Most founders vet a custom software development company the way they’d vet a restaurant – reviews, a portfolio, a gut feeling on the sales call. That process works fine for a €40 dinner. It’s a poor way to choose a partner for a €70,000 decision that will shape your product for the next two years, precisely because a bad restaurant choice costs one evening and a bad vendor choice compounds for the life of the codebase.
Step 1: Check References Against Your Project’s Actual Complexity
A reference list full of happy clients tells you the vendor can deliver something. It doesn’t tell you they can deliver your something. Ask specifically for two references whose projects matched your integration count, compliance requirements, or technical complexity – a vendor with twenty glowing testimonials for simple marketing sites is not automatically qualified to build a multi-tenant SaaS platform with payment processing.
Step 2: Request a Discovery Phase Before Any Fixed Quote
A vendor willing to run a paid or lightweight discovery phase before issuing a fixed quote is signaling they intend to estimate based on your actual requirements, not a guess from your feature list. A vendor that quotes a firm number after a single 30-minute call is either extremely lucky at estimation or planning to make up the difference with change orders later.
Step 3: Ask to See a Real Code Sample or Architecture Decision
Portfolios show screenshots. Screenshots don’t show code quality, test coverage, or architectural decisions – the things that determine whether your product can actually scale past its first hundred users. Ask a technical advisor to review a sanitized code sample or to sit in on a technical Q&A with the vendor’s lead engineer, not just their account manager.
Step 4: Clarify the Change-Order and Communication Process in Writing
Get the escalation path, sprint reporting cadence, and change-order pricing policy documented before signing – not discovered mid-project when requirements inevitably shift. A vendor who can’t clearly explain how they’ll handle a scope change before you’ve asked is unlikely to handle it gracefully once it actually happens.
Step 5: Confirm Who Specifically Will Work on Your Project
The person who runs your sales call and the person who writes your code are frequently different people at different seniority levels. Ask by name and role who will be assigned to the project, and whether that team is dedicated to you or split across multiple concurrent client engagements – a shared team means slower turnaround on your specific issues.
Step 6: Run a Small Paid Pilot Before the Full Engagement
A two-to-four week paid pilot – a discovery sprint, a single well-defined feature, a technical audit – reveals more about communication quality, code standards, and delivery speed than any number of sales calls. Vendors confident in their work welcome this. Vendors who resist a pilot are often protecting a gap between how they pitch and how they deliver.
Why a Pilot Is a More Reliable Signal Than a Promise
Economist Michael Spence’s signaling theory, part of the body of work on information asymmetry recognized by the 2001 Nobel Memorial Prize in Economic Sciences, offers a useful lens for why a paid pilot reveals more than any verbal assurance a vendor can give during a sales conversation. Spence’s original insight was about job markets – why a costly, hard-to-fake credential like a degree can convey real information about a candidate’s ability precisely because it’s expensive to acquire, unlike a claim that costs nothing to make. The same logic applies to vendor vetting: a claim [“we’re great communicators,” “our code quality is excellent”] costs a vendor nothing to say, which is exactly why it carries little information. A two-week paid pilot costs the vendor instant and real reputational risk if it goes badly, which is exactly why a vendor’s willingness to do one – and how they actually perform during it – carries genuine signal a sales pitch cannot.
This reframes several items on the six-step list above. A vendor’s willingness to name specific past clients for reference calls is a costly signal, because a bad past client experience would surface in that call and the vendor knows it. A vendor’s willingness to let a technical advisor review real code is costly for the same reason. A vendor’s refusal to do either isn’t necessarily proof of poor quality, but it does mean the founder is left relying entirely on cheap, unverifiable claims – exactly the situation information asymmetry research says a rational buyer should be most cautious about.
The practical upshot is a useful reframing of the entire evaluation process: don’t just collect answers from a prospective vendor, notice which of those answers cost the vendor something to give. A confident verbal claim is the cheapest kind of evidence available. A named reference, a real code sample, and a paid pilot are progressively more expensive for a vendor to offer if they aren’t confident in the underlying quality – which is precisely why they’re progressively more trustworthy as evidence.
Manifera’s Approach: Built to Survive This Exact Vetting Process
- Amsterdam (Governance/Transparency): Dutch project leads run structured discovery before quoting, document communication and change-order policy in writing from the first proposal, and are available for direct technical Q&A with a client’s advisors before any contract is signed.
- Vietnam (Execution/Velocity): The assigned engineering pod is named and dedicated, not a shared resource pool – clients know exactly who is building their product from kickoff, and can request a paid pilot sprint before committing to the full engagement.
This is Dutch Management × Vietnamese Mastery applied to vendor trust itself: European process transparency paired with a dedicated, named delivery team a client can actually verify. Learn about Manifera’s way of working before you commit.
Reading the Six Steps as a Cost Ladder, Not a Checklist
Treating these six steps as a simple checklist – did the vendor say yes to all of them – misses the more useful structure underneath: each step costs the vendor progressively more to satisfy honestly, which means a vendor’s pattern of responses across all six tells you more than any single answer in isolation. A vendor happy to talk at length (cheap) but reluctant to name specific references (moderately costly) and unwilling to commit to a pilot (most costly) is showing a specific, informative pattern – not failing a binary pass/fail test, but revealing exactly where their confidence runs out.
This is also why founders should be wary of treating any single step as sufficient on its own. A glowing reference alone is weaker evidence than it appears, since a vendor selects which references to offer – the real signal comes from combining a reference call with a pilot, so the founder isn’t relying solely on a data point the vendor chose to share. Each additional costly signal a vendor is willing to provide narrows the gap between what they claim and what a founder can actually verify before committing real budget.
Case Study: A Munich Logistics Firm’s Pilot-First Decision
Ravenholt Logistics, based in Munich, shortlisted three vendors for a route-optimization platform and requested a paid two-week discovery pilot from each before choosing. Two vendors declined, citing “standard process doesn’t include a pilot.” Manifera’s Amsterdam team ran the pilot, producing a documented requirements analysis and a working proof-of-concept for the highest-risk integration.
That pilot became the basis for an accurate fixed-price quote, and the Vietnam pod delivered the full platform in fourteen weeks with no scope disputes – because both sides had already tested how they’d work together before signing anything larger.
“The two vendors who wouldn’t do a pilot were telling us something important before we even asked. We just had to listen to it.”
Vendor Vetting Checklist
Before You Sign Anything
A vendor that welcomes all six of these steps is signaling confidence in their own delivery. A vendor that resists more than one or two is asking you to take their word for something they could easily demonstrate instead. Get a custom team proposal within 48 hours – pilot sprint included.
Scoring the Six Signals: A Weighted Vetting Scorecard
Not all six steps carry equal weight, and treating them as equally important is a common mistake when comparing multiple custom software development companies side by side. Industry surveys of failed software engagements consistently point to two root causes above all others: mismatched team seniority (cited in roughly 35% of failed vendor relationships) and undocumented scope-change handling (around 28%). That puts “confirm who specifically will work on your project” and “clarify the change-order process in writing” ahead of reference checks in practical weight, even though reference checks are the step founders default to first.
A workable scoring approach: assign 25 points each to the pilot sprint and the named-team confirmation, 20 points to the change-order policy in writing, 15 points to the discovery-before-quoting step, 10 points to the code/architecture review, and 5 points to reference checks. A custom software development company scoring below 60 out of 100 on this weighting – meaning they resisted the pilot, wouldn’t name the team, or wouldn’t put change orders in writing – has failed the two highest-weighted signals regardless of how polished their references sound. Score every vendor on the same rubric before comparing quotes; a €15,000 cheaper proposal from a vendor scoring 40 is rarely actually cheaper once the first unplanned scope change arrives.
FAQs
[Scenario: founder unsure how much vetting is really necessary] Is a full six-step vetting process overkill for a smaller project?
Not necessarily all six in full depth, but references, a discovery conversation, and a written change-order policy are worth confirming even for smaller engagements – the cost of a bad vendor match scales down with project size, but rarely to zero.
[Scenario: founder worried about the cost of a paid pilot] Is it worth paying for a pilot sprint before the full contract?
Yes, in almost every case. A two-to-four week pilot typically costs a small fraction of the total project and prevents the much larger cost of discovering a poor fit six months and half a budget into a full engagement.
[Scenario: founder trying to compare team quality across vendors] How do I actually evaluate code quality if I’m not technical myself?
Bring in a technical advisor for a single paid review session, or ask the vendor’s lead engineer to walk a non-technical stakeholder through an architecture decision in plain language – how clearly they can explain it is itself a useful signal.
[Scenario: founder deciding between a large agency and a smaller boutique firm] Does company size matter more than the six vetting steps?
Less than you’d think. A large agency can still assign a junior, shared team; a smaller firm can still offer a dedicated, senior pod. The six steps reveal the actual team and process you’ll get regardless of company size.
[Scenario: founder wanting to avoid an awkward vendor conversation] What if a vendor gets defensive when I ask for references or a pilot?
Treat that reaction as data. A vendor confident in their delivery welcomes scrutiny before a large commitment; defensiveness at this stage is a preview of how they’ll handle disagreements once the contract is signed.
[Scenario: founder comparing several custom software development companies on price alone] How do I compare quotes from different custom development companies when the numbers vary wildly?
Normalize each quote against the six-step scorecard before comparing totals – a lower number from a vendor who resisted a pilot or wouldn’t name the team is not actually the cheaper option once change-order costs are factored in.
[Scenario: founder about to sign with a software development company found through a marketplace listing] Should I trust vendor ratings on freelance marketplaces instead of running my own vetting?
Marketplace ratings measure past client satisfaction on unrelated projects, not fit for your specific complexity or compliance needs – use them as a first filter, not a substitute for the six-step process.
[Scenario: founder worried a custom software development company is subcontracting the work] How do I find out if the company I’m vetting will actually build it themselves or subcontract it?
Ask directly in writing whether any part of delivery will be subcontracted, and request the subcontractor’s name and location if so – a vendor unwilling to answer this plainly is hiding a material fact about who controls your codebase.
[Scenario: founder needing a quick gut-check before a first sales call] what is the single fastest question to ask a custom software development company to gauge their honesty early?
Ask what a recent project’s biggest failure or delay was and how they handled it – a vendor with a specific, honest answer is far more trustworthy than one who claims to have never missed a deadline.
Suggested Posts
How useful was this post?
Share it with your friends
Get our latest articles here!
Do you have any questions?
Help us improve the content of this Insightful blog by asking us questions. Manifera's team of experts will help you answer these questions as soon as possible.








