The good news is that you do not need technical knowledge to choose well. You need a small set of questions that separate people who will serve you from people who will bill you, and the answers are legible to anybody.
What actually goes wrong
Understanding the common failures tells you what to protect against, and they are more mundane than horror stories suggest.
The most frequent is ownership. Work is delivered, paid for, and the business discovers later that it does not control the domain, the advertising account, the analytics property, or the source files. Absent a written assignment, the person who created something retains the copyright to it in the United States, which means a logo or a website you paid for may not be yours to modify. This is entirely preventable with one clause and entirely irreversible once a relationship has ended badly.
The second is dependency. A site is built on a platform only that provider can host, or in a system that requires them for routine changes. Every price update becomes a request. The provider may not have intended this, and the effect is the same: you are renting something you believed you had bought.
The third is silence. The work starts, communication thins, deadlines pass with explanations, and the business has no way to tell whether progress is happening. This is the most common complaint and the hardest to write a contract against, which is why the questions that predict it matter.
The fourth is misdirection, where a business asks for a specific thing, receives exactly that, and it does not help. A new website will not fix a positioning problem. Advertising will not fix a page that does not convert. A provider who delivers what was requested without questioning whether it addresses the problem has technically performed and has not helped.
The questions worth asking
Six questions, and the answers tell you more than any portfolio.
What would you need to know before recommending an approach? This is the single most revealing question. Somebody who responds with questions about your customers, your pricing, and where your work currently comes from is going to solve a problem. Somebody who moves straight to what they would build is selling a service they already had.
Who will the domain and the accounts be registered to? The correct answer is you, in your name or your business name, with them granted access. Any other answer, however reasonably explained, means you do not fully control your own presence. This question also tells you how they treat clients generally, because a provider who registers domains under their own account has decided that convenience for them outweighs control for you.
Will I receive the source files, and can this be hosted anywhere? Editable design files, not just exports. A site that can move to another host. Both should be immediate yeses, and hesitation here is a portability problem you will meet at the worst possible moment.
What happens when I ask for something outside the agreed scope? Every provider says there are no hidden fees. What distinguishes them is what happens when you request a change halfway through. A good answer describes telling you at the time and what it would cost. A vague answer means the conversation happens at invoicing, which is where surprise bills come from.
Under what circumstances would you tell me to spend less? The answer reveals whether their recommendation is independent of their revenue. Somebody who can describe a real situation where they advised a client to do less, or nothing, is answerable to the outcome. Somebody who cannot has never separated the two.
What does this look like if we stop working together? How much notice, what you keep, and how long a handover takes. Ask it early, while everybody is optimistic, because the answer is much harder to get later and it is the one that protects you most.
Reading the answers
Certain patterns are reliable enough to act on.
Specificity beats enthusiasm. A provider describing exactly what they would do first, and why, is more likely to deliver than one who is excited about your business. Enthusiasm is easy to produce.
Willingness to disagree is a strong positive signal. Somebody who tells you the thing you asked for is not what you need, in a first conversation, before any money has moved, is demonstrating the behaviour you actually want. It is uncomfortable and it is the tell.
Guarantees about ranking or results are a reliable negative. Nobody controls search placement, and a promise of a specific position means either misunderstanding or something that will eventually cost you the listing. The same applies to guaranteed traffic figures.
Proprietary metrics deserve suspicion. A score with a name nobody else uses, which cannot be compared against anything, exists to produce a number that goes up regardless of what happens to your business.
And be careful with a portfolio of businesses much larger than yours. Impressive work for a company with a marketing department tells you little about how somebody serves an owner making every decision themselves, and the process that suits one is frequently wrong for the other.
What to agree before work begins
Four things in writing, and they can fit on a page.
Scope, stated specifically enough that both parties would agree on whether a given request is included. Vagueness here is the source of most disputes, and they are rarely bad faith. They are two people who genuinely remembered the arrangement differently.
Ownership, assigning copyright in the work to you on payment, along with delivery of source files. This is the clause that prevents the most expensive category of problem and it is one sentence.
Payment terms tied to stages rather than dates, so money moves when work does, plus what happens if the engagement ends partway.
And how you will know what is happening: how often you hear from them, in what form, and what you should expect to see at each stage. Written updates are more useful than calls, because a document can be reread in six months when a decision comes up again.
The judgement that matters most
Beyond any checklist, the question worth holding is whether this person is answerable to your outcome or to their deliverable.
A provider answerable to a deliverable builds what was ordered. A provider answerable to an outcome asks whether the order was right, says so when it is not, and occasionally talks you out of spending money with them. The second is rarer, harder to find, and worth considerably more, because in a first year the largest risk is not paying too much for good work. It is paying a fair price for the wrong work, which costs the money and the months.
If you are unsure after a first conversation, ask what they would do with your budget if they could not sell you anything. The answer, and whether they can give one, tells you most of what you need to know.