Feature work often starts too vaguely. A ticket says "improve onboarding" or "make the upgrade screen clearer," then the team jumps from opinions to implementation. I use Mobbin to put real product evidence between those steps.
Referral disclosure: this article includes my Mobbin referral link. If you are a new user who becomes a qualifying paid customer through it, I may receive a reward from Mobbin. Check the current offer and eligibility in my Mobbin referral and discount guide.
Before searching, write the decision
I start with one sentence:
We need to decide how this feature should handle [moment] for [user] without causing [risk].
That sentence stops the research from turning into a mood board. Mobbin contains individual screens, ordered flows, and website sections. Its MCP tools can also search with natural language and return screen images inline.
1. Search for the first meaningful action
Prompt:
Find onboarding flows where a new user reaches a useful product state before completing every setup task.
I use this to challenge onboarding that asks for too much too early. I compare the first useful action, which setup steps are deferred, and whether the product shows value before requesting a larger commitment.
2. Search for progressive setup
Prompt:
Find products that move account setup into a checklist inside the main experience. Compare progress, optional steps, and resume behavior.
This is useful when a long wizard feels brittle. A checklist can reduce the distance to the product, but it can also leave users in a half-configured state. The research should expose that tradeoff.
3. Search for empty states with one clear job
Prompt:
Find empty dashboards that help a new user complete one meaningful first action. Compare the primary CTA, explanation, and example content.
Empty states often become decoration plus a generic "Create" button. Better examples explain the outcome, preview the finished state, or offer sample data.
4. Search for permission pre-prompts
Prompt:
Find mobile flows that explain the benefit of notifications, camera, location, or contacts before opening the system permission prompt.
The official Mobbin MCP page highlights permission research as a useful query. I compare timing, benefit copy, privacy explanation, and the path after denial. The real decision is when the request earns the right to appear.
5. Search for honest paywalls
Prompt:
Find subscription paywalls that show the price near the CTA, explain billing terms, provide a visible dismissal path, and avoid artificial urgency.
Paywall research is easy to misuse because aggressive patterns can look effective without showing their long-term cost. I compare plan hierarchy, trial language, renewal terms, restore purchase behavior, and whether the free path remains available.
Mobbin's current MCP examples compare paywalls from shipped products and call out concrete details such as price placement, risk reversal, and visible cancellation language.
6. Search for upgrade moments in context
Prompt:
Find upgrade prompts triggered by a real product limit. Compare how the screen explains the blocked action, paid benefit, and return path.
A standalone pricing page does not show how an upgrade feels during use. I check whether the prompt preserves the user's work and lets them recover without starting over.
7. Search for failure and recovery states
Prompt:
Find checkout, upload, or account-linking flows with recoverable errors. Compare error placement, retained input, retry behavior, and support options.
Failure-state research exposes whether the flow was designed as a complete system. I look for what failed, what remains safe, what the user can do now, and whether retrying creates duplicate work.
8. Search for cancellation and downgrade
Prompt:
Find cancellation flows that confirm consequences clearly, offer relevant alternatives, and still provide a direct way to leave.
I compare data retention, effective dates, pause or downgrade options, refund language, and confirmation states. Retention offers can be useful, but hiding the exit damages trust.
9. Search for settings information architecture
Prompt:
Find settings screens for products with accounts, billing, notifications, privacy, and integrations. Compare grouping, labels, search, and destructive actions.
Settings pages grow until every team owns a section. Mobbin references help test whether categories match a user's mental model or mirror the organization chart.
10. Search for the complete handoff
Prompt:
Find flows that end with a clear success state and hand the user into the next useful action. Compare confirmation, education, and undo options.
Teams often design the form and forget what follows it. A good handoff answers: Did it work? What changed? Where am I now? What should I do next? The final screen is part of the feature, not cleanup.
Turn references into a short implementation brief
After the searches, I do not ask an engineer or coding agent to copy the winning screen. I write a brief with:
- the user decision and risk
- three patterns observed across multiple products
- the approach selected and why
- the existing design-system components to reuse
- loading, empty, error, success, and permission states
- accessibility and analytics requirements
- the smallest test that proves the complete flow works
If I am working in Codex, I use the process in my practical Mobbin MCP and Codex guide: search focused flows, compare the reasoning behind them, then let the repository and product requirements determine the implementation.
The rule that keeps inspiration useful
One reference is taste. Several references reveal a pattern.
I try to compare at least three different products before writing a recommendation. Repeated choices may indicate a useful convention, but frequency is not proof. The right answer still depends on the user's goal, the business model, accessibility, platform conventions, and the constraints already present in the product.
Mobbin is worth paying for when this kind of research happens regularly. It shortens the path from a fuzzy feature request to concrete options the team can discuss. If that fits your workflow, open Mobbin through my referral link and confirm the current offer at checkout.
The best result is not a screen that looks familiar. It is a decision with enough evidence that implementation can become boring.