What six interviews revealed
Users wanted earlier warnings, understandable reporting, and specific steps to resolve bad contacts instead of trial-and-error cleanup.
Keap / Thryv / R&D
AI-Powered Email Health
Turning an opaque email-health report into a clear score and useful next steps. A one-sprint exploration of deliverability, engagement, and complaints inside the existing CRM.
- Role
- Product Designer
- Timeline
- April 2025 / 1 sprint
- Research
- 6 in-depth interviews
- Scope
- R&D, reporting, AI concepts
My contribution
I used six user interviews and competitive research to design the widget, report, and assistant concepts. In a workshop with product, research, content, and customer success, we narrowed the scope to helping users understand problems and take action. The work remains a prototype.
Discovery
Users learned about email problems too late.
Campaign suspensions, unclear bounce warnings, and manual list cleanup left users dependent on support. The existing experience scored 48.8 SUS in baseline testing.
The product opportunity
Combine clearer diagnostics with proactive guidance as Keap and Thryv move toward a unified platform.
I wish the system gave me a heads-up.
User interview
Identify the missing safeguards.
Competitive analysis examined suppression, bounce thresholds, and verification to understand where the current experience left users exposed.
Process
Make the score explainable and the next step visible.
A cross-functional workshop with design, research, product, content, and customer success focused the MVP on insight, guidance, and control.
Replace ambiguous metrics
Use a 0-100 score as an entry point, then explain what contributes to it rather than leaving the number on its own.
Organize around three actionable categories
Deliverability explains inbox reach, Engagement explains recipient interaction, and Complaints explains negative feedback.
Keep resolution in the workflow
Pair warnings with cleanup or improvement actions so users can respond without leaving their day-to-day CRM work.
Product Design
From a status report to a useful diagnosis.
The widget offers an at-a-glance view. The report provides category breakdowns, trends, and contextual recommendations for a deeper investigation.
A clear score with an action path.
The redesign groups key metrics and places resolution actions beside the information that motivates them.
The original widget required interpretation.
Ambiguous metrics and weak hierarchy made it difficult to judge health or know what to do next.
Connect the diagnosis to a recommendation.
The proposed report makes the score prominent and adds trends, bounce breakdowns, and AI-powered guidance.
A metric without a clear next step.
Text-heavy explanations and buried actions made the original report harder to use.
The redesigned report inside Keap.
A prototype walkthrough shows how health information and recommendations fit into the existing CRM.
Open prototype on Vimeo ↗A contextual assistant, explored as a concept.
The AI side-panel concept offers recommendations based on the current health data. It is an exploration, not a shipped capability.
Open prototype on Vimeo ↗Next Steps
What still needs testing.
The sprint ended with widget, report, and assistant prototypes. Before building them, the Reporting Center and Email Deliverability teams need to review feasibility and priorities.
- Baseline SUS
- 48.8
- User interviews
- 6
- Health categories
- 3
Existing experience before the redesign, not a result of the new design.
In-depth research into reporting and deliverability pain points.
Deliverability, Engagement, and Complaints organize the proposed experience.
Can users trust the advice?
Check whether recommendations match the email data and explain why an action is suggested. We still need to test what happens when the advice is wrong or incomplete.
Does it help users resolve the problem?
After launch, track whether users complete the recommended action, need less support, and return to the feature. Also check deliverability and complaints; more feature use alone would not prove it works.