Start with the shape of the work, not the shape of the supplier
Before you compare suppliers, write down three things about the job. How long the result has to live (a campaign site for one quarter is not the same thing as an internal portal that five departments will lean on for six years). How often it will change after launch. And who has to be able to open the code if the relationship ends. Those three answers remove most of your shortlist before you meet anyone.
A rule that holds up: the shorter the life of the thing you are building, the less the supplier's structure matters. The longer it lives, the more it matters who exactly will be sitting with your code in month fourteen.
What each option is genuinely good at
The big agency
Buys you breadth and paperwork. Multi-market campaigns, brand systems, media buying, security questionnaires answered by someone whose actual job that is, professional indemnity insurance, and continuity when a person leaves. If your procurement department needs a supplier that survives an audit, this is the safe answer. The cost is layers: the people who sold the work are usually not the people who do it, decisions travel through account management, and a two-day change can take three weeks of process. You also pay for capacity you never use.
The freelancer
Buys you speed and a direct line, for a single clear scope. A good senior freelancer will out-deliver a large team on defined work and cost a fraction of it. The risk is not competence, it is singularity: one holiday, one illness, one better offer, and your project stops. There is no second pair of eyes on the code, no cover in August, and nobody to escalate to. Excellent for scoped work with a defined end. Fragile as a long-term operational dependency.
The senior studio
Buys you senior people doing the actual work, with more resilience than one person and far less overhead than an agency. It fits mid-market builds that have to be maintained for years, and retainers where the people who built the thing are the ones keeping it alive. The limits are real: capacity is finite, a studio cannot staff six workstreams at once, and specialisms outside its core (large media buying, broadcast production, heavy enterprise integration programmes) come from a network of specialists rather than from in house. Ask about that plainly.
Compare them on what actually differs
- Who does the work: agency, mixed seniority behind an account manager; freelancer, the person you met; studio, the person you met, with backup.
- What breaks it: agency, slow change control; freelancer, one human being; studio, capacity limits.
- Price anchors published in Europe in 2026: fixed-price prototypes and discovery workshops from around 3,500 EUR; fixed-scope MVP builds from around 12,990 EUR; AI implementation phases from around 25,000 EUR; unlimited-design retainers from around 2,500 EUR a month; small maintenance retainers from around 120 EUR a month. Most agency websites publish no figure at all, so publishing one is itself a shortlist signal.
- Continuity: agency, high but impersonal; freelancer, low; studio, medium and personal.
- Breadth: agency, wide; freelancer, narrow; studio, deep in a few things plus a network.
- Where the code ends up: this depends on the contract, not on the supplier type. Ask in all three cases.
The questions to ask before you sign
- Who, by name, will write the code, and what share of their week is on us?
- Can you show me a repository from a live client, with their permission? Commit history and code review tell you more than a slide.
- What happens in August, and what happens if that person leaves?
- Which accounts are in our name from day one: repository, hosting, database, domain, analytics, app store?
- What does the first month after launch cost, and what is included in it?
- What have you got wrong recently, and what changed as a result? A supplier with no failures is either new or not telling you.
- If we stop next quarter, exactly what do we receive, and how long does it take?
The contract terms that decide how this ends
Ownership of code and data
Say it plainly: all source code, database schemas, content and data are the client's property, assigned on payment, with no licence-back conditions. Then check that it is true in practice. Ownership on paper is worthless if the repository sits in the supplier's account and the database lives inside a platform you have no admin login to. Ask for owner-level access on the day work starts, not at handover. Polargate has moved several client sites off vendor platforms onto infrastructure the client owns (repository, hosting, database), verifying that schema, access rules, users, storage and functions matched before cutover. It is doable, and it is far cheaper to avoid than to fix.
Exit
An exit clause should name the artefacts and the clock: repository access, a database export in an open format, environment variables and secrets transferred through a secure channel, documentation of deploys and third-party accounts, and a defined number of paid transition hours. Thirty days notice with a written handover is normal and reasonable. If a supplier resists putting that on paper, you have your answer to a different question.
Response times
Be specific and be realistic. Define severity levels (site down, feature broken, cosmetic), a response time per level, working hours, and what happens outside them. Be careful with anyone promising 24/7 without telling you who is awake. Agree how the retainer works too: what a normal month includes, what an urgent request costs, and whether unused hours roll over. Polargate runs client maintenance through a ServiceDesk with per-client hour contracts and a visible hour balance, precisely because "we'll take a look" is not a service level.
Honest signals of trouble
- Nobody's name on the website. No founder, no team, no location, only a form.
- No repository or environment in your name after kick-off.
- Numbers you cannot check: awards without a year, "300+ projects" without one you can phone.
- Their own house does not match what they sell. Reviewing studio websites in 2026 we found one whose TLS certificate had been expired for months, several whose achievement counters rendered as zero without JavaScript, and agencies selling AI-search visibility whose robots.txt blocked the AI crawlers.
- The senior person you met disappears after the proposal.
- No written scope, or a scope so broad that any dispute is unwinnable.
- Estimates with no range and no stated assumptions. Every honest estimate carries both.
- Reluctance to show a mistake they made. One of ours: a routine audit round on a client site raised 37 findings and only 4 survived independent verification. The useful part was the 4, and saying so out loud.
How to run the decision
Buy a small, paid piece of work first. A fixed-price discovery sprint that ends in a written scope, an architecture, a cost range and a schedule tells you more about a supplier than three sales meetings, and if you dislike the result you own the document and can take it elsewhere. That is why Polargate starts every client relationship this way. Choose the structure that fits the life of the work, then put ownership, exit and response terms in writing before the first invoice is issued.
