How to Choose a Web Development Agency in 2026: 12 Questions That Save Six Figures

We have been hired to rescue about twenty projects that another agency shipped or failed to ship over the past two years. The pattern is always the same: the buyer asked the wrong questions, compared on the wrong axis, signed, and paid twice to get what they should have paid once. Here is the checklist we give to friends before they sign with any agency, ours or otherwise.

Question 1: Who specifically will write my code? Get names. Real names of senior engineers. Not 'our team', not 'a dedicated project manager'. Put the names in the contract. The #1 bait-and-switch in agency sales is sending seniors to the pitch call and juniors to the actual work.

Question 2: Can I see two recent codebases with client permission? Portfolios are decorated. GitHub commits are not. A serious agency will get permission from a recent client and let you see the actual code quality, commit hygiene, test coverage. Refusing this request is a red flag.

Question 3: What is your rescue-project rate? This is a trick question. If they never rescue projects, they are either too new or too small to have seen how bad things can get. If every second project is a rescue, they either do bad initial work or cannot scope new builds. The sweet spot: 15-30% of their volume is rescue work, and they can tell you what they learned from it.

Question 4: What happens if the direction is wrong at milestone one? A serious agency has a written answer: full refund, partial refund with deliverables, or contract exit clause. 'We will talk about it if it happens' is not an answer. TheCodeVolt's version: full refund after first milestone if direction is wrong, no questions asked.

Question 5: Do I own the source code, designs, and deploy access on day one? Not at final payment, not at project acceptance, day one. If the answer delays ownership, that is a vendor-lock trap. Domain, DNS, hosting account, analytics account, source repo access should all be in your name from week one.

Question 6: How do you price maintenance after launch? Agencies that avoid this question are the ones that will charge 2-5x market rate later because you have no alternative. Good agencies quote a clear retainer (hours per month at a fixed rate) or a pay-per-ticket model with documented response times.

Question 7: What does your weekly demo look like? You should be able to open a URL every Friday, click through the week's work, and leave feedback that the team sees by Monday. If 'weekly demo' means a Zoom call where they share a screen but you cannot click, they are not running a real weekly delivery process.

Question 8: What is your stack opinion, and why? Beware of agencies that pick technology by popularity or by what they happen to know. A 2026 Next.js 16 build does not belong on WordPress. A 5-page contact-form site does not need a headless CMS. The right agency argues for a specific stack based on your project, not their comfort zone.

Question 9: How do you handle scope changes? Fixed-price projects without a scope-change process either fail or turn into fights. Good process: up to 10% scope change absorbed, above that a change request with written cost and timeline impact signed by both sides before work starts. Not billed hourly, not assumed 'included in the spirit of partnership'.

Question 10: Can I talk to a recent client without you on the call? If the answer is no, they are filtering references. A confident agency introduces you to a recent client over email and steps out of the conversation. The client will tell you things a sales pitch cannot.

Question 11: What is your post-launch churn? How many clients cancel retainer after year one? If they will not tell you, it is high. If they tell you it is zero, they are lying. A healthy number is 15-25% annual churn because some clients graduate to in-house teams.

Question 12: Why did your last two failed projects fail? Every agency has failed projects. Ones that refuse to admit it either have not done enough volume to hit a failure, or are lying. The answer should be specific and self-aware: 'the client changed their founder and the new one wanted a different direction' is real; 'we have never failed' is a red flag.

Red flags to walk away from: no named engineers in contract, refusal to show a real codebase, no written scope-change process, no clear IP ownership clause, no weekly demo structure, vague maintenance pricing, reference filtering, stack choices driven by the agency's comfort rather than your project, and sales people who promise timelines they cannot know in the first call. Any two of these appearing together means you should walk.