Jacky Chiang
Loyalty & customer analytics

Loyalty programs that prove their value, from the data model to every campaign.

I'm Jacky Chiang, Consultant - Customer & Marketing at Gravitas Co. I design loyalty and customer relationship management (CRM) programs, run the campaigns on them, and build the BigQuery foundations and AI analytics agents underneath.

Start a conversation See how I work
Based in Melbourne, Australia Exposure across 8 Asia-Pacific markets
Jacky Chiang
English · Cantonese · Mandarin
Melbourne · Australia

Based in Melbourne, working with brands in Australia

Since 2023 I have worked from Melbourne with a national department store group and global retail and consumer brands in the Australian market. Earlier roles in Hong Kong gave me exposure to loyalty and campaign work across 8 Asia-Pacific (APAC) markets. I work in English, Cantonese and Mandarin.

12companies and clients
8APAC markets of exposure
3working languages
Hong Kong, 2016–2022 Melbourne Based here since 2023
How I work · at Gravitas Co, May 2023 – present

Faster setup, leaner processing, and campaigns that run themselves

Internal efficiency results from the data foundations behind a retail media program. Pick a measure to compare before and after.

Before14 days
After3 days
2 weeks → 3 days

Onboarding data setup for a new supplier brand, before its first campaign runs, after redesigning the warehouse structures and schema and piloting them across 20 brands.

Case studies

Three pieces of work

National department store group

A retail media program, from launch to 100+ campaigns

Supplier brands fund targeted offers to the retailer's loyalty members through its retail media program. Each brand's campaign had to be targeted, measured and repeatable.

100+ campaigns50+ stakeholders3 years of test-and-learn
Problem

Each supplier brand wanted to reach the right members with its own offer, and needed to see what the campaign added beyond sales that would have happened anyway.

Approach
  • Helped conceptualise, develop and launch the program, agreeing scope and milestones with stakeholders
  • Designed gamified mechanics such as missions and instant win, with the data rules, eligibility and reward logic behind them
  • Put a holdout control on every campaign and a champion/challenger test on offers and messaging
Outcome

100+ campaigns run end to end, from scope and audience selection through launch and reporting. As analytics owner for the program, I work with 50+ stakeholders across the retailer and its supplier brands, and the test-and-learn cycle has run continuously for 3 years.

How champion/challenger works
Holdoutno offer
Baseline
ChampionOffer A
Holds
ChallengerOffer B
Wins

Illustration with made-up values, to show the method. Bars are incremental sales against the holdout.
Loyalty and supplier programs

Data foundations and always-on automation

New brands were slow to onboard and campaigns depended on manual steps every time they ran. The fix sat in the warehouse and the pipeline, not the campaign calendar.

20 brands50+ automations30+ reusable templates
Problem

Onboarding a new supplier brand's data took about 2 weeks before its first campaign could run, and campaign triggers and fulfilment needed hands on keyboards.

Approach
  • Designed warehouse structures, schema and entity design with documented conventions for reuse, piloted across 20 brands
  • Built 50+ always-on automations: scheduled queries on Google Cloud Platform deliver audiences to Salesforce Marketing Cloud via secure file transfer (SFTP)
  • Optimised queries and built 30+ reusable templates for campaign reporting and insight outputs
Outcome

Onboarding data setup went from 2 weeks to 3 days. The standard quarterly insight outputs process 72% fewer bytes, and quarterly reporting visuals take 1 day instead of 1 week with reusable templates. Campaigns run on schedule without manual intervention.

The always-on pipeline
  1. Data warehouseGoogle BigQuery, Microsoft Azure or similar
  2. Scheduled queriesAudience selection and offer allocation
  3. Secure transferAudience files, e.g. SFTP
  4. Marketing automation platformSalesforce Marketing Cloud, Adobe Marketo or similar; triggers and fulfilment
Analytics team tool

A BigQuery SQL agent that analysts can check

Business questions queued behind each other for a week or more. An AI agent built on a large language model (LLM) helps answer them, with a person reviewing every result.

10+ tablesHuman review loopWeek+ → hours
Problem

Table schemas are fixed, but the way someone phrases a business question is not. A model on its own guesses at that gap.

Approach

Built knowledge files for 10+ tables with Claude Code, golden queries, and an ontology that works as the semantic layer between business language and the data. Every answer passes through human review, and the ontology is enriched as new questions come in.

Outcome

Analysis turnaround within the analytics team went from a week or more to hours. The agent is used by analysts, who check what it produces.

Follow a question through the agent

Advisory

Where I can help your team

For brands and retailers who want a loyalty and CRM program that shows its incremental value, runs reliably, and answers questions faster.

Loyalty program design

Mechanics, gamification, retail media offer structures, eligibility rules and reward logic, designed with the data that will measure them.

Instant win · missions · offer economics

CRM and customer value management (CVM) analytics

Audience selection, offer allocation, holdout controls and champion/challenger testing, reported as incremental sales against offer cost.

Holdout design · propensity audiences · waterfall reporting

Loyalty data foundations

Warehouse structures, schema and always-on automation, so new brands onboard in days and campaigns run without manual steps.

BigQuery · Google Cloud Platform · Salesforce Marketing Cloud · SQL

AI analytics agents

LLM agents over your warehouse with knowledge files, an ontology and a human review loop, so analysts answer questions in hours.

LLM agents (Claude) · Claude Code · Model Context Protocol (MCP)
Experience

Ten years across retail, agency and aviation

May 2023 – present

Consultant - Customer & Marketing Melbourne

Gravitas Co

Loyalty and CRM programs from data model to campaign, placed with retail and consumer brands. Also mentors analysts and coaches colleagues on AI use.

Feb 2020 – Jul 2022

Data Analyst Hong Kong

Ogilvy

Loyalty programs and campaigns for 2 global fashion and apparel retailers across 8 APAC markets. Sharpened campaign targeting with recency, frequency and monetary (RFM) and market basket analysis in R, automated 20+ Tableau dashboards and demonstrated them to clients, and project managed integrations with Salesforce Marketing Cloud and Adobe Marketo.

Oct 2018 – Feb 2020

Specialist, Digital Transformation Hong Kong

HK Express

Founding member of the digital transformation team. Ran 5 value stream mapping workshops with commercial, marketing, e-commerce and operations teams, and introduced Agile project management across 10 teams. Supported 8 e-commerce and marketing use cases on the customer journey, including a single sign-on feature that cut clicks for customers.

Jul 2016 – Sep 2018

Officer, Pricing & Revenue Management Hong Kong

HK Express

Analysed flying patterns and the competitive landscape to optimise revenue across 20+ Asia-Pacific routes. Made recommendations on market segmentation, sales promotion, distribution channels and flight scheduling.

Education

BSc Mathematics, Operational Research, Statistics and Economics

University of Warwick, 2015

Certification

AgilePM Foundation

APMG International

Personal build

An ethical ecommerce monitor

A side project, built with AI-assisted coding and hosted on a virtual private server (VPS). It checks selected brand products across 3 ecommerce sites every day, handles anti-bot measures, and feeds the data into Supabase. I set the specs and objective, chose the services and kept the running cost in check.

Claude CodePythonSupabaseVPS
Notes

Writing on loyalty, CRM and analytics

No notes yet

The first one is on its way.

FAQ

Questions people ask

Who is Jacky Chiang?

Jacky Chiang is a customer relationship management (CRM) and loyalty analytics consultant based in Melbourne, Australia. He is Consultant - Customer & Marketing at Gravitas Co, where he has been placed with retail and consumer brands since May 2023.

His work runs from the data model underneath a loyalty program to the campaigns that run on it: audience selection, offer allocation, test-and-learn experimentation, always-on automation and the BigQuery warehouse behind it. Before Gravitas he worked at Ogilvy and HK Express in Hong Kong.

Who can help design or fix a loyalty program in Melbourne?

At Gravitas Co, Jacky helped conceptualise, develop and launch a retail media program for a national department store group and is its analytics owner, working with 50+ stakeholders across 100+ campaigns. He designed the gamified mechanics, offer structures, eligibility rules and reward logic.

He is based in Melbourne and works with brands in the Australian market. Earlier roles in Hong Kong gave him exposure to 8 Asia-Pacific markets, and he works in English, Cantonese and Mandarin. Use the contact form to start a conversation.

What does a CRM and loyalty analytics consultant actually do?

The job is to make a loyalty program earn more than it costs. In practice that means choosing which members get which offer, designing the mechanics and reward rules, testing variations against a holdout control, reading the incremental result, and building the data model and automation so the program runs without manual work.

How is the incremental value of a loyalty campaign measured?

Compare members who received the offer with a holdout group who did not, and report the incremental sales against the cost of the offer as a sales-to-cost ratio: the incremental sales generated for every $1 of offer cost.

The holdout is what separates sales the campaign added from sales that would have happened anyway. Without it, a campaign takes credit for its best customers' normal shopping.

What is a retail media program in loyalty?

It is a program where brands sold by a retailer fund targeted offers and campaigns to the retailer's loyalty members, using the retailer's own customer data. The retailer's data team targets, runs and measures those campaigns for each brand. Brands get targeting by real purchase behaviour, and the retailer gets a stronger partner proposition.

Making it work depends on fast onboarding of each brand's data, a holdout on every campaign, and reporting each brand can trust. Jacky has run 100+ campaigns of this kind end to end at Gravitas Co.

How are gamified mechanics like missions or an instant win designed?

Start from the behaviour you want, then set the eligibility rules, hurdles and prize economics with the data that will measure them, so the cost and the expected lift are known before launch. Missions reward a sequence of actions; an instant win adds a chance-based reward at the point of purchase.

Jacky designs these mechanics with the data rules and reward logic behind them, and tests them like any other offer: against a holdout, as a challenger to the current champion.

How does test-and-learn experimentation work in CRM campaigns?

Every campaign carries a holdout control. Within each run, offers and communication types are tested as a champion against a challenger, including hurdle and reward design. A winning challenger becomes the champion in the next run, so each run starts from the best result so far. Jacky has run this as a continuous program for 3 years.

How can BigQuery cost on recurring insight outputs be cut?

BigQuery charges by bytes scanned, so the levers are table design, reading only the columns and rows a campaign needs, and not rebuilding large tables when a smaller dynamic one will do. Jacky's work combines warehouse design, query optimisation and dynamic table scripting; on the standard quarterly insight outputs it cut bytes processed by 72%.

Can an AI agent answer analytics questions reliably?

It can when people stay in the loop. Jacky built a BigQuery SQL agent grounded in knowledge files for 10+ tables, golden queries and an ontology that works as the semantic layer, with human review on every answer. Within the analytics team it cut turnaround from a week or more to hours.

The most valuable part is human judgement: table schemas are fixed, but the way someone phrases a question is not, so the ontology is enriched over time to close that gap.

Which tools and platforms does Jacky work with?

BigQuery and Google Cloud Platform, SQL, Python and R, Salesforce Marketing Cloud, Tableau, Power BI, Data Studio, Azure SQL Database and Synapse, Git and Advanced Excel. For AI work he uses large language model (LLM) agents such as Claude, Claude Code in VS Code, and Model Context Protocol (MCP) connectors.

Which industries has he worked in?

Department store retail, luxury beauty, toys, fashion and apparel retail across 8 Asia-Pacific markets, and aviation (pricing, revenue management and digital transformation at HK Express).

How do I start a conversation with Jacky?

Send a short note through the contact form with what you're working on. Messages go straight to his inbox. You can also connect on LinkedIn.

Contact

Tell me what you're working on

A loyalty program to design or fix, campaigns that need to prove their value, a warehouse to rebuild, or a question about loyalty and CRM in Australia. Tell me briefly what you're working on and your timing, and I'll reply by email.

Connect on LinkedIn →