Driving digital engagement and repeatable growth through marketing automation and accelerated by Publicis Sapient
TALA: Test-and-Learn Marketing Automation
Test-and-Learn Automation (TALA) helps organizations like quick-service restaurants (QSRs) and other consumer-facing industries improve the impact of marketing campaigns.
Within a few months, businesses get results that help them serve the best offer, content or recommendation to the right customer at the right time. Outside of marketing campaigns, TALA can also be used to improve the customer user experience on mobile apps and sites.
Identify an initial small set of use cases.
Select use cases that are core to the business and likely to improve key performance indicators (KPIs) such as customer retention, visit frequency or basket size. Use cases are obtained from a user experience to drive customer actions like downloading an app, converting to a transaction or redeeming a reward or offer. Businesses should select use cases that can be easily measured within the intended user experience.
Collect data for the use case.
Data should be gathered through a new customer data platform (CDP) or a preexisting data platform.
Setup an analytics environment to:
Enrich customer data. Most CDPs are equipped to apply proven algorithms to populate customer preference and propensity attributes such as behaviors, demographics and history with a brand.
Develop a segmentation strategy. CDPs can also be used to identify and select data attributes across customer behavior, demographics, purchase history, preference and propensity. A robust set of attributes should have values strongly correlated with customer actions taken in the prioritized use case.
Design and execute test-and-learn experiments.
Designing experiments is both a quantitative and qualitative process that combines the knowledge of the business with data to generate test hypotheses that are qualified, prioritized and executable. TALA automates the final test design and preparation steps by generating audiences made up of test and control groups. It then connects these audiences to experiments. Once tests are executed, TALA automates the generation of test performance reports. Performance measurement reports confirm if tests should be scaled, pursued once modified or simply abandoned. Once tests show which offers work, those learnings can be scaled to larger campaigns.
Convert experiments into campaigns.
Multiple tests are run simultaneously to create a repository of results to be used for planning campaigns. In return, campaign planning and performance monitoring generate their own set of questions that are quickly explored and answered. This collaboration generates growth at scale.
Test-and-Learn vs. Artificial Intelligence (AI)
Customer behaviors are often unpredictable. Relying on AI alone to uncover trends in data risks the technology missing new behaviors. Test-and-learn, combined with AI, is more efficient for the following reasons:
- AI generates well-qualified hypotheses, but it cannot predict whether these hypotheses will generate more revenue or better margins. A business will never have enough recent and complete historical data for AI algorithms to provide these insights. Test and learn generates new data through short, high frequency experiments to identify the probability of a profitable outcome.
- Guesswork is not a predictable recipe for success when working with massive amounts of data in a fast-paced, high-transaction environment. Test-and-learn rigorously identifies and validates hypotheses for marketers during campaign planning.
- Test-and-learn eliminates unproductive campaigns and avoids wasted time, resources and money (e.g., discount offers).
- A positive side effect of test-and-learn experimentation is the creation of a repository of proven facts about customer habits, product preferences and trigger mechanisms that also answer “what-if” questions about the business.
A key learning example from test-and-learn came from a quick-service restaurant (QSR) testing the following hypothesis: If customers receive offers they like, will they spend more money? We learned that visitor frequency and number of items purchased generally increased. Interestingly, margins for high-value customers dropped slightly, and margins for customers just below that level went up. In other words, high-value customers would have ordered from the restaurant without offers, and the discount represented lost revenue.
Start simple, scale fast and demonstrate results along the way
Publicis Sapient’s deployment of TALA within various client environments and well-chosen case sets generated the following successful results within months:
- A global QSR saw its revenue grow by millions without margin loss.
- One of the world’s largest logistics companies increased its revenue by more than $200 million.
- A North American airline saw its online bookings increase by double-digit percentage points when applying test-and-learn to its online booking user experience and associated content.
Start simple and scale up fast
In this diagram, the top row shows the quick start path to enable TALA, and the lower row explains how each activity scales up over time.
Why now and why Google?
Accessible and affordable cloud data processing and machine learning technologies are enabling the development of personalized insights on large data sets. The deployment of data-driven analytics and marketing automation capabilities on cloud platforms like GCP can be done in a matter of weeks.
Google offers a variety of analytics tools, including Looker and Vertex, which are directly integrated into their cloud data warehouse, BigQuery. GCP is also an environment for building applications and APIs, and Publicis Sapient is combining these native GCP features into the TALA platform for clients. The benefit of working with GCP is the simplicity of assembling its capabilities for a defined purpose and the deep level of security provided.
Why Publicis Sapient created TALA
To offer organizations a low-risk marketing and data platform to support a high rate of productive campaigns.
To generate test results quickly at a low cost by building on platforms and capabilities already used by an organization or with small, incremental investments.
To use and complement the power of AI to enable personalized offers.
To provide a tool that is attuned to the business environment and can accommodate running a high volume of experiments.
To assist with the development of facts and insights a marketer can use for strategic planning.
To setup an optimization framework for personalization that starts with campaigns and further extends to user experiences.
Start a Conversation
Vp Technology, Publicis Sapient