How to choose a software development partner: the same six written artefacts for every quote

Who is publishing this, and what to discount

This blog describes itself as "Field notes on AI automation, custom software, recruiting advertising, and lead generation — written from the work we ship for clients, not from a content calendar." The byline is the organisation Niro Digital, not a named person. Niro Digital is NiroDigital d.o.o., registered in Mursko Središće, Croatia, and serving primarily the Slovenian market — a Slovenian-market agency registered in Croatia.

That matters for reading the worked example later: the framework is being demonstrated on pages published by a party that has a commercial interest in the result. The pages are useful as test material, not as neutral procurement guidance. Where a published page does not answer a question, the honest treatment is to say so in the grid rather than fill the cell with an inference.

Three quotes are not comparable until the same six artefacts are in writing

The three prices cannot be treated as a price difference until each quote has been checked against the same written artefacts; this article has not seen your quotes and does not assume that their scopes differ. A quote silent on code ownership and handover is not offering the same thing as one that addresses them.

Each of the six artefacts has one defined evidence test: the specific document or written answer that satisfies it. A verbal assurance in a meeting, a values page, a case study, or an answer to a question the buyer did not ask in writing does not satisfy the test. The scorecard below is blank on purpose. The weight column is yours to set.

Record Evidence status as 1 where the required written evidence is supplied, and 0 for silence or refusal; carry that value into Score. Calculate weight × score for each row, total the six weighted scores for each supplier, and record an equal total between two suppliers as a tie that requires a documented tie-break decision rather than an automatic winner.

Scores produced with this scorecard are professional judgements against a disclosed framework, not measured performance results. The scorecard evaluates documents, terms and evidence. It does not measure engineering quality.

Now set the weights from your own operations, not from this article's preference. For a scheduling, field-reporting and invoicing system covering 22 technicians, the artefacts whose failure stops invoicing or loses job history outrank those whose failure is an inconvenience. If the exit terms are absent, the day you want to leave is the day you discover the real price of the engagement. If the acceptance criteria are absent, "done" is whatever the supplier says it is. Write the reason next to each weight; that reason is what you will defend in front of the board.

Who owns the code if nobody wrote it down

Paying the invoice does not by itself settle ownership. In the default position described by the UK Copyright, Designs and Patents Act 1988, section 11(1), as retrieved from the point-in-time version dated 1 February 1991 and marked "Version Superseded: 01/12/1996": "The author of a work is the first owner of any copyright in it, subject to the following provisions." Section 11(2), from the same superseded version, makes the employer first owner where an employee creates the work in the course of employment, subject to any agreement to the contrary. The current in-force wording was not retrieved, so these are presented as the statutory default as it read at that point, not as current consolidated law.

UK IPO guidance states that the author or creator is usually the first owner, with the employee exception subject to agreement to the contrary; the retrieved portion of that guidance contains no explicit sentence about commissioned works resting with the contractor. A law-firm commentary from Burness Paull makes the commercial point directly: "Without express wording to this effect, that is unlikely to be the case" — that is, bringing in a contractor and paying for the work is unlikely by itself to transfer ownership.

The Slovenian material is weaker. The available wording of ZASP Articles 99 and 100 comes only through an unofficial consolidated text whose amendment chain stops at 56/15; the state register's consolidated text is described as an informational working aid for which the body gives no guarantee, and the wording has not been verified against the later consolidation amended through 130/22. On that limited basis, Article 99 frames a copyright commission contract in which the author undertakes to create a specific work and deliver it to the client, and provides that, unless the Act provides otherwise, the rules on the contract for work apply; Article 100 concerns collective works created on a client's initiative with a large number of co-authors. No conclusion is drawn from this about the buyer's position without contract wording. This is not legal advice. The term you need — who owns the commissioned code, and what licence you receive — must be written down, and the contract wording itself should go to a lawyer qualified in the relevant jurisdiction.

The operative articles of Directive 2009/24/EC were not retrieved, including any rule on programs created by an employee, so no "EU rule on employee-created code" is asserted here.

For a buyer, the only usable conclusion is narrow: the default can leave the term unstated unless it is written, and a published value statement will not change that.

On the day the relationship ends, what do you actually take away?

The exit test for a bespoke build is not the same as leaving a SaaS subscription. On the day the relationship ends, you need to be able to take away: the source code and its licence, the handover documentation, repository and hosting access, the data in an extractable form, and enough written knowledge for a different developer to continue. If one of those items is absent from the written exit terms, record the gap and ask the supplier and your lawyer what continuation or transition would require in your specific arrangement.

The switching-charge rule that sounds like it protects you — and does not

The retrieved switching-charge provisions (Data Act, Article 29) address providers of data processing services and their customers; the definition of that term was not retrieved here, so this article does not treat the rules as establishing what a bespoke-development contract must say about code or documentation handover. The retrieved Article 29 wording says that from 12 January 2027, "providers of data processing services shall not impose any switching charges on the customer for the switching process." Until that date, reduced switching charges may be imposed and "shall not exceed the costs incurred by the provider … that are directly linked to the switching process concerned." Providers must give prospective customers information on standard service fees and early termination penalties before entering into a contract, and must tell customers about services that involve highly complex or costly switching; the retrieved wording of the publication duty is truncated in the source, so it is described only as far as it reads. The regulation's general application date was not verified in this run and is not published here.

A UK government publication restates the same rules and records that Google announced a global free-egress programme on 11 January 2024 and AWS on 5 March 2024; those announcement dates are reported within the document and were not independently verified.

None of that by itself establishes what a bespoke software development contract must say about handing over code and documentation. Treat the switching-charge headline as relevant to data-processing subscriptions, not as a substitute for the exit terms you need in writing.

Which case studies are evidence and which are decoration

For this article's case-study check, treat a quantified figure with a visible denominator or period as stronger published evidence than a logo wall, an anonymous client description, or an unexplained award. This check assesses what the page lets you verify; it does not prove overall supplier quality. Apply that test to any supplier's case-study pages, including the ones published by Niro Digital.

The case-study figures used as examples below are the publishers' own published claims; they were not independently verified, and they are used here to show what an evidence test can and cannot settle. Niro Digital's published Lektobot case study states that on a real 31,204-word master's thesis it inserted 250 reviewed comments in 1 hour 24 minutes, that 400 proposals entered the funnel and 250 verified comments came out, and that 11 independent checks must pass. The Evroproces recruitment case study states that 2,000+ potential workers were generated at an average cost per lead under 1 €, with a 2–3% conversion rate. The published web case studies follow the same pattern: RAIN smart home claims 95+ Lighthouse scores across all pages, WCAG 2.1 AA compliance and first contentful paint under 1.2 seconds; Weedorizers claims Lighthouse 90+ across nine EU markets; Munchies claims Lighthouse 90+ with a bilingual storefront. Those figures carry a denominator or period where the publisher states one; they do not prove anything about code ownership, handover or exit terms. The content-agent platform case study is an internal product rather than client work; it is evidence about the platform Niro Digital built for its own publishing pipeline, not a customer outcome.

You can practise the same reading on the published project library.

Run the same six tests against our own published pages

The credibility test is whether the framework fails the publisher's own material. It does.

The content-agent platform case study publishes a 3,046-word sourced article for $0.37 across 24 model calls, a 14-stage pipeline, 15 deterministic validators, a weighted evaluation scoring 9 quality dimensions, one day of cost engineering that took an article from $8.40 to $0.37, and about $5 per month in model spend for a client on three posts a week. Those are the publisher's own claims, not independently verified. They describe how the platform runs; they do not answer the six evidence tests. You can check this scoring yourself on the case study page.

What a published price page can and cannot tell you

Niro Digital's pricing page publishes indicative price ranges for web development, marketing setup and retainers, and performance optimisation, with the caveat that the ranges reflect typical projects. This article does not reproduce those figures; read them on the page itself. Whatever the ranges say, they cover a class of work and include none of the six artefacts: who owns the code, what the handover pack contains, who is assigned and for how long, how acceptance is defined, how changes are priced and approved, and what happens at exit. Those are contract terms, not price ranges.

How acceptance and changes are defined — where the price you did not see lives

The fourth and fifth artefacts are where invisible cost lives after signature.

Acceptance criteria must be written, testable, and tied to a named verification method. For each criterion, the quote or contract must say how it is verified and what happens when it fails. "We test thoroughly" is not an evidence test.

Change-request pricing must be a written rate or pricing method, not the words "standard rates apply". There must be a named approval route: who approves a change on your side, who prices it on theirs, and the threshold at which the change-control process triggers. An answer given verbally in a meeting is not evidence.

Fixed price or time-and-materials: what each protects you from, then stop

This part is professional judgement against the disclosed framework, not a recommendation, and no retrieved evidence supports one shape over the other for this project. Fixed price protects you from overrun on a scope that is defined; the risk of unknown work sits with the supplier, which is why fixed-price quotes often price that risk in. Time-and-materials protects the supplier from a scope that is not fixed; the risk of discovering more work sits with you. Neither shape is safer in the abstract. What matters is whether the scope and acceptance criteria are specific enough for fixed price to mean anything.

What has to sit behind €18k, €34k and €61k for them to be the same product

The comparison you cannot currently make is across rows, not down a price column. For each artefact, the question is not "which quote is cheaper" but "which quote puts this in writing, and which leaves it silent".

Caption: Constructed illustration written by this article. €18k, €34k and €61k are illustrative price points — not three real quotes the article has seen, not Niro Digital's prices, and not market benchmarks. No supplier is named or implied in any cell.

A lower price is not automatically a missing artefact. It can legitimately reflect a smaller team, a different seniority mix, lower overheads, or genuinely less scope. The scorecard cannot tell you which of those is true from the price alone; it only shows you which artefacts to demand in writing.

Are the three quotes even for the same category of work?

Before reading the terms any further, check each quote against the category. Niro Digital's published custom-software category covers CRMs, dashboards, employee tools and logistics systems shaped around how the business actually works. Your job is scheduling, field reporting and invoicing. If a quote describes a website with a scheduling module, or a marketing site with a form, that is adjacent work, not the same category. A category mismatch creates a price gap no contract term will fix. The custom software development page defines the category; use it as a test, not as a recommendation.

Who is in the meeting versus who does the work

The named-team-and-availability artefact exists because the person in the discovery meeting is often not the person building the system. The evidence test is written: named individuals, their roles, and their committed time. A published team page is not evidence of who will be assigned, no matter how specific it looks. Niro Digital's own site names two people — the founder and the CTO — and reading the about page with this artefact in mind shows you a key-person risk; it does not show you an assignment.

The same discipline applies to capacity. No website can tell you whether an agency is too small or too big for your project. The publisher of this article publishes no threshold for the project scale or budget it does not serve; whether a project sits inside the scale it serves cannot be assessed from the published material, and this article does not answer it.

Three questions that should produce a document, or a recorded gap

In your next meeting, ask these three, and ask for each answer in writing:

1. Which document states who owns the copyright in the commissioned code and what licence we receive? Diagnostic because a supplier who holds the term can produce the clause; a supplier who does not can produce only reassurance.
2. Who is named as assigned to our project, for how many days, and who replaces them if they leave before go-live? Diagnostic because it forces a capacity claim into writing.
3. What exactly is transferred on termination — source code, documentation, repository and hosting access, data — in what format, at what cost, and with what transition support? Diagnostic because exit terms either exist in writing or they do not.

Record a refusal or lack of written evidence as a gap in the scorecard; for ownership and exit terms, take that gap and the proposed contract to your own lawyer before deciding.

What a written artefact can settle, what only your lawyer can settle, and what no website can tell you

A written artefact can settle the first six questions: ownership and licence wording, handover contents, named assignment and availability, acceptance criteria and test method, change pricing and approval route, and exit and extraction terms.

Only your own lawyer can settle the contractual and jurisdictional questions. Frame them as questions, not answers: who owns copyright in the commissioned code absent express wording; what the licence should permit; which jurisdiction's defaults apply to a cross-border contract; where the data sits; and what obligations fall on you as the buyer. This article is not legal advice and does not assume one jurisdiction's defaults apply.

No website can settle code quality, architecture or stack appropriateness. A published stack statement — Niro Digital's own site says it primarily uses Next.js, React and TypeScript, with Shopify and custom builds for e-commerce — is an example of the kind of claim a non-technical buyer cannot adjudicate, not guidance to act on. The scorecard does not measure engineering quality, and a supplier with tidy documents can still build badly. Definitional questions are covered in the FAQ; the contractual questions in this section have no website answer.

Two cases the method does not decide for you

If the €18k quote is silent on an artefact, that does not prove it is deficient. It may genuinely reflect a smaller team, a different seniority mix, lower overheads or less scope. The method records silence as silence; it does not convert silence into a verdict. Ask the supplier to put the artefact in writing, then score the document.

If a supplier refuses to put one of the six artefacts in writing, record the refusal and continue. For ranking purposes, treat silence as a gap rather than an inclusion.

The email to send before the board meeting

The deliverable is a ranked comparison with each weight and its operational reason written next to it, so you can defend the recommendation in the room without reconstructing your reasoning. Send each supplier the same written request for the six artefacts. Then ask the two questions no page can answer: who will actually be assigned and available, and whether your project sits inside the scale the supplier actually serves.

If, after scoring, you want a first digitalisation step defined, or you want Niro Digital assessed as one of the candidates, write via the contact form. The published contact details are info@nirodigital.com and +386 70 630 880. Any question about code ownership, contract terms, jurisdiction or data protection should go to your own lawyer, not to the contact form.