Skip to content
← Work

CASE STUDYBuilt on SagUI

Sales Analytics: from signal to action.

A dashboard for a regional sales leader running 10 convenience stores. It finds the best and worst sellers, compares stores against target and watches stock — then turns what it finds into alerts with a recommended next step, and actions you can track to done.

ROLE
Product Designer & Design-Engineer (solo)
STACK
Next.js 16 · React 19 · TypeScript · SagUI · Vercel
DATA
MDM master data + two years of seeded daily sales
Stores from MDM master data
10
Alert types, each with a next step
6
Days of seeded daily sales
730
New components — all SagUI
0
Sales Analytics screens laid out at an angle as panels: overview metrics, revenue trend, best and worst sellers, store targets, category map, stock health, alert inbox, recommended action, take action, action tracker, inventory health and product detail

One system, three layers

Sales Analytics is the third piece of one system. Each layer does one job, and each makes the next one possible.

  1. DATA

    MDM

    Governs the master data — stores, vendors, categories, products — behind a staging-and-approval gate.

  2. INTERFACE

    SagUI

    Provides every component, chart and token. Screens are assembled, not designed one by one.

  3. DECISIONS

    Sales Analytics

    Consumes MDM's data, is built from SagUI, and turns sales signals into tracked actions.

The problem

A regional leader’s day isn’t short of numbers. It’s short of decisions. Most retail dashboards answer “what happened?” with charts, then leave the leader to work out what to do, who to call, and whether anyone did it.

  1. 01

    What needs my attention today?

    Not every metric — the few things costing money right now.

  2. 02

    What should I do about it?

    A specific, sensible next step, not just a red number.

  3. 03

    Did it get done?

    A loop that closes, so the same problem doesn't quietly come back.

Built on MDM’s master data

Sales Analytics doesn’t invent its own stores and products. Its 10 stores, 10 vendors, 16 categories and 25 products come from the same convenience-store data set the MDM platform governs — and that pays off in the details.

  • Alerts know who to callEach product carries its MDM vendor, so a stockout recommends “Reorder 900 units from Healthy Beverage Corp.” — not just “reorder”.
  • Stores know their managerA falling store recommends asking its named manager for a recovery plan.
  • Categories group the analysisThe category map, product filters and margins all roll up the MDM category tree.

On top of that master data, a seeded generator builds two years of daily sales and a stock snapshot — 244 store-product series × 730 days. The same seed always gives the same numbers, with planted scenarios so the product has a story to tell: an energy drink surging and out of stock at two stores, a salad in decline and overstocked, a store losing traffic, and the airport store as the busiest location.

MDM keeps the master data trustworthy. Sales Analytics shows what that data is worth.

Information architecture

Six areas, split into analysis and operations. Every analysis view shares one store filter and one period control (7 days, 30 days, 90 days, 12 months), and every metric is compared with the period before it.

AREAJOBKEY VIEWS
DashboardToday at a glanceRevenue, units, margin, open alerts; revenue trend vs. the previous period
ProductsWhat sells and what doesn'tHighest and lowest selling, biggest movers, category map; per-product detail
StoresHow each store tracks to targetProgress-to-target gauges, revenue ranking; per-store detail
InventoryWhere stock is wrongOut of stock, below reorder point, overstocked, healthy
AlertsWhat needs attentionAn inbox by severity and type, each with a recommended next step
ActionsWhat's being doneReorders, transfers, markdowns and follow-ups, with an activity timeline
Sales overview dashboard: revenue, units sold, gross margin and open alerts, with a revenue trend against the previous period
Dashboard — today at a glance, every number against the period before.
Best and worst sellers panel: ranked product table with revenue, units, change and marginStore targets panel: progress-to-target gauges for all 10 storesCategory map panel: treemap of revenue by category and productStock health panel: out of stock, below reorder point, overstocked and healthy counts with a stock tableProduct detail panel: revenue, units and margin with a revenue trendInventory health panel in dark mode: waffle chart of healthy, low and out-of-stock lines

From signal to action

The core of the product is one loop: signal → alert → recommended step → action → resolved.

Six alert types, each with a default answer

ALERTRAISED WHENRECOMMENDED STEP
StockoutA product is at zero on the shelfTransfer from a store with surplus, else reorder from the vendor
Low stockBelow its reorder pointReorder back to the shelf maximum
OverstockFar more cover than it needsMark down 20% for two weeks, or transfer surplus
Falling salesSales dropping against the previous periodA 15% promotion and a shelf review with the vendor
Demand surgeSales climbing chain-wideA chain-wide order covering two weeks of new demand
Store declineA store's revenue is downAsk the store manager for a recovery plan

Alerts carry their cost in plain language — “usually sells 30 units a day here, so every day out of stock loses about $99” — so severity is something a leader can feel, not just a colour.

Alert inbox: tabs for needs attention, action taken and resolved; filters by severity and type; a table of alerts with signals and status
Alerts — triaged by severity and type, with tabs for every state an alert can be in.

One click from alert to action

Recommended action panel: an alert's evidence, store, product, vendor and the recommended next stepTake action panel: a pre-filled form with action type, store, product, quantity, due date, owner and noteAction tracker panel: a table of actions with status and an activity timeline
  • Evidence firstAn alert opens on the store, product, vendor and cost — then the recommended next step.
  • Pre-filled actionTake action fills the type, store, product, quantity, due date, owner and note, so accepting the recommendation is one confirmation.
  • A loop that closesCreating an action marks the alert actioned; completing it resolves the alert. Nothing disappears just because someone looked at it.
  • Ready for an AI assistantCopy prompt hands the situation and recommendation to an AI assistant — to draft the vendor email or the note to a store manager.

Built entirely on SagUI

Sales Analytics is the proof case for SagUI: a real, data-dense product assembled from the design system rather than designed screen by screen.

  • ComponentsApp shell, metric cards, sortable data tables, tabs, chip groups, segmented controls, selects, number fields, dialogs, drawers, alerts, badges, toasts, timelines, empty states, skeletons.
  • ChartsLine, bar, donut, gauge, treemap, slope, waffle and activity heatmap — SagUI's own charts.
  • TokensColour, type, radius and elevation from @sagui/tokens, so light and dark mode work everywhere.
  • Shared shellThe same AppShell, sidebar and top bar as the MDM apps — two products that read as one family.
Store detail for Lubbock Tech Park in dark mode: revenue, units, margin and target metrics, revenue trend and inventory health
Store detail in dark mode — the theme switches through SagUI tokens, with no app-level styling.

What this demonstrates

  1. 01

    Systems thinking across products

    Data, interface and decisions as three layers of one system, each doing one job.

  2. 02

    Design for decisions, not displays

    Every alert ends in a next step, and every step can be tracked to done.

  3. 03

    A design system earning its keep

    A six-area, data-dense product built without inventing a single new component.

NEXT →A live feed from the MDM databaseNotify managers and vendors from an actionMeasure the loop: alert to action, actions done on time
NEXT CASE STUDY · SYSTEMSagUI →