Data Analytics Services in Melbourne, Australia

Retail / E-commerce

A Melbourne-based multi-brand retailer (4 retail brands, 45 physical stores across Victoria and NSW, 3 e-commerce sites, wholesale channel, A$85M annual revenue) was making strategic decisions with incomplete and delayed data. The problems: (1) Data silos: each brand operated on different systems — Brand A on Shopify (e-commerce) + Lightspeed (POS), Brand B on Magento + Vend, Brand C on Shopify + Square, Brand D (wholesale only) on MYOB. Customer data: fragmented across 7 different systems. A customer who bought Brand A online and Brand C in-store: appearing as two separate customers — no cross-brand view. Inventory: each brand managing inventory independently — Brand A having excess stock in a product category where Brand C had stockouts. No visibility across brands. Financial data: each brand reporting separately in different formats — the CFO manually compiling a consolidated view in Excel, a process taking 3 days per month. (2) Reporting delays: monthly performance reports: produced manually by the finance team, available 12-15 business days after month-end. The reports: backwards-looking summaries of what happened — not forward-looking insights about what to do. The CEO: receiving January's performance report in mid-February — by which time, the information was too stale to inform February decisions. Weekly reports: produced by each brand manager in their own format — inconsistent metrics, inconsistent definitions (Brand A defining "conversion rate" differently from Brand C), making cross-brand comparison impossible. (3) Marketing black box: the retailer spending A$1.8M annually on digital marketing (Google Ads, Meta, Instagram, email) across 4 brands. Marketing ROI: unmeasurable. The marketing team could report clicks and traffic but couldn't connect ad spend to actual purchases (because purchase data was in the POS/e-commerce system, not the marketing platform). The question "which marketing channels drive profitable customers?" — unanswerable. During peak periods (Boxing Day, End of Financial Year, Black Friday): marketing budget allocated based on gut feeling rather than data. Post-campaign analysis: manual, approximate, and completed weeks after the campaign ended. (4) Inventory inefficiency: the retailer holding A$12M in inventory across 45 stores and 3 warehouses. Stockout rate: 8% (items customers wanted but couldn't buy — lost sales). Overstock rate: 15% (items marked down to clear — margin erosion). Both problems occurring simultaneously: overstocked in some locations while understocked in others. The buying team: ordering based on last year's sales plus intuition — no demand forecasting, no location-level inventory optimization. Markdown decisions: made late (when the buying team noticed excess stock during store visits) rather than early (when data showed sell-through rates below plan). (5) Customer blindness: the retailer collecting loyalty program data (280,000 members across all brands) but not analysing it. Customer acquisition cost: unknown. Customer lifetime value: unknown. Customer retention rate: unknown. Customer cross-brand purchase behaviour: unknown. The loyalty program: functioning as a discount mechanism rather than a customer intelligence platform. The CEO: "We have 280,000 loyalty members. I can tell you how many there are but not who they are, what they want, or how to keep them."

We built a unified data analytics platform over 8 weeks: (1) Data extraction: Fivetran connectors for: Shopify (2 stores — orders, products, customers, inventory), Magento (1 store — same data points), Lightspeed POS (Brand A stores), Vend POS (Brand B stores), Square POS (Brand C stores), MYOB (Brand D financials + consolidated chart of accounts), Google Ads (4 brand accounts), Meta Ads (4 brand accounts), Google Analytics 4 (3 e-commerce sites), and Klaviyo (email marketing — 4 brand accounts). Data flowing to BigQuery (Google Cloud, australia-southeast1 region — data remaining in Australia). Sync frequency: hourly for POS and e-commerce (near-real-time sales visibility), daily for marketing and financial data. (2) Unified data model (dbt): customer identity resolution — matching customers across 7 systems using email address, phone number, and loyalty card number. 280,000 loyalty members resolved into 245,000 unique individuals (35,000 duplicates — same person with different email addresses across brands, or loyalty cards created twice). Cross-brand purchase behaviour now visible: 18% of customers purchasing from 2+ brands (previously invisible). Product master — unified product catalogue across all brands, with consistent categorisation (category, subcategory, season, price band) enabling cross-brand product analysis. Financial consolidation — chart of accounts mapped across all brands, enabling consolidated P&L, brand-level P&L, and store-level P&L from a single data model. Standard calculations — gross margin (after cost of goods, freight, and duties), marketing cost per acquisition (ad spend attributed to first purchase), customer lifetime value (12-month forward-looking CLV based on purchase history and predicted retention), and inventory health metrics (weeks of cover, sell-through rate, aged stock percentage). (3) Marketing attribution: multi-touch attribution model connecting ad spend to purchases — tracking the customer journey from ad click → website visit → purchase (including store visits linked to digital touchpoints via loyalty card). Attribution logic: last-click attribution as the primary model (simple, defensible), with data-driven attribution as a supplementary view (Google's algorithmic attribution providing weighted credit across touchpoints). Marketing ROI by channel: the first time the marketing team could see which channels generated profitable customers (not just traffic). Finding: Google brand search ads (customers searching for the brand name) had 12:1 ROI, while broad Google Shopping ads had 1.8:1 ROI — the marketing team had been allocating equal budget to both, unaware of the massive difference. Meta prospecting ads: generating high traffic but low conversion rate (0.3% vs. 2.1% for Google). The high spend: driven by impressive click metrics that didn't translate to purchases. (4) Inventory analytics: demand forecasting model (time-series forecasting using Prophet — trained on 3 years of sales history by product category, store location, and week). Forecast incorporating: seasonality (Australian retail calendar — Boxing Day, EOFY, Mother's Day, Father's Day, Easter, school holidays), trend (growing and declining categories — athleisure growing 15% YoY, formal wear declining 8% YoY), and external factors (weather — Melbourne's cold weather driving coat sales in predictable patterns, promotional calendar — known sale dates boosting volume). Inventory optimization: recommended stock levels by product category and store — minimising both stockouts (lost sales) and overstock (markdowns). Markdown recommendations: automated alerts when a product's sell-through rate falls below plan — recommending early markdown timing and depth rather than waiting until end-of-season. (5) Customer analytics: RFM segmentation — 245,000 customers segmented into: VIP (top 5% by spend — 12,250 customers generating 35% of revenue), Active (regular purchasers — 48,000 customers), At-risk (previously active, declining frequency — 31,000 customers), Lapsed (no purchase in 12+ months — 82,000 customers), and New (first purchase in last 90 days — 15,000 customers per quarter). Customer lifetime value: calculated for each customer — enabling the marketing team to determine: how much to spend acquiring a customer in each segment, which customers to invest in retaining (VIP and at-risk), and which lapsed customers to attempt reactivation (based on predicted response rate and CLV). Cross-brand analysis: 18% of customers purchasing from 2+ brands — these cross-brand customers having 2.8x higher CLV than single-brand customers. This insight: informing a cross-brand marketing strategy (encouraging Brand A customers to try Brand C, and vice versa) that had never been attempted because the data to identify cross-brand opportunity didn't exist. (6) Dashboards: CEO dashboard: consolidated revenue (all brands, all channels), YoY comparison, brand performance ranking, top-line KPIs (revenue, margin, customer count, average order value) — updated daily (vs. 12-15 days previously). The CEO checking this dashboard every morning on his phone. Brand manager dashboards: brand-specific performance — sales by store, sales by category, inventory health, marketing ROI, customer metrics — enabling each brand manager to make data-driven decisions independently. Marketing dashboard: spend, impressions, clicks, conversions, and ROI by channel and campaign — updated daily. The marketing team: reallocating budget weekly based on performance rather than waiting for monthly reviews. Buying dashboard: inventory health by category and store — weeks of cover, sell-through rate, aged stock alerts, demand forecast vs. actual — enabling the buying team to adjust orders and markdown timing proactively.

Reporting speed: from 12-15 business days after month-end to next-day automated reports. The CFO: reclaiming 3 days per month previously spent on manual report compilation. The CEO: making decisions based on yesterday's data rather than last month's data. Marketing ROI: first-ever measurable marketing ROI across all channels. Budget reallocation based on data: shifting A$320,000 annually from low-ROI Meta prospecting to high-ROI Google brand and category search. Estimated impact: A$180,000 incremental revenue from the same marketing spend (higher-converting channels receiving more budget). The marketing director: "For three years I've been asking for ROI data. Now I have it, and it's changing every decision we make." Inventory efficiency: stockout rate from 8% to 4.5% (improved availability driving incremental sales). Overstock rate from 15% to 9% (earlier markdown decisions preserving margin). Combined impact: estimated A$620,000 annual margin improvement — reduced markdowns (A$380,000) plus captured sales from reduced stockouts (A$240,000). The buying team: using demand forecasts to adjust orders 4-6 weeks before the season rather than reacting mid-season. The head of buying: "The forecast isn't perfect, but it's much better than guessing. And when it's wrong, I can see it early enough to adjust." Customer intelligence: cross-brand customer identification enabling a targeted cross-brand campaign — emailing Brand A customers about Brand C products (based on category affinity). Campaign result: 12% of targeted Brand A customers making a first Brand C purchase within 30 days (vs. 2% baseline cross-brand conversion). VIP retention program: identifying the 12,250 VIP customers (35% of revenue) and implementing targeted retention (exclusive previews, priority access, personal styling) — VIP retention rate increasing from 68% to 79% in the first 6 months. At a 2.8x CLV premium for cross-brand customers: the cross-brand and VIP programs contributing an estimated A$1.2M in incremental annual revenue. Self-service analytics: brand managers creating their own ad-hoc analyses — exploring data without waiting for the finance team. The number of data requests to the finance team: decreased 65% (brand managers answering their own questions). The finance team: spending time on strategic analysis rather than report production. Development cost: A$95,000 (8-week engagement including data architecture, pipeline development, modelling, dashboards, and training) + A$24,000/year (Fivetran, BigQuery, and Looker licensing). Annual value: A$620,000 (inventory efficiency) + A$180,000 (marketing optimisation) + A$1.2M (customer intelligence programs) + CFO time recovery + faster decision-making. ROI: analytics investment recovered within first 6 weeks from inventory efficiency improvements alone.