Product design
More data was never the problem. Knowing what to do next was.
Sales Intelligence
Platform Design
Role
Product Designer
Timeline
3 months
team
Sole Product Designer
platform
Web
The Real Problem
The platform pulled together CRM records, public filings, hiring activity, funding events and website behaviour, and turned them into a connected view of an account and the people inside it. On paper that is everything a seller could want.
In practice, users opened it and did nothing.
The MVP had strong underlying logic and a weak journey. Signals were visible but not ranked. Contacts were listed but not connected to the reason the account mattered today. Warm introduction paths were informative and somehow still not actionable.
What users said, almost word for word:
"I do not need more data. I need to know which account matters now, who can open the door, and what to do next."
"I need a system that helps the team act on the right accounts faster, not another dashboard that creates more interpretation."
"I can see who could introduce me. I cannot tell whether that introduction is worth asking for."
The product was not missing information. It was missing a through line from insight to action.
Finding the Fix
I mapped the existing MVP across the six stages a user actually moves through, and named the specific point where each stage lost them.
The journey analysis surfaced this:
Discover accounts.
Signals were wide and priority was unclear, so scanning was slow and arbitrary.
Review buyer signals.
Signals were visible, but ranking, urgency and explanation were not clear enough for a fast decision.
Open account detail.
Context, signals, contacts and actions sat too close together, producing weak hierarchy and high cognitive load.
Inspect contacts.
Contact relevance was never tied back to the signal that made the account interesting.
Find a warm introduction path.
Relationship paths felt exploratory rather than confidently actionable.
Take the next action.
Primary next steps were not visually dominant, so users hesitated and then left.
Synthesised, that is six problems and one shape: the product asked users to switch between exploring and acting far too often, and never made the act half louder than the explore half.
The three solutions I focused on:
A ranked entry point.
Discovery opens on a prioritised set with the reason for the ranking stated, rather than a wide field of equivalent signals.
Three layers, clearly separated.
Signal, contact and action each got their own hierarchy inside the account view, so the eye moves in the order the decision does.
Explained relationship intelligence.
Introduction paths show strength and basis, not just existence, so a user can tell a good path from a technically valid one.
What Actually Happened
The first pass at discovery ranked accounts and said nothing about why. It tested worse than the unranked version, which was the most useful failure of the project. Users did not want a better list. They wanted to know what the list was reacting to. Adding a short plain language reason next to each ranked account changed how the whole screen was read.
The account view took three attempts. Version one separated signal, contact and action into three equal columns, which just moved the hierarchy problem sideways. Version two collapsed everything into one scroll and buried the action. Version three gave signals the top of the view, contacts a middle band tied explicitly to the active signal, and pinned the next action so it was never off screen.
Warm introductions were the part I most misjudged at first. I designed a graph view, because the data was a graph. It was interesting to look at and nobody could act on it. I replaced it with a ranked list of paths, each with the strength of the relationship and what that strength is based on. Less impressive, dramatically more usable.
The AI assistance also moved. Initially it lived in its own area, which made it a feature users had to remember to visit. I dissolved it into the workflow, so a draft or a summary appears where the user already is rather than somewhere they have to go.
What Changed
The redesign gave the product the journey its intelligence deserved:
Discovery that opens on a ranked set with the reasoning shown
An account view with genuine hierarchy between signal, contact and action
Introduction paths presented by strength rather than by existence
AI assistance that sits inside the workflow rather than beside it
A component system covering the dashboard, account and contact surfaces
Underneath all of it, the finding that mattered most was structural: the product needed a better information architecture before it needed anything else. I owned the research and the design end to end.
What I Had to Work With
I was the sole product designer, which meant I owned the research as well as the interface. That is the reason this project has evidence behind it rather than just screens.
What the work relied on:
Product analytics covering usage, funnels, cohorts, retention and feature adoption
Session recordings and heatmaps, plus session replay tied to frontend errors and network activity
Surveys and onboarding questionnaires
Unmoderated prototype testing and moderated sessions with recruited participants
Three personas came out of that, built from real segments rather than invented: a senior account executive in mid market B2B, a revenue operations manager at a growth stage company, and a founder still selling personally. Each was documented with goals, pain points, behaviours, tool ecosystem, success metrics, working style and the specific redesign opportunities they implied.
The hard constraint was that the underlying intelligence was good. This was not a case of a weak product needing a better interface. It was a strong product whose value was buried in its own information architecture, which meant the redesign had to be structural rather than cosmetic.
What I'd Do Differently
I would run the journey analysis before the first design pass rather than alongside it. It produced the sharpest insights in the project and it produced them slightly too late to save two rounds of work.
I would also take further:
Outcome tracking, so the platform can learn which signals actually preceded a closed deal
Team level views, since revenue operations and the seller need the same data at different altitudes
Explicit confidence framing on every inferred score, rather than presenting a number as a fact
What I Learned
A ranking without a reason is not trusted.
Users will accept a shorter list. They will not accept one they cannot interrogate.
Hierarchy is the whole job in a data heavy product.
Signal, contact and action are three different questions. Giving them equal visual weight means answering none of them.
Show the useful version, not the accurate one.
The relationship graph was the truest picture of the data and the least useful thing on the screen. A ranked list of paths was neither, and it worked.
Research is what makes the redesign defensible.
The interface changes came out of a documented journey analysis, which meant every decision had an answer behind it when someone asked why.
