Enterprise ERP (NDA)
2023
Nexus
Designing an enterprise analytics experience that helps finance teams move from high-level subscription metrics to the underlying accounts, products, and transactions.
Overview
Making subscription metrics easier to investigate
Nexus is a product designed to surface recurring revenue, retention, and unit economics through accessible dashboard portlets for CFOs and controllers at SaaS companies. I designed the interfaces across the project — dashboards, KPI drilldowns, preferences, narrative insights, and later interface variants.
The Problem
Finance teams needed a clearer view of retention
Finance leaders could see that a subscription metric had changed, but not easily find why — which customer, product, or transaction was behind it. The Retention Tab requirements identified this as a clear product gap: a large set of related measures with no structured way to move from a headline number to the evidence behind it.
Research
Grounding the design in real requirements
Rather than running net-new discovery, I worked from the Retention Tab requirements document that identified missing retention capabilities as a product gap. The brief was clear on user need, understand what changed, compare periods, identify the affected segment, and find the records behind a metric, and on business intent: support retention decisions and strengthen the company's offering for SaaS companies, with loyalty, higher ARR, and reduced churn as stated goals.
What the requirements made clear
The design challenge wasn't simply adding more charts—it was organizing a large set of related measures while keeping definitions, filters, and comparison context understandable.
Ideation
Turning requirements into a visual hierarchy
I translated the retention requirements into a hierarchy of summary, trends, groups, and accountlevel detail — giving every visualization a clear job tied to a specific question a finance user would ask.
Requirement | Design Response | Question it answers |
|---|---|---|
KPI scorecard & comparisons | Headline cards + period/variance table | What changed? |
Renewal, churn & retention trends | Time-series portlets with period controls | Is this a pattern over time? |
Churn by product / renewal by item | Item-level breakdowns | Which part needs attention? |
Cohort analysis | Signup-month rows x elapsed month columns | Which customer groups are weakening? |
Net retention by customer | Account rows across calendar months | Which accounts should we inspect? |
Design consideration
Portlet-level date controls add flexibility, but their scope has to stay visible so users don't mistake two differently filtered views for a like-for-like comparison — a tradeoff I had to design around rather than test away.
Designs
Connecting dashboards, drilldowns, and investigation
The final design connects a retention overview, cohort and account-level detail, a full KPI drilldown, and metric-definition preferences into one investigation path — so a controller can go from "retention dropped" to the specific accounts worth a closer look without leaving the workflow.
Retention Tab: KPI cards sit above a comparison table — value, comparison period, and variance together so users inspect a change instead of reacting to a single number.
Cohort Analysis: Signup-month cohorts tracked by elapsed month. The heatmap surfaces which customer groups are weakening over time, with numeric values kept visible alongside color.
Lessons
Designing for structure, not just display
This project reinforced that in financial interfaces, confidence comes from being able to trace a number back to its basis — not from the number itself. The clearest lesson was to treat a narrative summary as useful only when a user can check it against the selected period, segment, and underlying values; warnings, definitions, and a path to supporting detail have to ship together, not as a trust badge bolted on top.
If the initiative continues, the next step is validating whether finance users can actually explain a KPI change, identify a deteriorating cohort, and trace a figure to a record — with task success, interpretation errors, and time-to-evidence as the measures that would tell us.
