Getty Images/iStockphoto

Marketing agency uses Looker to build data platform, grow

When overhauling its analytics operations, a digital marketing agency turned to analytics vendor Looker's tools to build data marts and enable self-service exploration and analysis.

The Looker analytics platform enabled digital marketing agency Wpromote to build data marts and embed customized business intelligence in the workflows of its employees and clients to foster data-driven decision-making.

Founded in 2001 and based in El Segundo, Calif., Wpromote grew quickly and is now among the largest independent digital marketing agencies in the U.S. From 2020 through 2022, its average deal size increased 225%, with its clientele now including more large enterprises whose needs are more complex than those of small and medium-sized businesses, according to the company.

To meet those needs, Wpromote requires data. But the amount and complexity of data is increasing as an ever-growing number of sources are creating it.

We need a lot from our data. We need to integrate data from many different sources across many different clients, and democratize data science for our teams so more people can access advanced capabilities.
Paul DumaisCTO, Wpromote

"We need a lot from our data," said Paul Dumais, Wpromote CTO, on Oct. 11 during a breakout session at Google Cloud Next '22, a virtual conference hosted by the tech giant. "We need to integrate data from many different sources across many different clients, and democratize data science for our teams so more people can access advanced capabilities."

And those needs are becoming even more important as Wpromote expands to work with more enterprise clients that have complex data needs and multichannel marketing strategies, he continued.

To meet Wpromote's rising need for data and deal with its expanding intricacies, the agency theoretically could have hired reams of data experts. But that would have come at great cost, and experienced data scientists are hard to find.

Instead, Wpromote determined that process automation was the key to dealing with the agency's growing data needs. It needed to automate repeatable tasks and also the delivery of personalized data to employees dealing with Wpromote's varied clients, as well as to the clients themselves.

So three years ago, Dumais and his team began developing Polaris, a data platform built in part with tools from the Looker platform.

The results have been increased efficiency, more trustworthy data and more informed data-driven decisions in real time, according to Dumais.

"My first directive when I was brought on board [three years ago] was to automate everything," Dumais said. "Automation would free up time for human critical thinking, and that would lead to better outcomes for our customers and for the company."

Wpromote CTO Paul Dumais speaking at a virtual conference
Wpromote CTO Paul Dumais speaks during Google Cloud Next, a virtual conference hosted by the tech giant.

Building Polaris

Upon arriving at Wpromote in 2019, Dumais discovered that the company's employees spent copious amounts of time wrangling data and producing reports.

In addition, he realized that employees were using different versions of the company's data to inform their decisions rather than a single, consistent version of the data. As a result, they couldn't trust that the data was accurate.

Therefore, Wpromote started to build Polaris with the development of an automated data pipeline that could upload data from its sources into a single location where one -- and only one -- version of the data could reside and be used for analysis.

"If I could automate the process of getting all the data into a single source of truth, it would have a huge impact on our business," Dumais said. "The complete solution would break into four basic functions -- get the data, analyze the data, explore the data and take action."

Dumais determined, however, that using a general extract, transform and load tool built by a platform vendor wouldn't meet Wpromote's needs.

Given that the agency has hundreds of customers, each with unique goals and methods, and hundreds of employees assigned to work with various -- but far from all -- clients, the data pipeline needed to be able to move data from disparate sources into a single location while keeping the data unique to each client. In addition, it needed to keep up with the exponential growth in the amount of data being created.

So Dumais and his team built a customized data pipeline.

"We needed greater control over the data formats and workflow. We needed to be able to quickly add custom data sources in the future," he said. "And I knew it was the first stage in an automation pipeline, so it had to be consistent and predictable."

Next, Wpromote needed a data warehouse.

The company's data scientists were already using Google BigQuery, and after some exploring, Dumais said he determined that BigQuery was ideal for Wpromote's needs.

Beyond a data warehouse, however, Wpromote needed to build data marts, which are smaller data repositories within the data warehouse dedicated to specific businesses and users. It was the data marts that would enable Wpromote's employees to address the specific needs of different customers.

"The data mart is the most important piece," Dumais said. "It's the solution to the original challenge -- a single source of truth that allows our teams to access our customers' data that is both consistent, so we can automate, and uniquely represents [the client's] business."

Like the data pipeline, Dumais planned to build the data marts with his team.

But then he found the Looker platform.

Role of Looker

In developing data marts, Wpromote needed to build hyperspecific data repositories that could be updated using code, and it needed to create a user interface including data visualizations so that employees could explore the data.

"We needed to make it simple for people and machines to make business decisions based on the data we were collecting in our data warehouse," Dumais said.

The Looker platform provided the means to do just that, he continued.

Specifically, the Looker platform enabled Wpromote to model its data with LookML, which is a language for querying data; create dashboards for exploring and visualizing data; and set user management access controls so that data is used safely and securely.

But despite all that Looker could do, Dumais was dubious at first.

He didn't want to entrust too much of the performance and development of Polaris to one vendor, fearing that the vendor's tools might not evolve fast enough to meet Wpromote's quickly expanding data needs.

"I wanted to build a solution with no ceiling -- the only limitation would be our vision and our ability to execute," Dumais said.

And the Looker platform actually enabled that, he said.

As the platform's partnerships with rival vendors such as Tableau and Sisu demonstrate, one of Looker's attributes is openness. If a customer wants to enhance their deployment of Looker with the capabilities of another vendor or a custom-built tool, Looker will work with them.

That openness convinced Dumais to ultimately choose Looker to develop Wpromote's data marts and meet its BI needs, including embedding analytics into custom applications.

"Looker, out of the box, is a very powerful and complete product, and it can be a turnkey solution for many companies," he said. "I can get started with the default behavior and then build on top of it. I can customize it and I can extend it. It aligned with my designs for the Polaris platform."

Now, Polaris is a complete data platform with a custom-built automated data integration pipeline, BigQuery for data warehousing, and Looker for managing data marts and self-service exploration and analysis of data.


The primary result of Wpromote's development of Polaris is greater efficiency, according to Dumais.

In addition, Dumais noted that since there is only one version of a client's data and the integration of that data is automated, decisions are more accurate than before the development of Polaris.

"A business challenge that used to take hours for many hands to deliver can now be done by one person in less than 15 minutes, and we can do it more accurately and consistently than we did before," Dumais said. "Looker freed up time and resources that we could use elsewhere."

Dig Deeper on Business intelligence technology

Data Management
Content Management