@ LeadCrunch.ai
LeadCrunch.ai is a marketing technology startup that uses AI to facilitate more effective marketing strategies. When I joined LeadCrunch, it had recently pivoted to focus on selling AI guided qualified lead campaigns. The AI creates the targets, and then humans engage the leads. As demand surged for this new product I was tasked with designing solutions to help our team keep up.
Shortly after LeadCrunch launched its Qualified Lookalike leads product, demand surged and outstripped our capacity to fulfill orders. The manual process used to fulfill these orders could not keep up. Our internal fulfillment team developed an ad hoc process to this demand in the short term - but quality issues began to creep in. Clients were quick to notice these issues, damaging their trust in our product and forcing us to run expensive "make good" campaigns. To keep up, our team was working long hours - our key players were burning out.
The data being generated was scattered and unusable, making it impossible to analyze and optimize our process. This also prevented the product from displaying that data to clients. Finally, LeadCrunch used third-party partners to help engage leads, and the lack of data clarity made it difficult to evaluate their performance.
Finally, the messy onboarding process led to delays fulfilling orders and questions from clients. Every extra day it took for us to start work on an order was a day our clients spent wondering if they’d wasted their budget.
In order to help keep up with demand, collect the needed data, and give clients a better service experience it was critical that we find ways to scale the order fulfillment process.
Improving our fulfillment process was a multi-phase journey that spanned many design and development cycles. To explore how to solve the problem, I first mapped out the process used when fulfilling campaigns. I worked with our customer success team to understand the order intake and fulfillment process and identify bottlenecks and breakdowns that they encountered.
We identified three key workflows as major opportunities for improving the scalability of our order fulfillment process.
First, the AI target list creation process used at the start of every campaign was error prone and repetitive.
Second, the manual process of managing our third party partners caused too much overhead for our customer success team. Expensive miscommunication errors often occurred in this process.
Finally, the quality assurance process was unstructured and ripe for optimization. Many steps could be automated and the data generated during it could be centralized for use throughout the business.
We approached each of these topics as separate design initiatives, although they were ultimately connected as one seamless order fulfillment experience.
The first bottleneck in our fulfillment process was the mechanism used to generate AI targeted leads. At the start of every campaign, we used data on the client's existing customers and generated a set of prioritized targets for outreach. Creating these targets involved the customer success team working with the data science team to tune a model for each campaign. This back and forth took up a lot of their time and delayed the launch of campaigns.
This process had to be repeated for every new order placed by a client. Errors when creating these models could cause expensive outreach to invalid leads. If a model was not made quite right, we could wind up engaging leads that the client would not accept. As a result, we faced constant delays and opportunities for mistakes to be made when launching a new campaign.
To improve this process, I worked with the data science team and customer success teams to map out the steps needed to create AI models. I then iteratively designed and tested a new process to create re-usable models once per client. This minimized the time needed to set up each campaign, and removed opportunities for errors on subsequent campaigns.
The result of this feature was a substantial decrease in the time to value for clients. The changes also allowed the data science team to focus on developing our technology, rather than helping with customer support. Finally, the new feature reduced opportunities for errors by limiting the number of times models needed to be created. This reduced the number of rejected leads that needed to be replaced, saving thousands of dollars per campaign and improving our margins substantially.
After generating AI targeted leads, LeadCrunch used a variety of in-house and third party partners to promote content to those targets. However, setting up and managing these third-party services was a huge amount of work for our customer success team. For each order, multiple days were spent communicating with these partners to make sure they had the accurate material and information about the campaign.
To help alleviate these problems I conducted a design sprint with the customer success team to map and optimize the partner management process. I also conducted user interviews to better understand our partner’s workflows and how we could integrate more seamlessly into their process. Finally, I rapidly prototyped the best ideas from the design sprint and tested a partner management product designed to streamline the process of working with our partners.
As a result, we created a partner management platform that automatically provided them with the information and materials they needed to do their work. This product also provided a place for partners to upload their lead data directly into our system, removing the need for the QA team to receive and process it via email. This allowed us to apply various QA checks at the moment that partners uploaded data. As a result, a whole set of easy to catch formatting issues were taken off the plate of the internal QA team.
The partner management experience became the foundation for how the customer success team set up and managed partner services. It reduced the risks of miscommunication with partners, and removed much of the partner management overhead involved in fulfilling client orders. As a result, campaigns could be fulfilled more rapidly, with fewer mistakes along the way - ultimately resulting in more satisfied clients.
The final major bottleneck in scaling our fulfillment process was the core of the QA workflow. This complex process involves dozens of review and approval steps for each lead that is generated. The customer success team had developed a piecemeal method for completing this process using spreadsheets and google docs that worked for a while. However, as we scaled to sell millions of dollars of leads each month, this process couldn’t keep up with demand.
Besides being unscalable, this process had a number of other drawbacks. The data it generated was not captured in a usable format, which prevented the product from displaying critical information to clients. Expensive mistakes due to simple typos and errors in spreadsheets. There were many validation tasks that could be automated, but with such an unstructured process there was no way for the product to interface with the data.
To make this workflow more trackable and scalable, I worked with the customer success team to model their process and rebuild it as a productized QA tool. The team was stretched thin so it was imperative not to add any additional steps in the QA in the transition period. Through an iterative design process working with product managers, customer success stakeholders and other designers, I created a workflow that brought all of their QA and fulfillment tasks into the product.
With this new interface, the client success team could do all of their QA work from a centralized location. They could upload client requirements and the product would automatically validate incoming leads against them. The data from the QA process was also automatically tracked and recorded, and fully QA'd leads could be automatically delivered to clients.
These internal fulfillment products became the bedrock of LeadCrunch’s business model. They enabled the creation of tens of millions of dollars in revenue. With these products, LeadCrunch could keep up with growing demand and scale from struggling to fulfill a few orders each month to handling hundreds of campaigns simultaneously.
Pressure was taken off the customer success team and clients were happier with the quality of leads they received. Before these optimizations, each customer success team member could only manage 2-4 campaigns at a time. After these changes, they were able to manage 30-40 campaigns each. These internal tools laid the groundwork for the larger product strategy: by collecting data in a structured format LeadCrunch was able to optimize its internal business process and use the data to create a client facing product experience.