The Intelligence Era and the Virtuous Loop

We are now entering a fourth era of computing — the era of intelligence — where we can combine huge volumes of data with learning algorithms to make software automatically get better over time. This is yielding products that make our lives better every day, from highly accurate speech recognition to autonomous vehicles.

To put this into context for investors, the first era started in the 1970s and was about investment in new hardware and semiconductors, characterized by Moore’s law. This was the true beginning of technology venture capital. The second era — the software era — started in the late 1980s and was characterized by the development of applications to process our real-world workflows on these semiconductors. This was essentially ‘data in, data out’ software. The third era — initially Software as a Service and then named Cloud Computing — was characterized by the distribution of computing to significantly reduce costs and allow the development of new types of software. We could run software and store data but few applications from this era used the data, except to generate reports. However, more data from a myriad of users created an opportunity for a whole new class of cloud-based software that learns over this combined data to make smarter applications. And while the pursuit of intelligence and prediction can be traced back to the 1950s at IBM, these ‘intelligent’ cloud-based applications are finally replacing the applications from the previous era, giving good reason to qualify this as a new era of computing.

The Virtuous Loop

Cloud computing offers many advantages over on-premise applications such as lower total cost of ownership, easier manageability, faster time to implementation, more frequent updates, and greater security (some would argue). Thus, customers migrated from on-premise to cloud vendors over the past two decades.

The first wave of cloud-computing vendors pioneered a multi-tenant architecture where each customer could access the same software stack but their data was stored separately. Thus, insights were separately created for each customer using only their own data. The first wave of cloud computing vendors produced more efficient and responsive — but not more insightful — workflows. However, as Google and other consumer-based cloud companies proved, there was value in analyzing automatically created, implicit data (the ‘data exhaust’). The new generation of enterprise cloud-based vendors realize they can also benefit from analyzing this cumulative, cross-company data exhaust. These insights are of a different quality than anything one can glean from a single customer’s dataset, no matter how large. Architecture that combines customer data, a data exhaust, public data, data science and machine learning algorithms can yield an even higher quality of insights.

Some have referred to these new applications as data-driven applications or data-first applications but we think of them as applications architected with a ‘virtuous loop’. A virtuous loop architecture feeds the anonymous customer data, data exhaust and public data into machine learning algorithms to generate both cleaner data and insights. The clean data is enhanced by more incoming data, fed into more appropriate algorithms to produce even better insights and thus generates even more useful data. And this goes on. This virtuous loop helps the applications learn from the data and get smarter.

The early cloud companies were so focused on resolving the multi-tenant security risk they gave away the most crucial asset a cloud company could have built: cross-company data that could be used to derive more meaningful intelligence. Google and other consumer cloud companies maintained rights to cross company data exhaust but their enterprise counterparts did not.

The Virtuous Loop

InsideSales: Exemplar Enterprise software

InsideSales is a Sales Acceleration platform that blends customer interactions with a rich set of profiles. The company applies Machine Learning to growing database of 90B sales interactions, adding 5B/month. InsideSales constantly mines the data in real-time to serve sales reps leads mostly likely to close, what time of day to make contact, and what by means of contact (phone/email) leads should be contacted. Customers often see an increase in sales conversions of 30% in their first month of using the product. The integration of the data, algorithms, and the core software is what provides the value and allows them to have a barrier to entry against potential competitors. InsideSales has trademarked its virtuous loop engine as Neuralytics.

Simplified Neuralytics Architecture

Startups Well-positioned to Win

Startups are well positioned to disrupt existing multi-tenant application vendors with next generation, virtuous loop-based applications. This is especially true because they can kick-start their virtuous loop in a few, cheap ways.

Firstly, companies can kick-start their database by gathering lots of data from many, small customers and then learn over that data. As the software gets smarter and produces better results, large companies are attracted to become customers as well. Those early customers can often be acquired by startups while still on their seed funding; a well-priced minimum viable product (MVP) can be easily adopted through word-of-mouth among small customers. These MVPs are usually simple, for example a cloud-based sales dialer (InsideSales), conference-room scheduler (Eventboard) or facilities management dashboard (Lucid). However, the value is in the generation of data, vertical skills and clear product roadmap to add analytics features.

Companies like InsideSales generated 400 small customers before raising any funding. Today they have 3,000 customers, many large enterprises with thousands of users. Eventboard started with 300 small customers on just $200K of funding. Today they have 1,300 with many enterprises doubling in use within 90 days. Lucid started with a small number of college campuses on its seed funding and today have 500 enterprise campuses and 30 cities and manage 1.3B square feet of office space. In each case, they are well positioned to mine billions of pieces of data to dramatically improve the results for all their customers. Every new customer, whether large or small, benefits from the models created by the base of customers already there. Thankfully, startups today benefit from a more amenable climate to ask customers for the right to retain cross-company datasets for broader learning.

Secondly, there is a proliferation of publicly available data to combine with the proprietary data. For example, there were only a handful of public government datasets available at the start of the Obama administration and there are over 180,000 today. 40 countries have joined to create hundreds of thousands more.

Thirdly, there are more companies collecting and cleaning data than providing it in a structured way to developers through APIs. For example, Clearbit, a Zetta portfolio company, is appreciated by its customers for providing the highest quality and the most recent data on people and companies.

Partnering with Zetta

The market opportunity for virtuous loop-based startups is in the trillions of dollars given the advantage for every line of business in the enterprise, every SMB, the industrial sector through the Internet of Things, and previously data-starved industries such as agriculture and cities.

Our focus at Zetta is to help find and build startups like InsideSales, Eventboard and Lucid that generate lots of proprietary data, creating platforms with barriers to entry by combining data with workflow software. To be seed funded today, it is important to create a purposeful and long-sighted data strategy from the beginning. And, as the era implies, our focus helps us to get smarter with every new investment.


Thanks to Jared Haleck for his comments on this post.