Pioneering Automation in The Boutique with Signal Hub

Blog Post
May 14, 2018

The Boutique continues to nurture website visitors by providing valuable content in the right moment. We strive to do this in the most human way possible, just as if we were in a real brick-and-mortar store, helping visitors face to face. To understand our visitors in the digital world, we must pick up on every signal we can find. We find these signals by sorting through web traffic data using a few different tools, all providing different pieces of data. Further information is found through reading press releases, the news, and social media. Every signal, no matter how small or irrelevant it may seem, is part of the buyer's journey. Each buyer's journey, at a company level, is mapped by the Boutique team into an excel spreadsheet. 

In my last post, Meeting Visitors “Face-to-Face” Thanks to Automation, I discussed how we have been working with our partners at Wipro to automate this manual process of mapping company website visits through an automation platform called Signal Hub. Signal Hub is a powerful data analytics platform that will create the same Excel spreadsheet report that the humans in the Boutique currently produce. Not only will it map the journeys in the correct format, but Signal Hub will begin to analyze the journeys for us as well by classifying browsers, shoppers, and buyers based on specific patterns and algorithms. In other words, Signal Hub will start doing the thinking for us, and it will do it 100 times faster! 

Shifting from manual to an automated process will allow the Boutique to have more time to focus on personalized outreach efforts based on the signals we uncover. In addition, automation will increase our efficiency and ultimately scale the operation, allowing us to have more quality conversations. 

Our partners have worked tirelessly—and I mean tirelessly—to begin automating the work we do in the Boutique. For example, one of them is located in India—11-and-a-half hours ahead of us. You can imagine that aligning with them on daily phone calls has proved to be a challenge. But, we persisted, covering every single detail that needed to go into setting up the automation process. Fast forward 3 months, lots of phone calls, even more emails, and a constant consumption of coffee, and we were ready for our first Signal Hub report.  

Cue the report—and the crash

We received an email with a Signal Hub report attached in a link. Of course, the file was too large to send via email. We clicked the link to being downloading. The excitement was exhilarating! Everyone was eager to see how the first report turned out after our months of preparation. Loading…loading…loading. Spinning circle of death.

The report had pulled in so much activity, 23,000 Excel tabs to be precise, that the computer crashed. Our previous manual mapping pulled 400 tabs, so downloading that file was already painfully slow. Now we're pulling signals for every single company that has been to the website in the past year. I think you get the point.  

Our solution: A daily Boutique summary 

As a solution, we decided that it wasn’t necessary for every company to be mapped. For example, a company that comes on to our homepage one time for 3 seconds, leaves, and doesn’t return doesn’t need to be mapped. That is not a journey indicating they are having a having a conversation with us and therefore would require no action by the Boutique. 

Additionally, we will download a daily summary file from the Signal Hub platform to get a high-level overview of company activity. The report will pull activity for the prior year, providing us with the following information: company name, last visit, time on page, and number of page views. This summary will still be massive, with thousands of rows—but will not include each individual buyer's journey. Furthermore, by sorting the document, we can pull the most active recent visitors towards the top of the file. We will manually condense the names of all of the companies we would like to investigate further into a separate file. As soon as we have identified the current website visitors engaging in a deep digital conversation, we can then upload this list back into Signal Hub, and we will receive a report of each company's journey map in under an hour. 

To put this new process into perspective: when we manually mapped our company activity, we could get about two maps completed in an hour. With the new process, we can now get about 50 maps in an hour, substantially scaling our process—increasing the velocity at which we are able to have relevant conversations with our visitors.

The human brain versus the computer brain 

After going back and digging into the data, we found another major issue: we had not accounted for variable company names. For example, Signal Hub was pulling website activity in for company "XYZ." Simultaneously, it pulled different website activity for "XYZ Inc." It would be clear to the humans in the Boutique that "XYZ" = "XYZ Inc.," and we would have combined the website activity into one comprehensive journey. However, for Signal Hub, these two names were recognized and mapped as two entirely separate entities, and consequently, two separate journey maps were produced. Scenarios like this were not scarce. In fact, we began to see that this was a frequent occurrence in the data. In some cases, we were getting upwards of five maps per single company (e.g. XYZ, XYZ Corp, XYZ Corporation, XYZ US, XYZ US Corp.). 

The computer was not using the same human logic that we had been applying in the Boutique. Our human brains had no problem deducing that the above five names were all one company. Unfortunately, the computer brain was not able to understand that. It was evident that we needed to implement a name standardization process. Name standardization is important for two major reasons. First, it would give us one comprehensive journey map per company rather than multiple journey maps for each variation in name of the same company visiting our website. Second, condensing company activity appropriately would bring down the overall size of the report.

American Eagle and American Express are two VERY different companies 

Name standardization is trickier than it sounds. For example, it seems logical that Exxon Mobil and Exxon Corporation are the same company and can be grouped into “Exxon.” However, there are more complex examples that would be harder to standardize. Take American Eagle and American Express. Using the same logic as the prior example, these two companies could be grouped into one: "American." And while ripped jeans and credit cards are both a massive part of life as an American, that logic is clearly flawed. Because of scenarios like this, we knew we needed a sophisticated name standardization technique. 

Fortunately for us, Harte Hanks knows how to standardize a name—this is something we do for clients. We have a product called Global DataView™. It gives us the ability to cleanse and match data based on like components of names. By implementing our own product, we will be one step closer to producing our automated buyer’s journey maps.  

We knew changes to the first report would be inevitable. Going forward, we will continue to refine the process until we have the best report possible, putting us one step closer to contextual, human, in-the-moment conversations with our website visitors. User testing is something we will be getting quite familiar with in the upcoming months. Tune in next time to hear about how the changes are working!