Multi-touch attribution using Tapestry Passport for an industry-leading airline

CASE STUDY

Multi-touch attribution using Tapestry Passport for an industry-leading airline

Download Case Study

You’re restricted from further downloading resources.

Abstract

Multi-touch attribution (MTA) is undoubtedly a more effective approach than traditional heuristics for measuring marketing efficacy. However, implementing MTA presents challenges across martech, mathematics, and data engineering that combine to produce large technical hurdles that must be overcome:

Martech: custom tagging, consistent data collection
Data Engineering: ETL for large amounts of data, developing an appropriate data architecture
Mathematics: model selection and assumptions

Our approach leverages Softcrylic’s expertise in these three areas, resulting in an MTA solution that is durable and elegant in its simplicity, relying on readily available and easily understandable technology.

Overview

Our client is an industry-leading airline with thousands of advertising campaigns spanning multiple channels and triggering billions of server hits per month. They have a profound need for optimizing their marketing strategies and budget allocation across channels and campaigns. Relying on last-click attribution had become unsustainable, prompting their decision to pursue the development of a custom MTA solution.

Our client determined that Softcrylic’s range of digital expertise and central role in their digital marketing ecosystem put us in a unique position to develop an MTA solution that was tailored to fit their needs. Our Digital Analytics, Data Engineering, and Data Science practices gathered requirements, made recommendations, and implemented Softcrylic’s MTA solution, Tapestry Passport, that is now used as a primary measurement tool for our client and their advertising partners to guide and optimize digital strategy.

Implementation

Digital Analytics

An MTA solution relies heavily on accurate and consistent data collection. To that end, we created a system to collect advertisement touchpoint data from behind our client’s domain.

First-party cookies have longer lifespans, leading to higher fidelity identity resolution.
By our client “owning” the data collection mechanism, they are the first to know if there is a problem, resulting in faster error correction and minimizing data loss.
An organization owning its data opens up tremendous possibilities in customer modeling and data activation outside of MTA.

Data Engineering

A large organization can easily generate billions of server calls per month from its website, digital advertising, and email campaigns. To build an MTA model, these billions of data points must be transformed and distilled into distinct, contiguous user journeys. Our Data Engineering team utilized modern data processing tools to make this process as fast, efficient, and inexpensive as possible.

Spark is used to process the large server log files into a SQL database.
The Data Engineering and Data Science teams collaborated to design our MTA SQL database schema to be as efficient as possible for aggregating user journeys.

Data Science

Once the data has been collected, cleaned and transformed, it must be aggregated into customer journeys which serve as inputs for mathematical models that calculate individual touchpoints’ impacts on conversion probability. To that end, our Data Science team set up a cloud computing environment to:

Process in parallel large numbers of consumer journeys segmented by audience, timeframe, and conversion type.
Aggregate customer journeys (“pathways”) for consumption into an open-source package for statistical conversion attribution.

While there are many options for MTA modeling, open source packages have the advantage of transparency to us, our client’s team, and their partners – there is no black box.