What I am changing, and why
The health focus chips on the onboarding form.
All three users slowed down at this section. D. skimmed past it. A. said she wanted a free text field. The chips as currently designed force users into categories that feel either too broad or too specific for what they actually want to communicate. This is an execution problem at a critical moment: the onboarding form is where the curator learns what matters to the user. If users cannot express themselves accurately there, the curator cannot do their job. The fix is straightforward: replace the chip row with a free text field, with a placeholder that invites specificity. "e.g. high protein, no canola oil, love spicy food" gives users permission to be as granular as they need to be without forcing them into a predetermined taxonomy.
The kitchen vetting signal on the plan approval screen.
Two out of three users, D. and A., asked unprompted and independently how the kitchens are sourced and vetted. D. asked it practically. A. asked it from a quality standpoint. That is a pattern, not an opinion. The prototype currently shows restaurant names and neighborhoods but offers no signal about how those restaurants were selected. A single line on the plan approval screen, something like "Every kitchen is vetted by Plated for quality, sourcing, and cleanliness," would address the question before it becomes doubt. It costs one sentence. The benefit is trust at the exact moment the user is deciding whether to approve the plan.
What I am NOT changing — and why
The "No menu. No decisions." headline and landing screen structure.
P. and A. both read it carefully and understood the product immediately. The headline is doing its job for the core user. D. skimmed it, but D. is a lighter fit for the target profile: slightly younger, less settled in his food habits, and more willing to engage with a new product on curiosity alone rather than pain. His behavior on the landing screen is not a signal to change the copy. It is a signal that the landing screen is correctly optimized for someone who has already felt the problem acutely, which is exactly who Plated is for.
The "Request a change" button on the plan approval screen.
A. specifically cited this as the thing that made her feel safe enough to consider approving. "I have an out" is the exact reassurance the product needs to provide at that moment. Removing it or moving it would undermine the trust mechanic the entire approval flow depends on. It stays where it is.
The single-confirmation approval flow.
Nobody asked for more control over individual meal selection. Nobody asked to swap specific dishes or browse alternatives. The requests that came closest, A.'s ingredient specificity, were addressed at the onboarding stage, not the approval stage. The approval flow remains one tap. That is the core bet and the sessions confirmed it is the right one.
What I am doubling down on
The meal plan card with real restaurant names and neighborhoods was the single most effective moment in all three sessions. A. said "oh" when she saw it. P. immediately did the mental math on the weekly total and concluded it was fair. D.'s skepticism about meal kits dissolved the moment he understood the food was coming from local restaurants rather than a central production facility.
That moment, more than any copy or any UI decision, is where Plated becomes real to users. The local sourcing is not just a differentiator from Factor Meals. It is the trust signal. It is what makes the curator feel credible rather than algorithmic. The product should lean into this harder in the next iteration, both in the copy on the landing page and in how the meal plan card presents the kitchen context.
One thing this session surfaced that goes beyond the prototype
The "Is Maya a real person?" question from P. is not a prototype problem. It is a product problem. The curator attribution is currently decorative. It implies a human without proving one. If the answer to that question is yes, the product needs to think carefully about how it demonstrates that: whether through a brief curator profile, a short note with each plan, or some other mechanism.
If the answer is eventually no, the product needs to reckon with the fact that its core trust proposition rests on a human relationship that may not scale. That question does not get answered in this iteration. But it needs to be in the room when the next one is designed.
