Scaling an Enterprise company with Data

Interview with Govind Chandrasekhar, Co-founder —Semantics3

Govind Chandrasekhar, one of the founders for Semantics3, talks to us today about how his company uses data to drive product & business strategy.

The company has been operating in the enterprise/SaaS space for the last 5+ years and has offices in Bangalore, San Francisco and Singapore. It has built large-scale data pipelines to crawl the web and capture pricing information for various products across different retailers. In addition, they’ve built a sophisticated set of machine learning algorithms that operates on this data to provide various e-commerce data solutions for retail organizations.


Q: Govind, you and your co-founders have been at this for the last 5.5 years. Can you tell me about how the role of data has evolved at your company as far as determining product strategy is concerned?

Interesting question. We track and optimize for a range of metrics, and the importance we attribute to each of these has fluctuated over the years. That said, we’ve always used one metric as a north star to measure our performance — Revenue.

Just like most startups, we are always looking to make the most of limited resources; the best way to do this has been to build a product roadmap that maximizes our potential to generate revenue over a 6 to 12 month window.

Q: One follow up question. Have you always followed the philosophy of estimating revenue over a 6 to 12 month period or is this something that has evolved with time?

So this is a philosophy that we have always followed and it really speaks to our company’s roots. We did not raise any venture funding till a year and a quarter into our existence and this meant that we had to focus on revenue to keep our doors open. 
 
Believe it or not, during the early days, we used to look at elance.com (now Upwork)& freelancer.com for projects and tried to productionize everything we built. We did not realize it at that time but taking this strategy allowed us to build exactly what the customer wanted.

This philosophy has stuck with us till this date. We have an enterprising team and fancy ideas crop up all the time; it can be tempting to jump in headlong into a long product development cycle without involving the customer (and we have been guilty of this at times). For the most part, however, whenever we want to build something new, we try to validate this demand by sounding out existing customers with low-fidelity designs and proofs of concepts. Only when we get to a point where we know for sure that customers will pay for it do we go all out.

Q: Fascinating. Just to clarify, when you say revenue, do you mean Monthly Recurring Revenue (MRR) or Annual Recurring Revenue (ARR)?

MRR.

However, one of the things that we have struggled with is figuring out how to include one-time revenue from customers as part of the MRR.

With one-time revenue, there is a mutual understanding between us and the customer that they would only be using the service for a short period of time and will be churning out after that.

One-time revenue is tempting because it’s often right there for the taking, and can help us hit our monthly growth targets. The problem though is that it digs a guaranteed hole in revenue when the one-time revenue disappears from the books, making it that much harder for our sales team to achieve that month’s targets. As a team, tempting as it may be, we’ve collectively decided now to not include this revenue as part of our MRR.

Q: Alright, so how do you handle revenue from long term contracts?

We take the total contract value and divide it by the length of the contract and account for that monthly revenue as part of MRR — we essentially follow the accrual basis of accounting.

Q. Okay, so what are the top metrics that you look at beyond revenue?

Every single department in our company, including the engineering team, have their own set of metrics to optimize on.

On a weekly basis, the entire team has a call in which we do a deep-dive into the metrics for the week gone by. These metrics are accessible to everyone in the company, so if any of these numbers have been too stagnant for too long, it’s everyone’s responsibility to draw attention to this. Of course, we also have growth targets for a lot of these metrics that make it obvious when numbers stay stagnant

Q. Interesting, could you elaborate on the teams as well as the metrics they try to optimize?

Of course. So we have 5 top level departments — sales & marketing, data science, infrastructure, customer operations & the applications team.

The sales & marketing team owns the top level metrics — revenue. However, they also look at more granular numbers such as number of leads that entered the pipeline, number of signups on the website, number of outbound calls booked etc.

The data science team have metrics around the precision, recall & F1 score of the various algorithms important to the company. While these metrics are a little removed from revenue, this is actually very important from a product point of view — we want to be known for having the best-in-class algorithms.

The infrastructure team manages our entire web crawling operation and their goal is to ensure that the product and prices in our system are accurate. They look at metrics such as number of URLs crawled per day etc. They also handle secondary but critical metrics like overall health of our internal engineering infrastructure.

The customer operations team is the team in charge of retention. Since our contracts are enterprise in nature, we need to make sure that our customers get the most out of their relationship with us. This team tracks its performance through customer retention and the incremental revenue generated per existing customer — it’s their job to upsell the customer.

The last team we have in-house is the Application team, the most important of the lot in a lot of ways. They handle a broad range of tasks. One of them is to track the uptime of our public-facing APIs. Another is to independently “audit” the quality of the output generated by the other teams from a customer’s perspective. For example, they replay a lot of customer requests against our APIs and check to see if what the customer is seeing is in-line with what the other technical teams are reporting.

Q: So the one thing you have not mentioned yet (and I am curious to know how it is done in an enterprise company) is how you calculate Customer Acquisition Costs?

So this has actually been a very tricky for us. As a lot of our contracts are enterprise in nature, we actually need to have multiple touch points with the lead before we bring them on as a customer. Our first conversation is often a general discussion of the customers’ short and long-term needs, and a way for us to communicate what exactly is possible through our company’s offerings.

I realize that this is one of the most important metrics for game companies or eCommerce businesses. However, for us, the time and effort taken to close a customer as well as the total contract value is highly variable, and is not easy to attribute to one single acquisition effort.

Hence, rather than looking at the customer acquisition costs at a granular level, we track more meta metrics like call bookings and PoC conversion. We balance cost concerns by setting a budget for our sales and marketing team to work with and drive growth.

At the end of the day, revenue growth trumps all else for us. As long as that happens, everything else inevitably falls into place.


Lead with Data was built with the purpose of helping people make better data driven decisions. Be sure to follow this publication and checkout www.leadwithdatahq.com.