Section 1
Why portfolios are the weakest signal
A portfolio shows outcomes without context. You cannot see the brief the designer worked from, the client who mangled the copy, the budget that got cut, or whether that gorgeous homepage ever converted a visitor. Two equally skilled vendors can show wildly different portfolios because their clients differed, not their ability. Meanwhile the thing that will actually determine your outcome, the vendor's operating process, is invisible in a portfolio and fully discoverable in one conversation. W. Edwards Deming's line applies precisely: a bad system will beat a good person every time. A talented designer inside a chaotic engagement produces chaos; a solid designer inside a disciplined process produces results. So interview for the system. The vendors worth hiring will light up when you ask about process, because clients who care about process are the ones their process was built for. To see how this connects to the wider system, read [How to Run a Website Project Without Being a Designer: The Operator's Process](/blog/how-to-run-a-website-project-without-being-a-designer).
Section 2
The vetting questions that separate pros from decorators
Run every candidate through the same five questions and score what you hear against the table below. You are listening for sequence and accountability: copy before design, structure before polish, defined feedback rounds, named QA ownership, and a post-launch plan that involves measurement rather than disappearance. Price disparities usually make sense once you see process disparities, the cheap quote skips the phases that prevent rework, so you pay the difference later in change orders. One more filter that costs nothing: notice the questions they ask you. A serious vendor interrogates your goals, audience, and proof assets before quoting, the way Smashing Magazine's long-running collections of agency questionnaires document. A vendor who quotes from a two-line email is pricing a guess, and you will fund the gap between guess and reality.
Section 3
Freelancer, agency, or productized system?
A freelancer suits a tight scope with one decision-maker and an operator willing to project-manage: you get senior hands at the cost of bus risk and limited bandwidth. A traditional agency adds project management and breadth, but verify who actually does your work, the team that pitched is often not the team that delivers. The third option is a productized system, where the process itself is the product and your site is one module of a larger growth machine. That is the ConvertOS model: the brief, copy-first sequence, wireframe gates, QA checklist, and post-launch iteration cadence come pre-installed, because they are the same every time. McKinsey's Business Value of Design research found the financial winners were companies that treated design with operational rigor, measured, managed, iterated, rather than as episodic creative procurement. Choose whichever model gives you that rigor, but do not pay for talent and receive chaos. For a deeper look at this, see [AI-Generated Web Design: What It Actually Changes for Service Businesses](/blog/ai-generated-web-design-service-businesses).
Section 4
Managing the engagement after you sign
Vendor management is gate management. Put four approval gates in the contract, signed brief, approved copy, signed wireframes, staging review against the QA checklist, and release payments against gates, not dates. Then hold your side of the system: one decision-maker, feedback consolidated into a single document per round, and answers within two business days, because vendor timelines silently assume client responsiveness. Give feedback as business problems, not art direction: 'prospects ask about pricing on every call and this page buries it' gives the designer something to solve; 'move the box up and make it blue' merely makes them an order-taker. Escalate early if a gate slips, one slipped gate is a conversation, two is a pattern. Handled this way, most vendor relationships go from adversarial to genuinely collaborative, because both sides finally share the same definition of done. A useful companion to this piece is [How to Choose Lead Generation Channels When You Can't Do Them All](/blog/how-to-choose-lead-generation-channels-when-you-cant-do-them-all).
Section 5
What the research says
Research gives you several concrete standards to write into a vendor engagement. Craft matters measurably: Stanford's credibility study found visual 'design look' was the most cited factor in credibility judgments, at 46.1 percent of participants' comments (Fogg et al., 2003), so the visual bar is real, not vanity. But process is where the money is: Baymard Institute's testing found the average site can lift conversion roughly 35 percent through better design of an existing flow, the kind of gain a research-driven process finds and a portfolio-driven one misses (Baymard Institute, 2025). Demand a performance clause: Deloitte measured a 0.1-second mobile speed improvement lifting retail conversions by 8.4 percent, which makes a performance budget worth its line in the contract (Deloitte, 2020). Demand an accessibility clause: WebAIM found 94.8 percent of the top one million homepages had detectable WCAG 2 failures (WebAIM, 2025), and more than 4,000 US digital-accessibility lawsuits were filed in 2024, over a quarter against companies relying on accessibility overlay widgets (UsableNet, 2024), a vendor's QA checklist should name WCAG explicitly. Finally, McKinsey's Design Index found top-quartile design performers grew revenue 32 percentage points faster than peers, and the differentiator was operational rigor: design measured, managed, and iterated like any other business function (McKinsey, 2018). Vet for the vendor who already works that way.