9.1M
Live orders in production
Case Study
5 Dashboards on 9 Million Orders
A full replacement of a legacy Looker extension — rebuilt in React + TypeScript with 99% fewer API calls and custom chart rendering.
An e-commerce order-management platform was running a Looker extension that triggered 2,400+ API calls per page load, had no layout customisation, and was showing wrong KPIs due to SQL fanout bugs. We rebuilt it from the ground up — 5 dashboards, 9.1M live orders, $585M revenue volume.
E-commerce SaaS · USA·React 18TypeScriptRechartsLooker SDKBigQuery
L
E-commerce PlatformBigQueryLooker SDK
Live • 9.1M orders
Overview
Revenue
Fulfillment
Email
Attribution
99% fewer API calls
Total Orders (90d)
9.1M
▲ 12.4% vs prior
BigQuery live
Gross Revenue
$585M
▲ 8.7% vs prior
excl. refunds
Avg Order Value
$64.28
▼ 3.1% vs prior
all channels
Fulfillment Rate
97.3%
▲ 0.4pp vs prior
shipped/confirmed
API Calls / Load
~11
▲ 99% reduction
was 2,400+
Revenue Trend (90d)
BigQuery this period prior period
▲ +8.7% revenue growth vs prior 90d
Order Status Breakdown
LiveFulfilled
8.86M
Processing
182K
Cancelled
44K
Refunded
14K
SQL fanout bugs fixed — 5 KPIs corrected
Revenue by Channel
RechartsDirect
$241M
Email
$156M
Paid Search
$98M
Organic
$62M
Affiliates
$28M
Custom Recharts rendering — no embed tiles
Orders Volume Trend (90d)
BigQueryJanFebMarApr
▲ +12.4% order volume — peak single-day: 148K orders
5 Dashboards — 83 Tiles
TypeScript| Dashboard | Tiles | Status |
|---|---|---|
| Overview (KPI) | 22 tiles | Live |
| Revenue Analysis | 18 tiles | Live |
| Fulfillment Ops | 17 tiles | Live |
| Email Performance | 14 tiles | Live |
| Attribution | 12 tiles | Beta |
At a glance
Results That Speak for Themselves
99%
API call reduction (2,400 → ~11)
$585M
Revenue volume tracked
83
Tiles across all 5 dashboards
From challenge to outcome
The Problem & Our Approach
A Legacy Extension That Couldn't Scale
The existing Looker extension embedded native dashboard tiles via the Embed SDK — triggering over 2,400 individual API calls per page load. Every page visit was slow, blank-screen interstitials lasted 2-3 seconds, and there was no way to customise layouts or add date presets.
- 2,400+ Looker API calls per page load
- No drag-and-drop layout or date presets
- Blank tiles when underlying tables were empty
- 5 SQL fanout bugs inflating KPIs — wrong numbers shown to customers
- No per-tenant embed security
- 2-3 second blank-screen interstitial on every load
Five Dashboards Rebuilt From the Ground Up
We rebuilt the extension in React 18 + TypeScript using Recharts for rendering instead of Looker native tiles. By batching queries and parallelising SDK initialisation, we reduced API calls from 2,400 down to ~11 per page load.
- Recharts instead of Looker native tiles — full rendering control
- Parallelised Looker Embed SDK init — zero blank-screen interstitial
- 3-hour localStorage cache — instant repeat visits
- Client-side D|M|Y aggregation from day-level data
- React-grid-layout for drag-and-drop personalisation
- Cookieless embed v2 + service account for per-tenant isolation
- Holt-Winters revenue forecast built in
- Graceful degradation on empty tables
What We Achieved
Full extension replacement live in production with 9.1M orders
API calls reduced from 2,400+ to ~11 per page load (99% reduction)
Zero blank-screen interstitial — instant perceived load
Drag-and-drop layout personalisation per user
Graceful degradation on empty tables — no more blank tiles
Per-tenant embed isolation via cookieless embed v2 + service account
Holt-Winters revenue forecast shipped as part of the Summary dashboard
5 SQL fanout KPI bugs identified and corrected
The Five Dashboards
Each dashboard was designed around a specific decision — from executive revenue overview to granular asset tracking.
- Summary (23 tiles) — Revenue, orders, AOV, forecasts
- Shipping & Alerts (16 tiles) — Shipment volume, carrier mix, transit days
- Returns (17 tiles) — Return rate, reasons, ageing
- Asset Retrieval (13 tiles) — QR code usage, asset tracking
- Action Alert (14 tiles) — Email metrics, send volumes
Before vs. After
| Dimension | Before | After |
|---|---|---|
| API calls per page load | 2,400+ | ~11 |
| Blank-screen interstitial | 2-3 seconds | Zero |
| Layout customisation | None | Drag-and-drop |
| Date presets | None | Full range options |
| KPI accuracy | SQL fanout errors on 5 KPIs | Independently validated |
| Embed security | None | Cookieless embed v2 + service account |
Engineering Decisions That Mattered
- Client-side compute fills data gaps without a server round-trip
- 3-hour localStorage cache makes repeat visits instant — no API call at all
- Holt-Winters forecasting) is computed client-side on the fetched day-level data
- Bundle size: 961 KiB — within acceptable range for an authenticated B2B extension
- Parallel Looker Embed SDK initialisation eliminated the blank-screen interstitial entirely
Discuss a similar E-commerce SaaS project
Lets Viz runs a paid discovery audit ($500–1,000, credited toward the project) to scope your requirements, data model, and architecture before writing a line of code.
From Lets Viz
Helping companies make data-driven decisions to achieve their Goals
Experience superior data analysis with our custom dashboards. Specialising in all tools and technologies, we turn data into visually compelling, actionable insights.
NDA-safe · No obligation · Clear next steps