save the marked-up layout to the invoice record. When the auditor asks how the 60/40 split between Program and Fundraising was derived for the spring appeal, the answer is a PDF, not a memory.

Time-allocation logs for shared staff and shared events. The cost driver here is hours worked by function. The evidence requirement is contemporaneous time records: a weekly or monthly log captured by the staff member during the period, not a year-end estimate produced from a calendar review. For salary expenses, contracted-hours invoices from outside consultants, and event-staffing invoices that span functions, the time log is the authoritative record. Estimates produced after the fact carry far less weight when the auditor asks how the percentage was derived.

Square-footage allocation for shared facilities. The cost driver is occupied square footage by function. The evidence is a current floor plan with each room tagged Program Services, Management & General, or Fundraising, plus a calculation of total occupied square footage and the function-by-function percentages that result. The floor plan needs a date and a revision marker because programs move offices, and a key produced from a five-year-old layout will not match the current footprint. Attach the current plan to the occupancy invoices that ride on it.

Headcount or full-time-equivalent allocation for shared back-office services such as HR, IT, and finance. The cost driver is staff served, not the work the back-office team performs. The evidence is a roster reconciling each headcount to a function, with the percentage derived from the totals. This key works for shared-service invoices like an external HR retainer or an IT support contract that benefits the organization as a whole, where the cost driver is reasonably proxied by who is being served.

What changes at AP entry, in operational terms, is small. The AP coder attaches the evidence file (the marked-up mailpiece, the time log export, the floor plan, the FTE roster) to the bill record at receipt and stamps the allocation percentage on each split line in the accounting system. The methodology is documented before period-close. The audit trail consists of artifacts captured contemporaneously rather than spreadsheets reconstructed during fieldwork. When the auditor opens a sample of joint-cost invoices in the working papers, every sampled bill carries its own evidence record.

A brief note on the federal-grant intersection. Nonprofits that receive federal funding directly or as subrecipients are subject to 2 CFR 200 (Uniform Guidance), which layers separate cost-allocation requirements on top of functional reporting: direct cost versus indirect cost classification, federal cost principles, the negotiated indirect cost rate or the de minimis rate, and the cost allocation plan that supports it. Functional allocation for Form 990 and indirect-cost allocation for federal awards are related but governed by different rule sets, and the same invoice can require both treatments. The grant-side mechanics, including documentation thresholds, allowable-cost determinations, and rate negotiation, are outside the scope of this article; readers running federal awards should pair this workflow with the 2 CFR 200 federal grant invoice documentation walkthrough for the Uniform Guidance dimension.

---

Schedule M and the AP Inbox: Distinguishing Donations From Invoices

Schedule M is treated by most nonprofits as a tax-prep concern surfaced at year-end. That treatment quietly moves the routing decision into a place where it can no longer be made well. Form 990 schedule m noncash contributions reporting depends on documents that arrive in the same payables inbox as supplier invoices, look superficially similar to supplier invoices, and require a different journal entry. The decision happens at receipt, or it happens badly later.

Two trigger rules drive when Schedule M is filed at all. Schedule M is required when aggregate noncash contributions exceed $25,000 in the tax year. Independent of the dollar threshold, certain categories trigger reporting at any amount: works of art, historical treasures, qualified conservation contributions, taxidermy property, and archaeological artifacts. The AP team needs running visibility into both during the year, because waiting until year-end to discover an organization crossed the $25,000 mark, or accepted a single piece of donated art, leaves no time to rebuild the documentation backwards.

The harder operational problem is telling a noncash-contribution acknowledgment from a true supplier invoice when both arrive in the same queue. A pro-forma "invoice" marked "no charge" from a vendor donating goods. A formal donation letter that itemizes contributed equipment with fair-value figures. A vendor receipt that looks indistinguishable from a paid bill. They all hit the AP inbox; they all carry totals and line-item detail. Three observable signals separate them.

Header text and document title. Documents titled "donation receipt", "in-kind acknowledgment", "no charge", "complimentary", or "donor receipt for tax purposes" are noncash contributions regardless of how the line items look. Documents titled "invoice", "tax invoice", "bill", or "statement" with a vendor name and an invoice number behave as supplier invoices. The title is the first filter and is right most of the time.

Payable terms. True invoices carry payment due dates, payment terms (Net 15, Net 30, EOM), and remittance instructions or wire details. Donation acknowledgments do not, because there is nothing to remit. A document with line items and totals but no due date and no remittance information is almost certainly a contribution acknowledgment.

GL coding implication. The journal pattern is the test of last resort. Recording an in-kind donation against a nonprofit invoice debits the expense or asset that the donated good represents and credits gift-in-kind income. A supplier invoice debits the expense and credits accounts payable. If the AP team cannot answer "what does the credit side look like?" without thinking about it, the document is being routed too quickly.

Fair value at receipt is the other piece this section has to surface. FASB ASC 958-605-30-11 measures noncash contributions at fair value at the time of donation, and the basis for the valuation has to be captured at receipt. The basis varies with the asset: comparable market value for routine goods, a donor-supplied invoice or purchase receipt where the donor is the original purchaser, a professional appraisal for high-value items, art, or property where appraisal is required by IRS rules. The valuation memo, however brief, is attached to the contribution record at receipt. Reconstructing fair-value basis at year-end produces working papers that auditors discount and that examiners can poke holes in.

A clarifying note on what is not Schedule M. Donated services and donated use of facilities do not appear on Schedule M, regardless of estimated value. They appear in the Form 990 Part III Line 4 program-accomplishment narrative as part of the description of how the organization advanced its exempt purpose. AP teams sometimes route a free-rent letter from a landlord, or a memo from a law firm describing pro bono work, into Schedule M tracking. Both belong in the program-accomplishment narrative the program team drafts, not on Schedule M. Get this right at receipt and the year-end filing split between Schedule M items and Part III narrative items becomes mechanical rather than investigative.

Platform Fit by Tier: What Your Accounting System Can Actually Capture

The AP-entry workflow described so far assumes the accounting system can carry the codes the AP coder is stamping on each line. Whether it actually can is a real question with a different answer at each platform tier, and the honest answer matters because it changes how much discipline AP entry has to absorb.

QuickBooks Online with Class tracking and Custom Fields is the workhorse at the small end of the sector. The configuration most nonprofit consultants recommend creates one Class per functional category, Program Services / Management & General / Fundraising, with parent-child Classes carrying sub-program detail beneath the Program Services parent (one child per major program). Custom Fields can hold allocation-key data such as the column-inch percentage on a direct-mail invoice or the square-footage percentage supporting a shared-occupancy split. For a typical small-organization invoice that lands cleanly in one functional category, Class tracking is adequate. For 501c3 bookkeeper functional cost coding at this scale, QBO works and is what most QuickBooks-trained sector bookkeepers expect.

The honest limitation surfaces at line level. QBO Class is most natural at the header level on bills; tagging different Classes to different lines on the same bill is supported, but the entire interface and reporting model expects Class to behave like a transaction-level dimension on a system not built around dimensions. For an invoice that splits across two functional categories, the AP team has to either enter the bill as multiple lines and tag each line with its Class, or post the bill at header level and add a manual journal entry at month-end to reclassify a portion. Both work. Neither is what a multi-dimensional accounting system would do natively. QBO is not nonprofit-shaped; Class is repurposed to do work the platform was not designed for, and the workflow cost is paid in AP discipline rather than platform capability. The QBO and Xero side of this workflow has its own treatment in our cross-cutting article on those platforms; here, the fit note is enough.

Aplos sits a tier up and is shaped specifically for nonprofit accounting. Funds and functional categorization are native dimensions, not repurposed Classes, and the chart of accounts is structured around the Form 990 categories from the outset. Line-level functional splits are first-class features rather than workarounds. The fit cost is platform migration: moving from QBO mid-year is disruptive, and Aplos is not interchangeable with general-purpose small-business accounting tools. The fit benefit is that allocation logic stops fighting the data model. If an organization's joint-cost activity is light and its invoices are mostly single-function, the platform upgrade may not pay back. If joint-cost splits, allocation-key documentation, and per-program reporting are routine, Aplos's nonprofit shape makes the workflow noticeably less effortful.

Sage Intacct, and similar platforms in the same tier such as Blackbaud Financial Edge NXT, run true multi-dimensional accounting. Fund, Program, Department, Class, and Location can each be tagged on every transaction line, independently. Joint-cost splits, allocation-key data, sub-program classification, and grant-restricted-fund tagging all live in dimensions rather than account-segment hacks. At $5m to $50m revenue scale with federal funding, multi-program complexity, or multiple physical locations, the platform tier is where Form 990 Part IX populates from dimensional pivots rather than from year-end reclassification journal entries. The fit cost is real: implementation time, licensing expense, and the depth of accounting knowledge required to use the dimensional model well. The fit benefit at the right scale is that the AP-entry workflow this article describes is what the platform is built to support, rather than something the operator has to engineer around the platform's gaps.

The selection question, then, is not "which platform is best" but "what is the right tier given how complex our functional and joint-cost work actually is." A 501(c)(3) on QBO can produce Part IX-defensible AP coding; it just costs more discipline at AP entry to compensate for what the platform does not natively support. A 501(c)(3) on Aplos or Sage Intacct shifts the cost from operator discipline to platform tier and gets line-level dimensional splits in return.

At the small end, volunteer-staffed organizations have a different shape entirely. A congregation with one part-time bookkeeper and a volunteer treasurer is not running joint-cost activities, has limited noncash-contribution complexity, and processes a small number of invoices per month. The right operational model there is not the workflow this article describes; it is a leaner pattern adapted to the documents and decisions a small-org treasurer actually faces. The church invoice processing workflow for volunteer treasurers walkthrough covers that pattern for organizations whose Form 990 obligations are simpler.

Where AI Invoice Extraction Fits in This Workflow

The AP-entry workflow has two layers, and the distinction is the difference between automation that helps and automation that overreaches. The data layer captures supplier identity, invoice number, dates, totals, line items, and PO references: the structured fields that need to enter the accounting system from a PDF. The policy layer applies functional categorization, allocation keys, joint-cost determinations, and Schedule M routing decisions. AI invoice extraction is the right tool for the data layer. The policy layer is partly automatable through vendor-default routing but ultimately governed by the AP team and the written allocation policy that team is coding against.

The vendor-default mapping pattern is what makes the partial automation defensible. After a year of operating, a nonprofit's vendor file has identifiable patterns. The annual audit firm always codes Management & General, professional fees. The direct-mail vendor codes Fundraising on whichever line the chart of accounts assigns to printing and mail services. The conference venue used for the annual program-delivery training codes Program Services, occupancy. The IT support vendor codes back-office headcount-allocated. Roughly 70% of routine invoice volume, the recurring vendor base, can route to a default function from a vendor-keyed lookup applied during invoice intake. The remaining 30% are the line-level split candidates, the joint-cost candidates, and the new vendors that need human review on first invoice. The vendor-default rule does not replace the AP coder; it routes the easy invoices to a queue the coder can approve in seconds and surfaces the harder ones for the attention they actually need.

This is where AI invoice data extraction for nonprofit AP fits as a building block in the workflow, not as a replacement for it. The extraction layer reads supplier and line-item data from a PDF into a structured spreadsheet, and that structured output is what the accounting platform's import flow expects. The interaction model is a single prompt and a file upload: a controller describes what to extract, including any nonprofit-specific columns such as a Function Default tag populated from a vendor lookup, and downloads the result as Excel, CSV, or JSON. The kind of prompt a controller would save to a prompt library and reuse each month names the standard invoice header and line-item fields, adds the Function Default column with the recurring-vendor mapping (audit firm to M&G, direct-mail vendor to Fundraising, conference venue to Program), and leaves Function Default blank for unrecognized vendors so they fall to the AP review queue. Document filtering takes care of email cover sheets, remittance advice, and summary pages mixed into the same upload, so the input batch can be the raw payables PDFs without preprocessing. For the broader picture of how extraction tooling fits the nonprofit AP context, including treasurer-scale and volunteer-staffed cases, our walkthrough on AI invoice extraction for nonprofits covers the operational patterns at different organization sizes.

What AI extraction does not decide bears stating directly. PAC determination is a policy judgment that depends on reading the mailpiece, evaluating audience selection, and applying ASC 958-720-45 to the activity in question; that decision is not extractable from an invoice PDF. Allocation-key percentages are evidentiary and depend on the marked-up layout, the time log, the floor plan, the FTE roster; the percentage exists in the supporting evidence, not on the invoice. Schedule M discrimination can be heuristically pre-routed by document title and payable-terms signals, but the final call about whether a "no charge" pro-forma is a noncash contribution or a damaged-goods credit is a judgment the AP team owns.

The shape of the right answer is layered. Extraction surfaces clean line-item data and the vendor-default function code. The accounting platform carries the line-level Class or dimensional tags. The AP coder applies policy on the exceptions, attaches evidence at receipt, and posts. Form 990 Part IX populates from the trial balance the workflow produces, not from a year-end reclassification project.

---

Common Pitfalls and the Broader Regulatory Landscape

Five mistakes account for most of the audit findings and donor-trust damage that come out of functional-expense allocation work. They are worth naming plainly, because each one is the negation of something earlier in this article.

Re-coding functional allocation at year-end without contemporaneous evidence. The audit profession is increasingly explicit about the evidence it expects: the actual mailpiece for a column-inch allocation, the time logs for a salary split, the dated floor plan for an occupancy split. Reconstructing those records during fieldwork from staff memory and current documents is a finding waiting to be written. The fix is to attach the underlying evidence at AP entry rather than try to assemble it after the fact.

Treating allocation keys as set-and-forget. A square-footage key calibrated to the program footprint of three years ago does not represent the current allocation honestly when the program team has moved offices twice in the interim. A column-inch key set against last year's appeal piece does not apply to this year's piece, because the layout is different. Allocation methodologies are reviewed at least annually and updated whenever the underlying activity changes. The methodology disclosure under ASC 958-720-50-2 is a representation about the period reported, not about a key set five years ago.

Conflating donated services or donated use of facilities with Schedule M items. Pro bono legal work, donated use of a meeting room, and a free-rent letter from a landlord all describe value the organization received, but none of them belong on Schedule M. They appear in the Form 990 Part III Line 4 program-accomplishment narrative because they are not noncash contributions reportable as in-kind donations under the rules Schedule M governs. Routing them to the wrong place inflates Schedule M reporting and depletes the program-accomplishment narrative.

Letting Class become a coding habit rather than a coding policy. AP teams that have been coding the same way for three years will keep coding that way without thinking about it, even when the underlying activities have shifted. The fix is a written functional-expense allocation policy that the AP team codes against: vendor defaults, exception triggers, joint-activity routing, allocation-key methodology by key type, and the review cadence that keeps the keys current. The policy lives in writing; coding is an act of compliance with the policy, not a continuation of last year's pattern memory.

Inflating program-ratio metrics through aggressive PAC interpretation. State AGs read these ratios. Watchdog organizations read them. Major donors and grantmakers read them. An organization whose program ratio is two or three percentage points higher than its sector peers because it stretches the PAC test on direct mail will eventually have to defend the difference, and the defense will be harder than the original conservative coding would have been. The compliance posture is conservative interpretation. If a joint activity is borderline, the trust-preserving call is to allocate fully to Fundraising and accept the lower ratio.

Equivalent regimes in other jurisdictions follow similar logic with different forms and thresholds. UK-registered charities reporting under the Charities SORP split between charitable activities, raising funds, and governance costs, with SORP 2026 shifting some specifics; that workflow is covered in our walkthrough on UK charity invoice automation under SORP 2026. Australian charities reporting to the ACNC have their own program-versus-administration framework, covered in ACNC supplier invoice record keeping for Australian charities. The underlying instinct travels: functional allocation is a public trust signal that depends on contemporaneous evidence, and the work that produces it is done at the moment a supplier invoice enters payables.