A business-case method using invented inputs—not a price quotation, market benchmark or forecast of average savings.

A low price per invoice can coexist with an expensive implementation. An expensive subscription can also produce a good return if it removes enough unnecessary work. Neither conclusion can be reached from the subscription alone. The right comparison is between complete processes: what the company does today, what it will do after implementation, and which costs actually change.

I support e-invoicing because structured information creates a practical opportunity to reduce retyping and improve reconciliation. I oppose presenting that opportunity as a guaranteed saving for every business. A credible reform should acknowledge transition costs, recurring obligations and uneven benefits. It should also distinguish a genuinely lower cost from work merely shifted to an accountant, software supplier or public budget.

The saving comes from processing, not the label

Directive 2014/55/EU, which concerns electronic invoicing in public procurement, defines an electronic invoice as one issued, transmitted and received in a structured format allowing automatic electronic processing.14 That definition explains the mechanism behind potential savings. It is not evidence that a particular Slovak business-to-business implementation will achieve them, and the directive’s public-procurement scope should not be silently extended.

The Peppol BIS Billing documentation provides invoice and credit-note structures, validation rules, code lists and downloadable validation resources.5 Such common resources make it possible to build a more consistent exchange process. But a standardised document does not, by itself, clean a company’s supplier records, settle a disputed purchase or redesign an approval workflow.

For budgeting, distinguish transmission from automation. A company that manually exports a file, uploads it to a portal and then re-enters the status elsewhere may have complied with a new format without removing much work. Conversely, a well-integrated receiving process may save effort even when outgoing invoice volume is modest. Count the activities affected rather than assuming the sender receives every benefit.

Build the cost model around the business process

I would divide the budget into implementation, recurring operation, exceptions and exit. Implementation includes requirements work, software changes, data cleanup, testing, training and the period of parallel operation. These are categories to investigate, not claims that every business needs a new system or will be charged for every item.

Recurring operation includes the selected service plan, any relevant software maintenance, internal administration, support arrangements and the handling of retained records. Clarify what the provider includes and what remains with the customer. Do not assume every small business must personally manage network certificates; that depends on the service design and the division of duties.

Exception costs deserve their own line. A rejected invoice may require a person to investigate, contact another party, correct data and verify the outcome. An apparently inexpensive service with poor error explanations may consume more staff time than a better-supported alternative. That is a hypothesis to measure during a pilot, not a finding about any named vendor.

Exit costs include export, validation of retained evidence, changing integrations, staff adjustment and an overlap period if needed. Ask about these before signing. A low introductory price is less informative if future access to historical documents, acknowledgements or account data is unclear. A company should know what it can take with it and in what usable form.

Invented inputs—not an estimate of your price

Consider a fictional business processing 200 relevant invoices each month, or 2,400 a year. Assume the new workflow saves a net four minutes per invoice after ordinary exception handling. That produces 160 hours of annual capacity. At an assumed internal value of €20 an hour, the capacity has a modelled value of €3,200. These are illustrative inputs, not survey results, provider quotations or a forecast for Slovakia.

For this model, €20 per hour means an assumed opportunity value of internal capacity, not a measured wage cost or an accountant’s fee. Supplier prices are modelled excluding VAT, with VAT, deductibility and other tax effects outside the calculation; no universal recovery of VAT is assumed. A real cash budget must include any non-recoverable tax. The €25 subscription has no specified support allowance, archive, integration or service-level commitment, so it is not a recommended market price.

Now assume €1,200 in initial implementation costs, a €25 monthly service charge, €300 in annual additional software and support costs, and €200 in annual internal administration costs. Recurring cost is €800 and first-year cost is €2,000. Against the assumed €3,200 capacity value, the model shows a first-year net benefit of €1,200 and a subsequent annual net benefit of €2,400 before any omitted items or changes.

The most important word is “assumed.” If net time saved is only one minute per invoice, capacity value falls to €800. Under the same cost assumptions, it merely matches recurring cost and does not recover implementation expenditure. The result turns on workflow performance, not on whether the monthly fee sounds small. Sensitivity should also vary support cost and integration effort. The example assumes no exit in the illustrated years: price a switch in a separate scenario with an explicit year and overlap cost, rather than silently charging a full migration every year. Use the same horizon for public and commercial alternatives.

This profile is not a representative business. Measure low-volume recipients, firms processing manually and firms with existing EDI separately. The model shows which inputs drive the outcome, not their Slovak distribution or average benefit.

Capacity value is also not necessarily cash in the bank. If nobody works fewer paid hours and no external bill falls, the immediate benefit is time available for other work. That can be valuable, but it should not be counted again as a separate cash saving. A sound business case distinguishes released capacity, avoided future hiring, reduced external charges and actual expenditure reductions.

Measure exceptions before promising a return

A pilot should record the baseline time spent entering, checking, approving, correcting and reconciling invoices. Then measure the same activities under the proposed workflow. Include incoming and outgoing documents separately. Different teams may bear the cost and receive the benefit, so an overall gain can still leave one department with more work.

Record the proportion and causes of exceptions, but do not infer a stable annual rate from a convenient handful of uncomplicated documents. Include credit notes, multiple tax categories where relevant, missing references and recipient changes in an authorised test set. Use synthetic or appropriately protected data and record what the sample does not represent.

Avoid double counting. If the four-minute net saving already includes ordinary exception handling, do not subtract that handling again. If an integration charge includes training, do not add the same training as a separate supplier invoice. Conversely, employee time spent attending training remains a distinct resource cost if the model is designed to include it.

Compare services at the same boundary

One quotation may cover transport only; another may include an accounting connector, storage, support and document export. They are not directly comparable until the missing functions are priced or consciously excluded. The comparison should also separate optional commercial convenience from what is necessary to satisfy the applicable obligation.

Security and resilience have costs, but “premium security” should not become an excuse for an undefined baseline. For processing personal data, Article 32 requires risk-appropriate measures and explicitly takes implementation costs into account alongside the other specified factors.10 It does not mean that choosing a cheap plan removes the relevant duties.

For each quotation, ask what happens during an outage, what support includes, how pricing changes with volume and which functions remain accessible after termination. Obtain the conditions in writing. The objective is not to select the provider with the longest list of features but the service that meets the actual need at a defensible total cost.

Public money is still money

My policy preference is a direct public baseline channel with optional paid integration and value-added services. This could reduce the need for every small business to negotiate a separate commercial relationship simply to perform a mandatory task. It is a proposal, not evidence that a public channel would be cheaper in aggregate.

Changing the payer is not a saving in itself. Business fees measure what firms pay, not the total cost of providing the service. Government expenditure measures a budget burden, not all real resource use: staff time, infrastructure and work retained by businesses also matter. If a public contract replaces subscriptions, provider revenue may move from many business customers to one public purchaser. That transfer is not, by itself, a reduction or increase in societal cost. Keep payments and distributional effects separate from resource costs; do not count the same purchased work once as a fee and again as an additional societal cost.

The strongest counterargument is that a state-funded service may hide costs in taxation, compete unfairly or deliver a weak user experience. Those risks are real design questions. Compare startup procurement, development, integration and migration with lifecycle operation, support, security, upgrades, continuity and eventual operator replacement, including handover and parallel running. An initial development budget is not a lifecycle budget. Public procurement should preserve fair competition through comparable requirements, usable interfaces and a tested exit, rather than guarantee one supplier a permanent position. These are conditions of my proposal, not claims about an existing tender.

The same scrutiny belongs on the commercial model. Competition can encourage better service, but mandatory participation does not guarantee that switching is easy or that small customers can negotiate effectively. Compare the alternatives using consistent functions, service quality and time horizons. Publish assumptions and distributional effects, including which businesses pay more even if the aggregate balance is positive. The public variant must justify its total costs and purpose; the commercial variant cannot claim superiority merely because its costs sit outside the government budget.

Decide with evidence, then check the outcome

Before committing, write a short baseline, obtain comparable quotations, identify the process changes and test the assumptions that most affect the result. Assign ownership for collecting time and exception data. After implementation, compare actual effort and charges with the business case, and adjust the process rather than defending an outdated promise.

E-invoicing can be worth paying for. That argument becomes stronger when it admits which benefits are uncertain and which costs are unavoidable. The honest question is not “Is it free?” or “Will it ruin small firms?” It is: which work disappears, which work remains, who pays for the transition, and what verified improvement does the business receive in return?

Sources

[5] Peppol BIS Billing 3.0 - May 2026 Release

[10] GDPR adopted text, Article 32 (official reproduction)

[14] Directive 2014/55/EU, Article 2, adopted text (official reproduction)