By Andrew Hutson on Sep 4, 2020 9:42:25 AM
This article was originally published on Marketo Blog on August 15th, 2018.
As marketers, we talk about analytics and lead generation all the time. But what does the term “lead generation” actually mean? And how will prescriptive analytics guide lead generation in the coming years?
Lead generation is “the process of stimulating and capturing interest in a product or service for the purpose of developing a sales pipeline,” according to Marketo. It’s the first step in the sales supply chain. Without leads, we have no one to sell to.
Lead generation is often divided into two categories: inbound and outbound. Inbound attracts leads through a combination of tactics that typically center around social media and website advertising. Outbound refers to the act of finding prospects that could be interested in a particular product or service.
In this blog, we’ll focus on analytics related to outbound lead generation strategies.
Climbing the Analytics Maturity Ladder
To create successful lead generation strategies, companies need to analyze the data gathered from previous campaigns. Organizations that want to get more insights from their data will likely consider descriptive analytics as a starting point. Companies entering this phase typically operate on intuition, siloed data sources, and manual data entry. Maturing in this stage requires retiring disparate data sources, increasing data governance practices, and automating reporting. These practices increase data reliability and offer a better understanding of past performance.
Companies reach the next level, predictive analytics when they can leverage descriptive analytics to anticipate future performance. Companies have been trying to predict how they will perform based on external factors and past performance for ages, but it has become easier to quickly compare past performance data with external variables.
However, making a good prediction doesn’t mean making a good decision. Predicting typically means finding a quantity of something as a function of another variable based on collected data. Once a prediction is made, a company typically makes a decision based on evidence from descriptive and predictive data sources. A good decision requires consideration of facts and data against irrational or unknown aspects of a system.
So, what should a business do? Prescriptive analytics techniques aim to answer this question. Reaching prescriptive analytics maturity requires deliberate, disciplined collection of performance data, reliable predictions, and continuously improved prescriptive algorithms. It also requires a company culture dedicated to data-driven practices.
Prescriptive analytics could improve business processes through predictable interventions that are tracked and fed back into the prescriptive engine; it could shorten employee onboarding using programmatic design trees to guide and inform new workers, and it could allow customers to leverage a company’s expertise through its optimized algorithm. Prescriptive analytics could also help determine the best path forward for new lead generation campaigns.
Preparing to Use Prescriptive Analytics
Prescriptive analytics relies on the quality and accuracy of descriptive and predictive data sets. If your company is new or has been suffering from a low number of “wins” from your leads, you might be facing imbalanced data or data scarcity. To address data scarcity and imbalanced data, you have a few options:
1. Get More Data
This is often easier said than done. Few companies have access to large amounts of data generated from activities outside their own organizations. Therefore, getting more data means waiting for data to be collected or purchasing data from a third party.
When my company first launched, our team tracked every transaction produced by third-party apps and stored them in a structured database. Even if we weren’t certain how we would use the data, we still knew it would be valuable. Storing data is cheap, and we knew we might want to reference something in the future.
2. Use a Different Metric
For lead generation, this means using identifiers other than “sale versus no sale” to determine classifications for the data. For example, you could use “sale size,” “return business,” or “number of interactions” as a method of classification.
We explore many different metrics and attempt to identify the ways they relate to one another. We continually inspect, challenge, and interrogate our data to understand which attributes and metrics help us the most.
3. Resample Your Data
Resampling your data can offer you a different distribution, but that assumes you’re pulling from a large data set. If you don’t have enough data, this often isn’t an option.
We’ve run across this problem before. Sometimes, the sample population we pulled was imbalanced by chance, not because the overall population was actually imbalanced. But if we pull a different sample population, we sometimes get a better distribution.
4. Use a Service
An increasingly effective option is to capitalize on a third-party service that has more prescriptive data analytics, various metrics, and tested scenarios. In the case of lead generation, this would mean multiple clients, prospects, industries, and products. Our team uses machine learning techniques like PCA and classification and regression decision trees to understand which messaging produces the desired outcomes for our clients’ demographics and offerings.
Once you’ve used one or more of these strategies to solve any data issues, you can begin experimenting with different prescriptive algorithms, from linear regressions to decision trees or neural networks. Whichever route you decide to take will likely lead you to new ways of interpreting your data to create successful outbound lead generation strategies.