How MECCA is mirroring in-store customer experiences online—with dramatic results

Founded in 1997, MECCA brings together some of the world’s most coveted beauty and skincare products, selling over 120 brands to two million customers in Australia and New Zealand. It represents 10% of Australia’s $4.2 billion beauty market and around 25% of the prestige beauty sector, according to IBISWorld.

One of the secrets to MECCA’s success is its high-touch, high-service beauty boutiques. According to Vogue, MECCA spends around 3% of its revenue training staff to offer tailored, personalised advice to every customer.

MECCA is now looking to replicate this unique face-to-face experience online. “When you go into a MECCA store, the personalised service and recommendations you get from our hosts is something that we want to translate into the online world,” explains Lauren Shepherd, Head of CRM and Loyalty at MECCA.

MECCA is an early adopter of e-commerce, launching online shopping in 2001. It turned to its strategic partners, Amazon Web Services (AWS) and Servian, for help leveraging data and analytics to improve customer engagement online.

Delivering a hyper-personalised product and a bespoke journey are at the core of improving customer experience (CX) to drive business growth. Yet according to Gartner, over 70% of CX leaders “struggle to design projects that increase customer loyalty and achieve results”.

MECCA’s data transformation journey began in 2018, when it worked with AWS to build a data platform and employed Tableau as the business intelligence reporting tool.

After seeing a positive uptake across the business, MECCA’s technology team wanted to push the envelope by using data to add more value for customers. Their end goal is to enable different parts of the business to self-service – a form of business intelligence that enables anyone to drill into data that is relevant to their role. 

MECCA’s CRM and Loyalty teams were especially eager to use data to predict which products are most likely to appeal to different customers.

“We had lots of information about our customers but we weren’t really using it to optimise their experience – each customer would receive the same content and promotions,” says Shepherd.

The MECCA data team in partnership with Servian commenced a trial of Amazon Personalize, a machine learning service that provides a simple framework to build and run personalisation models. It is used by innovative retailers to provide specific product recommendations, personalised product re-ranking and customised direct marketing.

Before it could transition to AI-powered personalised marketing, MECCA needed to cleanse and restructure 23 years’ worth of data contained within its email platform to create an organised, clean and easily accessible data system.

This began the ‘Discovery’ stage of MECCA’s ML journey. An interdisciplinary technology team of CRM, data and Servian experts came together to develop a Proof of Concept (POC). It was their job to anticipate and address any roadblocks that would derail the POC, which uses Amazon Personalize to provide product recommendations as part of a targeted mascara marketing campaign. By utilising existing metadata, the campaign targets customers 90 days after their last purchase with mascara promotion e-mails.

Moving into the ‘Delivery’ stage, the technology team made multiple changes such as migrating to an Amazon S3 data environment and building new integrations to the marketing system to allow for automated end-to-end data processing. They also spent time helping executive stakeholders and managers understand how personalisation and the shift to machine learning will positively benefit their respective areas of the business.

This was key to the POC’s success, according to Paul Erskine, IT Delivery Manager for Digital and Data at MECCA: “Many executives had reservations about the complexity of data science in general. They had questions like: ‘Who will manage the model if someone leaves the business? What is the cost versus value? Who will support it over time?”

To address these concerns, the technology team shared their vision and plans at a data governance forum. They revealed the product recommendations generated by Amazon Personalize, and explained how data science can be used to optimise customer conversion and engagement rates. Their evidence was so compelling, MECCA’s executive team gave the POC a resounding green light.

When Amazon Personalize was released in Australia at the end of 2019, MECCA and Servian immediately deployed it – one of the first Australian companies to do so. Within hours, it began to produce tailored product recommendations, and today, it produces product recommendations across MECCA's entire catalogue.

“One of the benefits of using Amazon Personalize is the ease of training custom models using existing data on AWS managed services. This allows developers, not just data scientists, to build recommendation algorithms,” explains Erskine.

MECCA also deployed a long short-term memory (LSTM) propensity model to identify the best timing for product replenishment. In AB testing, 50% of emails contained personalised product recommendations while 50% did not. Those with personalised recommendations achieved significantly higher conversion rates.

“Since integrating Amazon Personalize, we are seeing a significant increase in e-mail click-through rates and an increase in email revenue relating to the products recommended,” says Sam Bain, Chief Digital Officer at MECCA.

MECCA is now running its personalisation model weekly for all active customers, generating more than 10 million product recommendations every week across all marketing campaigns.

Amazon Personalize is also outperforming MECCA’s previous system for implementing product recommendations using native capabilities from its email management system.

According to Shepherd: “We tested the Amazon Personalize recommendations against system recommendations from our email provider. In theory, the recommendations coming through the email platform are also based on purchase history, but they don't take into account as many measures as the Personalize model, making it less effective.

“We have really proven that by showing our customers products that are relevant to their life stage, their journey, and their purchase history, they're far more likely to convert.”

MECCA continues to work with AWS and Servian to reimagine the digital experience and delight loyal customers. Its goal is to leverage the power of machine learning to forecast what customers will like, and optimise MECCA’s ability to meet demand – all while improving its underlying data set to build ever-more predictive models.