Lending Club is an online financial community that brings together creditworthy borrowers and savvy investors to help both benefit financially. Founded in 2006, the company is the world’s largest online credit marketplace and has facilitated over $18 billion in loans to date, including personal and business loans, as well as education and medical financing.
Behind the scenes, the Product Analytics team at Lending Club is creating infrastructure to make the organization of over 1,000 employees as self-sufficient as possible. The team also manages A/B testing, web analytics, customer surveys, customer feedback, and SEO. Alan D’Souza, Director of Product Analytics, and Amanda Rosenberg, Senior Product Analyst, are two team members behind Lending Club’s analytics implementation and strategy.
Since implementing Heap, the team has already made a big impact. In one big win, they were able to discover small points of friction in the customer experience—a find that’s helped them to serve thousands of additional customers.
Searching for a Better Solution
Two years ago Alan joined Lending Club to lead web analytics. While the team had a solution from a well-established enterprise company in place for many years, it was used sparsely to view website visits and do light path analysis. It was never core to the team’s work due to concerns about data accuracy. Also, Lending Club products and pages evolved quickly, and new pages weren’t always immediately included.
According to Amanda, it was hard to get a full view of user behavior across numerous pages. “Either we’d try to find something directional to help inform our decisions, or it just was impossible to do, so we didn’t have answers to all of our questions, such as, ‘On this page, how many people clicked on View Agreements? And of the people who clicked on View Agreements, how many then went on to sign up or not?’ That formula—’for X, how many people did Y’—was hard to analyze with our old tool, and there was little transparency into how things were being calculated.”
When Alan joined, one of his first tasks was to figure out whether to stick with the established product or switch to something else.
“We realized that the old solution wasn’t flexible enough for the advanced analysis we wanted to do,” Alan said. “We wanted to answer questions like, ‘how many people took x, y, z actions in this order.’ We wanted to be able to track every click, to segment and create cohorts. For that, we needed a better, more flexible tool.”
The Analytics Tool Wishlist
Alan and team knew they wanted a tool that met the following criteria:
- Event-based vs. grounded in pageviews
- Quick to implement. In selecting an analytics tool, they wanted substance, not just style. One major concern was how quickly and easily they could implement it. Would the tool come ready out of the box?
- Provided a more complete dataset. Experience showed Alan that “it wasn’t possible to anticipate every possible question in advance. Questions would come up from across the org, such as ‘How many people click on this agreement popup in the footer of this particular URL?’ Those are peripheral things that I’d never think to tag.” In their next tool, they wanted event tracking to be easier by not having to decide in advance what to tag.
- Made raw data easy to access. No matter how great the UI in their new tool might be, Lending Club wanted access to their raw data, so they could extract it to power A/B testing, combine multiple data sources, and run predictive modeling.
When Alan started researching tools, he “simultaneously installed the Mixpanel, Amplitude, and Heap scripts. Ten minutes later, we’re getting all of this data in Heap, and nothing in the others. Heap just worked. I didn’t want it to work, because it was too simple—was it really that easy?”
Creating a Better Customer Experience
After deciding to use Heap, the team began to dig into their project backlog. Amanda wanted to analyze friction points that users received when requesting loans, particularly validation errors. With a validation error, a user doesn’t fill out a field in the way the system is expecting and asks the user to redo something rather than progressing them to the next step. With Heap, she could identify these, understand how many people were impacted by them, and therefore prioritize which ones engineering should solve first.
“I was like a private investigator, trying to find out why someone would get stuck,” Amanda said. “When we joined validation error data with Heap data, we were able to say how many validation errors we were receiving, narrow it down to specific ones, and ask whether or not a user reached the next conversion point. For example, if someone forgot to fill out a checkbox. If a user truly didn’t check a box, they’d likely just go back, click it, and move on. That may tell us we need to make the box more noticeable. But if they don’t proceed, it’s a signal that there may be some other friction in the process that we need to remove.”
To identify what were actually validation errors vs. one-time user errors, Amanda was able to take a user’s email address and user ID, go into Heap, and see how many times they received an error, tried to fix it, and where they got stuck. After that, she could see how many users experienced the error and didn’t continue to the next step.
“If 300 people a day experience the same friction, we know that it’s a good candidate to address. We calculate the impact of fixing that by multiplying how many people were impacted by average conversion rate,” Amanda said.
Added Alan, “The improvements we’ve been able to make are a big deal for us. They’re enabling thousands of people a week to have a slightly better experience using our site. A good number of people won’t be put off by these small frictions and will complete the task anyway, but if even 10 or 20 people a day get through who would otherwise have given up, that’s 10 or 20 more families that are one step closer to freedom from credit card debt or growing their business. That means a lot, for them and for us.”
Insights from Combining Data Sources in Redshift
In addition, the team has been using Heap for A/B test tracking and detailed persona creation. In both cases, Alan and Amanda pull multiple data sources into Redshift alongside their raw Heap data to glean deeper insights. With A/B testing, they can enrich Heap data with testing control information, export it to Redshift, and model it however they’d like.
“You can go beyond the data you typically see with A/B testing to look at every down-funnel and non-test page action. We’re still just scratching the surface, but will soon be able to form a complete picture around A/B testing analytics on top of our raw Heap data,” Alan said.
By combining Heap data in Redshift with user data that lives in their database (e.g. credit score, location, age, etc.), they’re creating holistic user personas to see how demographics impact site patterns. Because a big portion of traffic to their site comes from mobile, the team is using Heap to inform how to optimize the mobile loan application experience.
“In researching site patterns, we wanted to identify whether people preferred starting and completing applications on desktop, mobile, or a mix of both. That’s very important, because it informs whether you should optimize an experience for people switching devices, or make each flow the best it can be without worrying about cross-device consistency,” Alan said.
Making Everyone Data-Informed
Beyond their big wins with validation error analysis, A/B testing, and persona development, Alan and Amanda most enjoy the unexpected workflow changes that come from using Heap. Namely, fewer meetings and faster decision-making. The team is building out lightweight dashboards for the rest of the company, but also wants to help everyone go into Heap and explore raw data themselves.
“People are conditioned by incumbents to want these big metrics like page summary,” Alan said. “But those aren’t actionable. What improves things for our customers are specific questions, which is why we want to spend as little time as possible on these high-level, feel-good numbers and more time on the specific things that create value day-to-day. Heap is the only tool I’ve encountered that lets everyone instantly answer business questions.”