OCR
The real cost of manual invoice data entry
The cost of manual invoice data entry appears on no budget line: it has to be calculated. Time per invoice, fully loaded hourly cost, rework rate, monthly volume, then a comparison with the price per page of extraction. No figure is borrowed here: only yours count.
The problem
A real cost that shows up on no line
Nobody invoices data entry. It is done by people already paid, on time already budgeted, and it therefore disappears from the income statement as such. That is what makes it comfortable and impossible to weigh up: an invisible cost is never compared with a visible subscription.
The calculation starts with knowing what a machine can read: what automated extraction reads on an invoice (OCR, optical character recognition, being its first step), and then what it still leaves to be done.
Extraction is not the result. It is the first step. A price read correctly that breaches your contract is still a price paid in error. The calculation below therefore covers only what data entry costs, never what checking might recover: that second figure cannot be invented either.
The terms of the calculation, and where to find each value
No figure is supplied here, and that is deliberate: an average time borrowed from a study does not survive contact with a management accountant. The values below are measured in your own organisation, on your own documents, and the result belongs to you.
- Time per invoice
- Time the complete task, not the typing alone: open the document, retype the header and the lines, look for the purchase order, pick the coding, route it to the approver, file it. Separate the simple invoice from the multi-line invoice and from the invoice with no purchase order: those are three different costs.
- To be measured on a sample of real documents, with the person who processes them.
- Fully loaded hourly cost
- Gross salary, employer contributions, holidays, supervision, workstation: the employer cost of one hour, not the hourly rate on the payslip. Add the data entry provider and the overtime worked at close, where they apply.
- Management accounting or payroll will give you this value for your entity.
- Rework rate
- The share of invoices that pass through human hands a second time: a keying error corrected afterwards, a coding to redo, a credit note to request, a duplicate payment to recover. It is a multiplier on time per invoice, not a separate line.
- Count the adjustment entries and the credit notes requested over a closed quarter.
- Monthly volume
- The number of incoming documents per month, and the number of pages: the two are not the same. Take a representative month, then a peak month, because the cost of manual invoice data entry is paid above all at the peaks, when the backlog forms.
- The purchase ledger, the dedicated inbox, the incoming post register.
- The cost of the tool
- The published price, by volume, with no hidden entry fee and no per-user billing. Take the package that matches your page volume, not the lowest one on the grid, and check what the plan includes.
- The pricing grid, published per page and per monthly package.
The honest calculation then reads: time per invoice × fully loaded hourly cost × monthly volume, increased by the rework rate, compared with the cost of the tool at the same volume. The first three terms belong to you; the last one is published. We fill in none of the boxes on your behalf.
The costs that data entry never shows
- Re-entry after an error: it appears in no budget, because it is done by the same people who did the entry.
- Chasing suppliers over a lost or blocked invoice, and the time it takes on both sides.
- Late payment penalties, when the delay comes from processing and not from cash.
- Early settlement discounts lost, for want of having investigated the invoice in time.
- The absence of an audit trail: a check redone after the fact is taken on a sample, and paid for twice.
None of these costs is quantified here. They are quantified in your organisation, and they all are.
The threshold at which the tool pays for itself
The published price is a price per page: on the package of 500 pages analysed per month, it works out at €0.198 excl. VAT per page.
On the last package of the grid, 5,000 pages per month, the same page works out at €0.150 excl. VAT. The volume discount can be checked on a calculator, which is the least a published price owes you.
The threshold is crossed as soon as the entry cost of one invoice (the one you have just calculated) exceeds the price of the pages it represents. That is a comparison, not a promise: we know neither your fully loaded hourly cost nor your rework rate, and we do not invent them.
Capture
Gather and consolidate all your procurement data.
Collection by dedicated mailbox, upload, SFTP, scan, API or ERP export. Reads every format (native PDF, scan, photo, spreadsheet, structured feed), then extracts line by line: supplier, references, quantities, unit prices, discounts, taxes, terms.
The end of re-entry. A procurement history you can finally query.
What Capture detects
- Long documents and multi-page tables
- Unknown layouts
- Several languages and currencies
- Poor scans and handwriting
The plan covers extraction. Checking your negotiated terms is the next step and is sized with you: it is justified not by the cost of data entry, but by the variances it brings to light.
Nothing is lost. Everything can be checked, everything can be proven.
- The contract clause and the invoice line, highlighted side by side.
- Every extracted value stays linked to the exact place in the document where it was read.
- The same case produces the same decision, today as in six months: the rules are applied deterministically.
- No discrepancy is set aside in silence. Anything that matches no rule is raised, with its reason.
- The agent records what it did, in the order it did it: who, what, how much, when.
Frequently asked questions
How do you calculate the cost of entering one invoice by hand?
Multiply the time of the complete task (open, retype, find the purchase order, code it, route it) by the fully loaded hourly cost of the person doing it, then increase the result by the rework rate. Measure the time on real documents, separating the simple invoice from the multi-line invoice and the invoice with no purchase order.
Is the plan counted in invoices or in pages?
In pages analysed. A single-page invoice counts as one page; a bundle counts as many pages as it contains. Count a real month of incoming post rather than an average, then look at a peak month.
Does Zylio replace my ERP?
No. Your ERP runs the process and remains the source of truth. Zylio handles the exception, on top of it, and feeds the results back. No additional development inside your system.
Does the data need to be perfectly clean?
No. The Capture agent is designed for heterogeneous documents and incomplete reference data; structuring is part of the deliverable.
Measurable impact in every environment
More than 5 million procurement documents analysed
Between 1 and 7% of margin recovered
on the scope analysed
From 15 to 45% of time given back to teams, per FTE
depending on the scope and on data maturity
Zylio fits into your existing ecosystem.
The ERP runs the process. Zylio handles the exception and recovers the value that escapes it: invoices without a purchase order, line-by-line price discrepancies, duplicates and overbilling, off-contract spend.
Your data under high security.
Zylio meets the most demanding standards, and nothing is committed without your approval.
- Certifications
- Hosting
- Encryption
- Access
Read next
- Read a document, then check the priceThe pillar of the cluster: extraction, checking and evidence.
- Traditional OCR or AI OCRWhat one transcribes, what the other understands and applies.
- The fields of a supplier invoiceWhat is actually extracted from the document you key in today.
- Line-by-line extractionAn accurate total can hide ten wrong lines: every item, with its reference, quantity, unit price and discount.
- OCR for accounts payableEvery invoice line is read, the coding is proposed, variances are named before the approval to pay.
See what this looks like on your own data
Twenty minutes, on a spend category of your choosing. We show you what the agents detect, with the evidence behind it.
- No commitment, on your own data
- Result in 3 weeks
- 20 minutes, no sales pitch
- Your data stays hosted in France
- No change of tool or process

