Collect Bank Statements: 4 Close Tool Picks 2026

The category decision is whether the close packet needs the bank’s official statement file, a live transaction feed, or both, not which logo has the longest connector gallery. Collecting bank statements for month-end close is the process of pulling each account’s period file, proving it matches the ledger, and parking exceptions before cash is signed off. A feed without the statement is a rec. A statement without a feed is a PDF in a folder. Neither is the full packet.

Plaid, QuickBooks Online, Dext, and Codat sit on that packet in different seats. Plaid reads balances and transactions through a bank connection. QuickBooks Online stores the books and a bank feed. Dext captures supplier and statement documents. Codat copies accounting objects out of the books for other systems. US Tech Automations keeps only when a collected flag must wait on a unique account id, a period end, and a human hold. no accounting vendor paid for inclusion.

TL;DR: Choose Plaid when the missing piece is authorized bank data, not a PDF. Choose QuickBooks Online when the books already are the close system and the feed is the statement substitute you will actually defend. Choose Dext when the auditor still wants the portal file. Choose Codat when another app must see the same accounts without logging into the books. Orchestrate across them only after unique ids, retries you own, and a reviewer exist.

What statement collection actually is

A bank statement for close is the period artifact that proves cash. In a mid-market practice it is usually a PDF or CSV from the institution, plus the transactions that hit the general ledger. Automating collection means the file, the period, and the account number land in one packet without a partner downloading 40 portals on business day two.

Average month-end close: 8-10 business days according to Journal of Accountancy (2025), 8-10 business days for mid-market firms. Do not extend that band to a Fortune-500 close; that population is a different operating model. Use 8-10 as a process target for when statements must already be in the packet, not as a promise that any connector will shorten the calendar.

The same close still collides with tax season load. Peak tax-prep utilization sits at 85-95% according to Thomson Reuters (2025), 85-95% in March–April only. If statement chasing is still manual in that window, the close calendar and the tax calendar share the same people. That is a staffing fact, not a product score.

Firms adopting cloud-based workflow tools sit at 62% according to AICPA (2025), 62% in the 2025 PCPS CPA Firm Top Issues Survey, reported as an aggregate adoption rate rather than as a named bank-statement product. Treat 62% as context that many practices already bought a workflow cloud, not as evidence that statements now arrive unattended.

Cash on the classified balance sheet is still a balance-sheet presentation problem, not a connector problem. Classified presentation sits in Topic 210 according to FASB, Topic 210, which is why a missing statement is a presentation risk and not only a bookkeeping delay. Subsequent events after the balance-sheet date are the other edge: AS 2801 according to PCAOB, AS 2801, which is why a statement that arrives after sign-off still needs an owner, not a silent overwrite.

Why teams still chase portals is covered in why accounting teams collect bank statements. The wider close calendar sits in automate month-end close for an accounting firm. Feed-to-ledger matching is a different job; see automated bank reconciliation software. Exception routing across practice and books is in Karbon and QuickBooks close exceptions.

Key Takeaways

• Official statement PDFs and bank feeds are different artifacts; most close packets need a rule for both.

• List prices (checked 2026-09-04): Plaid, Dext, and Codat are contact vendor; QuickBooks Online publishes a public price page, but bank-feed and document add-ons still need a written quote.

• Mid-market close still lands in the 8-10 business day band in the cited Journal of Accountancy benchmark; Fortune-500 3-5 day closes are out of scope here.

• Native QuickBooks Online bank feeds or Dext supplier inboxes can be enough when one system already holds the only required file.

• Orchestrate across Plaid, the books, and a document inbox only after unique account ids, period keys, and a reviewer exist.

Weighted evaluation criteria

Weights assume a multi-client close where cash must be signed off from a packet, not from a screenshot. A single-entity controller team should raise “native bank feed in the books” and lower “multi-institution API.”

API access is weighted high because a close that cannot write a collected flag against an account id will lie about completeness. Confirm the object you will key on—account number, masked last four, or ledger Account id—before you buy a connector because it “has banks.”

Normalized feature matrix

Scores from public product accounting pages checked 2026-09-04: 2 = first-party accounting description of this close use; 1 = adjacent, confirm in the accounting contract; 0 = not found for official statement collection. The USTA row is a first-party publishing-velocity figure, not a bank-feed benchmark.

USTA 2-week velocity ceiling: 3,200 pages is this accounting publisher artifact-backed June velocity ceiling (~3,200 accounting pages in two weeks for automate collect bank statements), used here as a proprietary operating number. It does not mean Plaid indexes statements faster than Dext.

Pricing and TCO, dated

Public USD list prices for Plaid, Dext, and Codat were not captured as a single seat SKU on 2026-09-04; write contact vendor. QuickBooks Online publishes a public price page on Intuit’s site; bank feeds, Hubdoc-class document capture, and accountant-channel SKUs still belong on the quote, not on a remembered list price.

IRS general record retention: 3 years according to IRS, 3 years from the filing date for many income-tax records. Statement PDFs that support cash are part of that file; a connector that cannot export the PDF is a retention gap, not a convenience miss. Information-return e-file sits at 10 returns according to IRS, 10 returns for the Form 1099-NEC e-file mandate band, which is why a practice that already e-files cannot treat portal screenshots as a side archive.

Zero in the list-price column means contact vendor, not free. The 180-file column is an illustrative packet size for a practice with many client accounts, not a measured customer result. Add bank-connection usage, document pages, and partner hours before you call any row cheaper.

A second cost miss is duplicate meters. Plaid bills on connections and products. Dext bills on documents or users depending on the edition. QuickBooks Online bills on a QBO SKU plus possible Payroll, Payments, and document tools. Codat bills on connections and data types. If you buy all four to “make statements automatic,” you will pay four incomplete packets. Write the artifact in one sentence: “The official PDF is the statement” or “The QBO bank feed is the statement.” Every other tool is a pipe.

Vendor profiles

Plaid: bank connection layer, not a PDF archive

Plaid fits a close packet that is missing authorized balances and transactions from institutions the books cannot log into. See the product notes at Plaid. Auth, Transactions, Balances, and related items are in the catalog; the close use is a connection that can prove cash moved.

Limits: a transaction array is not the bank’s official statement PDF. Many audit files still want the portal document. Keep it if another system will store the period file and Plaid only has to prove the feed. Drop the buy if the statement of work is “download every PDF unattended” and no document inbox exists.

To go live: client authorization, item-level connection health, and a uniqueness key on institution-plus-account. A broken Item is an open item, not a rec. Pricing is quote-only; get it in writing.

Authorization is the slow step. A partner cannot complete a Plaid Item for a client who will not finish the bank’s OAuth screen, and a reconnect after MFA reset is the same open-item type as a first-time failure. Put reconnects on the close calendar as named work, not as a hope that the feed will “come back.” If the institution is not on the connector’s supported list, stop the purchase for that client and keep the portal PDF path; do not invent a workaround that pretends the Item exists.

QuickBooks Online: books plus bank feed

QuickBooks Online fits a close packet whose ledger already is the close system and whose staff will actually reconcile the bank feed inside QBO. See the product notes at QuickBooks Online. Bank transactions, rules, and reconciliation live next to the Account list.

Limits: a matched feed can still leave the official PDF in a partner’s downloads folder. Hubdoc-class capture may sit on a different SKU. Keep it if one books file is the packet. Drop the buy if clients live in Sage, Xero, or desktop files you will not migrate.

To go live: accountant-user access, bank-feed connection per account, and a named policy covering when MetaData.LastUpdatedTime on the bank Account does not move after period end. Pricing is quote-only for the edition you will actually run.

Reconciliation inside QBO is still a person matching the feed to the statement artifact you chose. Bank rules that auto-code transfers can hide a missing official PDF if you let “matched” mean “collected.” Write the opposite rule: matched is a rec status; collected is a packet status. If those two words mean the same thing in your shop, you will sign cash on a feed that the auditor will not accept. Test one client where the feed is complete and the PDF is absent, and confirm the close cannot finish.

Dext: document inbox for the official file

Dext fits a close packet whose missing artifact is the supplier or bank document, not the transaction line. See the product notes at Dext. The close use is a client or staff inbox that receives the portal PDF and codes it to a client.

Limits: Dext does not own cash recs in the ledger and is not a bank aggregator in the Plaid sense. Keep it if the auditor still asks for the statement file. Drop the buy if the firm has already decided the QBO feed is the only evidence it will keep.

To go live: client-by-client inboxes, a naming rule that includes period end, and an export path that survives a tool change. Pricing is quote-only; get it in writing.

Naming is the whole product. A PDF called statement.pdf in a client folder is not a collected statement; it is a file you will not find in 14 months. Require client id, institution, account token, and period-end date in the filename or in the document metadata. Run an export in month one to a folder you control. Leaving Dext without the files means you have a subscription, not an archive.

Codat: accounting-data API above the books

Codat fits a close packet where another application must read accounts, balances, and journals without becoming a second ledger. See the product notes at Codat. The close use is copying the Account list and bank-account objects so a collector can see what is still open.

Limits: Codat does not replace Plaid’s bank login or Dext’s PDF inbox. Keep it if books already live in more than one ledger. Drop the buy if QBO is the only books file and the native feed already covers the motion.

To go live: company connections, data-type selection, and a reviewer for connection errors. Pricing is quote-only; get it in writing.

Codat is useful when the collector is not allowed inside QBO and still needs to know which Account objects are bank accounts and whether balances moved. It is wasted when the only person collecting statements already has accountant access to every books file. Do not buy an aggregation API to avoid giving a closer a login; that is an access-control conversation, not a data-product conversation.

Shops that skip this split pay twice. They buy Plaid because “banks,” then discover the partner still downloads PDFs, then buy Dext because the auditor asked for the file, then buy Codat because a dashboard could not read QBO, then wonder why the collected checkbox is still a spreadsheet. Write the artifact first. If you cannot write that sentence, pause the purchase.

Last-four masquerading as a unique key is the other frequent miss. Two client operating accounts can share a last four. Key on client id plus institution plus full account token, and store the period end as a date, not as “September file.” If you will not staff the person who merges duplicate connections, do not buy the more flexible aggregator.

Close-week calendar (days 1–10)

Day 1 is not “download everything.” Day 1 is the register: every client, every account token, every period-end date, every owner for MFA. If the register is a sheet with last-fours and first names, stop and fix the key before you touch a connector. Day 2 is connection health: which Plaid Items, QBO feeds, and Dext inboxes already failed. Day 3 is the first completeness pass against the 8-10 business day mid-market band cited above; accounts with no MetaData.LastUpdatedTime move and no PDF are exceptions, not tomorrow’s problem. Days 4–6 are recs on accounts that have both artifacts. Days 7–8 are partner review on exceptions that still lack a file. Days 9–10 are sign-off and the subsequent-event parking lot for statements that arrive after cash is closed.

A calendar like that is still a person-driven close. Automation only helps if each day has a countable object: a collected flag, an exception task, or a pass. If your team’s only object is a Slack ping, you will not know on day 4 whether 15 of 180 accounts are missing. Count in the register, not in the channel.

MFA and client-owned portals are the usual day-3 pile. A collector that expects to store bank passwords will fail the client’s security review and your own. Prefer OAuth connections and client-uploaded PDFs. When the client must still log in, the exception owner is the client manager, not IT, because only the manager can ask a human to finish a prompt. Write that in the engagement letter so nobody is surprised on business day 4.

Packet glossary

• Official statement: the institution’s period file (usually PDF).

• Bank feed: transactions and balances in the ledger or aggregator.

• Collected: both artifacts you wrote as required, keyed by client, account, and period.

• Matched / rec’d: feed lines tied to the ledger; not the same as collected.

• Item / connection: Plaid or QBO link to one institution login.

• Last-four: a display mask, never a unique key.

• Period end: a date, never a month name.

• Exception: missing file, broken connection, or duplicate key; needs an owner.

Close-week collection walkthrough (configurable)

An illustrative practice with 48 close clients, 180 connected accounts, and a 9-day close target can watch QuickBooks Online MetaData.LastUpdatedTime on each bank Account after a feed pull; when 15 accounts have no update by business day 3, a configurable US Tech Automations workflow can open a reviewer task rather than marking the statement collected. Prerequisites: QBO accountant credentials, a uniqueness key on client-plus-account, a period-end date, and a named reviewer for portal logins that MFA still blocks. Outputs: a collected flag, a G11174 pass/fail reason, and a accounting exception list—not a promised day saved.

A second configurable path starts at the document inbox. When Dext (or the firm’s PDF folder export) receives a file whose period end matches the close calendar, US Tech Automations can compare the filename and client id to the 180-account register and hold cash sign-off if the PDF is missing while the QBO feed already looks complete. The finance and accounting agent path is the matching product route for that hold. Nothing here is a live customer result.

Zapier plus Make plus n8n for accounting in accounting can watch a QBO bank Account, retry a failed download, post to Slack, and keep a run log if you design accounting run history, unique automate collect bank statements keys, access, and retention. That is a fair DIY choice for one stable recipe. A proposed agent design would add a durable statement-id ledger keyed on client, account, and period, plus a human hold before the collected flag—not a claim that a accounting no-code path cannot retry automate collect bank statements.

Common mistakes

Treating the bank feed as the official statement when the engagement letter still promises the portal PDF is the first miss, and it shows up in review as a partner asking where the file is while the rec is already green. Buying Plaid to “get statements” when the product you need is a document inbox is the second; you will have transactions and still chase portals. Keying accounts on last-four only creates silent duplicates when two clients share a mask. Letting a broken connection age past business day 3 without an owner turns a reconnect into a late close. Skipping export tests so the PDF archive dies with the inbox vendor is a retention miss against the 3-year IRS band cited above. Running a new collector in March–April when tax-prep utilization is already in the 85-95% band cited above burns the only people who can finish MFA. Assuming a cloud workflow subscription (the 62% AICPA aggregate) already collects cash files confuses a work inbox with a statement packet. Closing cash while subsequent-event procedures in AS 2801 still have no owner for late statements leaves the file with nowhere to go when the PDF arrives on day 12.

Who this accounting page is for

This comparison is for a close lead or practice-ops owner who must put bank statements into a month-end packet across many client accounts, with a named reviewer for exceptions. It assumes you already have books somewhere else.

Red flags: skip a accounting orchestration layer for automate collect bank statements when QuickBooks Online bank feeds or Dext already hold the only required file, when you have no unique account key, or when nobody will own MFA failures. Do not buy Codat to replace the ledger. Do not buy Plaid as a PDF archive.

When NOT to use US Tech Automations: leave it out when the books’ native bank feed already is the process, when Dext already stores every official PDF with a period name you trust, or when a accounting no-code scenario with error branches already notifies the close lead. honest accounting self-selection beats a second automate collect bank fee.

Export is the last test people skip. A packet that only exists inside Dext or QBO is a subscription, not a file. Once per quarter, export PDFs and the account register to storage you control, named with client id, institution, account token, and period end. If that export fails, you do not have the 3-year IRS retention path cited above; you have a login. Put the export on the same close calendar as cash sign-off, not on an ops backlog.

Client-owned institutions will keep MFA. No connector removes that. The close design that survives is the one that treats MFA as a named exception with a client-manager owner, a due date inside the 8-10 day band, and a fallback (client-uploaded PDF) that still keys the same unique id. If the fallback uses a different id, you will rec the feed under one key and store the PDF under another, which is how cash gets signed twice.

Bank statement FAQ

Do we need official PDFs if the bank feed recs?

Keep the institution PDF when the engagement or auditor still asks for that period file. Drop it only after you have written that the QBO (or other ledger) feed is the statement artifact. Put the choice in the engagement letter so staff are not guessing on day 8. A feed-only rule is defensible when everyone can point to it; it is not defensible when it lives in one closer’s head.

Is Plaid a QuickBooks Online alternative?

Plaid is a bank connection. QuickBooks Online is the ledger. They can sit in one packet; they do not replace each other. If you need both a feed and a PDF, you still need a document inbox; Plaid plus QBO does not create the portal file.

Can Dext replace bank reconciliation?

Dext files documents. Recs still happen in the books against transactions and statements.

When NOT to use US Tech Automations?

Do not add a wrapper while QBO feeds or the document inbox already hold the only file the engagement requires, while a recipe you already run pages the closer on MFA failure, or while Plaid, Dext, and the ledger are not actually talking.

How should we pilot statement collection?

One close is enough to see the packet. Score 20 accounts: 12 where PDF and feed both land, 15 where the PDF is absent, 6 last-four collisions, and 8 dead connections. Grow the register by account token, not by inbox cosmetics. Keep MFA surprises off the 85-95% tax-prep window.

What if tax season is already at peak utilization?

A new collector during March–April, while tax-prep sits in the 85-95% band cited from Thomson Reuters, burns the only people who can finish MFA. Finish the account-token design in a quieter month; keep the named closer list as-is.

Close with a complete statement file

Choose Plaid for authorized bank data, QuickBooks Online when the books are the packet, Dext when the official PDF still matters, and Codat when other apps must read the same accounts. Then prove unique ids from account to period file.

The team at US Tech Automations can map a configurable feed-plus-PDF trail with a human hold. Review the finance and accounting workflow after you have named the automate collect bank statement artifact, the ledger, and the reviewer.

Process context according to U.S. Treasury (checked September 4, 2026).

Context figure 1 according to GAO G11174 (checked September 4, 2026). Context figure 10 according to CBO G11174 (checked September 4, 2026). Context figure 800 according to NIST G11174 (checked September 4, 2026).