After nearly three years of development, today, my team and I are extremely excited to finally launch our company Infer. Our goal is to help companies significantly win more customers by providing applications inspired by the deep data science and simplicity of the consumer web.
Why we’re doing this
My co-founders and I had the great opportunity to work on large scale data products at consumer web companies like Google, Microsoft, and Yahoo! There we witnessed first hand the impact cutting-edge data science and systems infrastructure have on making properties like Google Web Search so great and relevant. The rigor that takes place behind the scenes here is truly unbelievable.
However, when you compare the scientific approach that underpins popular consumer facing properties to how companies internally leverage their own data for important business decision making, it’s astoundingly poor in comparison.
For example, there is way more intelligence being applied in Facebook’s newsfeed telling you that your friends are getting drunk across the street than there is in helping companies make critical decisions that could make or break a quarter and lead to layoffs. We found that incredibly troubling and want to do everything we can to change that.
We looked at the data intelligence solutions that are out there today for companies and they simply don’t cut it. Many Business Intelligence (BI) and “Big Data” plays are in the business of delivering infrastructure and tools tailored for statisticians and engineers. These platforms are overly general, require heavy lifting, and don’t attempt to go deep on tackling specific problems end-to-end. And that’s precisely what we need. Where are the killer applications that leverage data science but don’t require you to be a data scientist to understand and run with?
For example, a typical VP of Marketing isn’t leaning back in his or her chair and thinking, hmm, I just need to copy all the historical data sitting in my CRM and marketing automation systems into Hadoop, crawl the web for more customer signals, and develop a machine learning model over all this data and plug those predictions back into my existing workflows to produce lift. It’s up to us to show them what’s possible and to build applications that can automatically do this for them while abstracting away all the complexity.
Applications require deep focus, so we decided to initially pick the biggest problem out there that affects the top line, and that’s helping companies close more customers – specifically, developing predictive lead scoring models over the historical customer records sitting in sales and marketing automation systems. Today, automation systems simply store and track sales and marketing activities and outcomes. However, automation alone is not enough to move the needle for companies.
They host great ground truth data for us to model, but automation lacks the intelligence to help companies answer critical questions like which potential customers are likely to win, which customers to sell more to, and which customers are at risk of churning. We need a new wave of applications designed with intelligence and consumer level simplicity (no training required and fast to deploy) at their core to help companies once and for all tackle these kinds of problems.
How do we do it? What’s so different?
We spent more than two years developing the science and experimenting with a variety of companies to nail a breakthrough solution for winning more customers based on data – and man do we got a solution we’re eager to roll out to more companies. We came up with many key tricks to develop this breakthrough.
For one, we don’t just machine learn models using the signals from the data sitting in automation systems – we also leverage hundreds of external signals about customers from a variety of web sources to ensure we’re modeling off the most complete customer profiles imaginable. For example, one signal we use is the number of recent job openings by department. Say Zendesk (customer shoutout) wants to know whether they should prioritize a particular lead (company “A”). One of the signals our system will look at is whether company A has any customer support job openings, because if they do, it shows potential intent and budget to invest in a customer support solution (such as Zendesk’s). This is just one example of an external data field that could help greatly in identifying winners but is typically not picked up or collected in automation systems today.
Second, we’re applying deep data science techniques (probably for the first time in this space) across the stack, from matching company records to the right external signals (teaching our machines that “IBM” and “International Business Machines” are the same company), to cleaning up data quality and handling sparsity (missing data) issues, to fine-tuning the predictive lead scoring models to ensure the most accurate scoring possible for our customers.
Third, we’re all about building great products with consumer level simplicity. Many data intelligence efforts can take months or years to complete. Our solution from start to finish deploys in mere days. Just give us a seat into the CRM system (such as Salesforce) and we’ll take care of the rest. We push our scores and intelligence directly into companies’ existing automation systems (we have connectors for Salesforce, Marketo, Eloqua, etc.) so reps and managers do not need to learn a new experience or sign-in to another portal with another set of credentials. Consumer level simplicity means going to where the customer is, and making it dead easy so that anyone can get this up and running very quickly and get the most out of our solution, with no training required.
We have deployed our solution in a variety of customers, ranging from the Fortune 1000 to some of the hottest companies out there like Box, Jive, Tableau, Yammer, and Zendesk. Our models have produced consistent and significant lift across the board. We have multi-year agreements in-place with many of our customers, are profitable, and most importantly have incredibly happy customers.
Focus is everything, and nailing how to accurately score any customer’s potential has enabled us to uncover a brand new business primitive. It’s not just about scoring leads. You can leverage Infer’s scoring applications to run better campaigns, do total addressable market analysis and planning, identify enterprise consolidation plays, visualize the health of the business at any point of time, and much more.
Many of our customers have referred to us as their secret weapon for achieving hypergrowth (sorry we’re launching!). We’re pushing the envelope when it comes to modeling and visualizing a company’s business. We’re completely re-imagining how companies should operate based on data.
For example, instead of a VP marketing saying they generated 5,000 leads this month (which doesn’t mean much – I could generate a 100K leads by dumping in irrelevant leads lists), they could instead say they generated 1,000 leads that Infer indicates will likely convert. Way more substantive and meaningful. It’s no surprise that our metrics are being leveraged in several of our customers’ board meetings.
What’s the Future?
Imagine company A, who competes with company B, is leveraging our solution and achieving 100%+ lift in their win or conversion rate. It would be irrational for company B not to follow suit as well to not just compete but survive. This is going to be an arms race powered by data. I strongly believe this will result in the biggest shift in business we’ll see over the next 5-10 years.
We’re very fortunate to be well-capitalized by world-class, long-haul investors and are ready to go out with a bang to get our solution into more customers and pioneer this paradigm shift. Despite the great traction, we’re just scratching the surface. We want to make our models even better, invest in many more data sources, and attack more problems with even better models. We want to build the very best engineering and science team ever assembled in this space to help us spark this data revolution. Many businesses simply have no idea what they’re missing. We’re going to make it our mission to show companies what they can really do with their data.