Dext vs Hubdoc: Automating Receipt and Data Entry for Bookkeeping

Data entry is the part of bookkeeping that adds the least value and absorbs the most time. Receipt-capture tools have solved most of it. A photograph of a receipt, an emailed PDF invoice, or a supplier invoice forwarded to a dedicated inbox is read by OCR, parsed into structured data, and published as a bill in cloud accounting software, ready to be matched against the bank feed. For a small business processing a hundred receipts a month, the time saved is hours per week and the audit trail attached to every transaction is materially stronger than anything that could be reproduced by manual filing. This piece is the practical guide to setting up and running that workflow with Dext and Hubdoc, the two market-leading tools.

It sits in the small business bookkeeping best practices hub, alongside the five most costly bookkeeping mistakes and using your chart of accounts for better reporting. The three pieces describe the failure modes, the automation that prevents most of them, and the structure that makes the resulting data useful.

What Dext and Hubdoc actually do

Both tools do the same core job: turn an unstructured document (a photograph, a PDF, an email body, a scanned image) into a structured bill or expense entry inside cloud accounting software. The pipeline is OCR, then field extraction, then a confidence-scored draft that the user reviews and publishes. Once published, the document is attached to the resulting bill, so the source evidence travels with the entry through the ledger for the rest of its life. The user never types the supplier name, the date, the net, the VAT, or the total: the software fills those fields in.

Dext (formerly Receipt Bank) and Hubdoc are both mature products with broadly similar capabilities. Dext is the standalone market leader and integrates with Xero, QuickBooks, Sage, and others. Hubdoc is owned by Xero and is bundled free with most Xero subscriptions; it integrates principally with Xero. The choice between them is usually decided by what is already in the stack: Xero users default to Hubdoc unless they need Dext-specific features; non-Xero users default to Dext.

How the OCR step works in practice

Both tools use a combination of optical character recognition and machine-learning field extraction trained on millions of supplier documents. For a clear PDF invoice from a known supplier the extraction is near-perfect: supplier name, date, net, VAT, total, line items, and even VAT number are recovered without intervention. For a faded receipt photographed in poor light the accuracy drops, and a human review step is needed. In typical small-business use, around 90 to 95 percent of documents publish cleanly; the remaining 5 to 10 percent need a quick correction before publishing.

Both tools learn from corrections. If a supplier is repeatedly miscoded to the wrong expense account and a human corrects it, subsequent invoices from that supplier are coded correctly. The same is true for VAT treatment, project assignment, and any tracking categories the business uses. After three to six months of use the rate of unattended publication climbs noticeably as the tool adapts to the actual chart of accounts in use.

How documents get into the tool

Both tools accept documents through several channels, and the right setup uses more than one:

• Mobile app: photograph a paper receipt at the point of purchase; the app uploads it immediately.
• Dedicated email inbox: a unique address provided by the tool; forwarding supplier invoices to that address pushes them straight into the queue.
• Auto-forwarding rules: configure the business email to auto-forward invoices from known supplier addresses to the inbox.
• Supplier-fetch (Dext): the tool can log into supplier portals (utilities, telecoms, some software vendors) and pull invoices automatically.
• Bulk upload: drop a folder of PDFs onto the web interface for historic catch-up.

A well-configured small business covers paper receipts by mobile app, emailed PDFs by the inbox, recurring supplier invoices by auto-fetch where supported, and forwards everything else manually. The aim is that no document touches a filing cabinet; it goes straight from supplier to the capture tool to the accounting software.

The publishing workflow

Once a document is in the capture tool with extracted fields, the user reviews and publishes. The review step takes seconds for a clean document: check the supplier, the date, the total, the VAT, and the suggested account, and click publish. The bill appears in the accounting software with the document attached. The bank feed brings the matching payment in when it clears, and the reconciliation step matches the bill to the payment in one click.

The full cycle, in a working setup, looks like this:

The total human time, across the whole cycle, is well under a minute per document, compared with several minutes of manual entry, attachment, and reconciliation. Across hundreds of documents a month the saving is dramatic, and the records produced are more complete than the manual version.

Auto-publishing for confident documents

Both tools support an auto-publish rule: documents extracted with high confidence from a known supplier with a stable coding pattern can be published directly into the accounting software without human review. For recurring monthly invoices from the same vendor (rent, utilities, fixed software subscriptions) this removes the review step entirely. The setting is per-supplier; the right approach is to enable it for stable recurring suppliers and leave human review on for everything else.

Handling the exceptions

Most documents publish cleanly; the exceptions are where bookkeeping judgement is needed. Common exception types and how to handle each:

• Illegible receipts: re-photograph in better light, or contact the supplier for a copy invoice.
• Mixed-VAT receipts (some items standard-rated, some zero-rated): split the bill across rate codes before publishing.
• Personal items mingled in a business receipt: split the bill into business and personal portions; post the personal element to the director loan or drawings.
• Tipped or rounded totals where the receipt does not break out VAT: estimate based on the supplier type, or query.
• Foreign-currency invoices: confirm the exchange rate applied, especially when the tool defaults to its own rate rather than the date of supply.
• Duplicate uploads (the same invoice forwarded twice): use the duplicate-detection feature, or delete the duplicate before publishing.

Exception handling is where the human judgement of the bookkeeper still earns its keep. The point of the automation is not to remove judgement but to remove typing, so that the judgement can be applied to the cases where it actually matters.

Setting up Dext or Hubdoc from scratch

A clean initial setup pays back across the life of the system. The steps:

• Connect the capture tool to the accounting software through the supported integration.
• Map the chart of accounts: confirm that every expense account the business uses is visible in the capture tool.
• Set up tracking categories (Xero) or classes (QuickBooks) if the business uses them for project or departmental reporting.
• Configure the unique inbox address and set business email to forward supplier invoices to it.
• Install the mobile app on each phone that will be used for receipt capture; sign in to the same workspace.
• Add known recurring suppliers and confirm their default coding.
• Decide which suppliers, if any, should be auto-published once confidence is established.
• Pick a review cadence (typically daily or twice-weekly) and stick to it so the queue does not build up.

The setup itself is a one-time exercise of perhaps two to four hours for a small business. The ongoing routine, once configured, is the brief review cycle described above, integrated with the monthly reconciliation discipline covered in the sister piece on the bank reconciliation process.

Cost versus saving

Hubdoc is included free with most Xero subscriptions, which makes the cost decision trivial for Xero users. Standalone Dext runs from around £10 per month for a single user up to higher tiers for multi-user or higher-volume use. For a business processing 100 receipts a month, the direct labour saving (assuming the alternative is a bookkeeper charging £30 per hour for manual entry) is several multiples of the subscription cost. The indirect savings, in better VAT recovery and a stronger audit trail, are larger still and harder to quantify because they show up as the absence of problems rather than as visible reductions in cost.

What the automation does not do

It does not replace the bookkeeper or the accountant. It removes typing; it does not remove judgement about how a transaction should be treated, whether a purchase is capital or revenue, or how a return should be prepared. It does not eliminate exceptions; it concentrates the human attention on the exceptions and removes the human attention from the routine. And it does not guarantee accuracy on its own: a misconfigured supplier default, an unreviewed exception, or a bank feed gap can all leave the records wrong even with automation running. The discipline of the monthly close still applies; the automation makes the close less tedious, not unnecessary.

Common setup mistakes to avoid

• Leaving auto-publish on for suppliers whose coding has not stabilised; errors slip through unreviewed.
• Forwarding only some supplier invoices to the inbox while filing others manually; the audit trail splits.
• Failing to map tracking categories at setup so all later entries lack the project or department dimension.
• Not training the team on the mobile app, so receipts continue to accumulate in pockets and wallets.
• Treating the review queue as optional, letting it grow into a backlog that defeats the point of real-time capture.

The bigger picture

Receipt-capture automation is one of the cleanest productivity wins available to a small business. It removes one of the most disliked parts of bookkeeping and replaces it with a workflow that is faster, more accurate, and easier to audit. Set up properly and integrated with the bank feed and the monthly reconciliation, it turns the purchase ledger from a chore into a near-automatic background process. The bookkeeper or owner who used to spend hours on data entry can spend that time on review, reporting, and the parts of bookkeeping that actually require human judgement, which is the right place for human time to land.