6 min read

Building predictability into product-led growth 🔮

How InVideo built predictability with product-led sales
Building predictability into product-led growth 🔮
Photo by Isaac Smith / Unsplash

We're InVideo, an in-browser video creation platform used by millions of users (freelancers, media agencies, influencers) across a global footprint of 195+ countries. The following is an account of:

  • Why we set up a sales team despite healthy self-serve driven growth
  • Our early experiments in sales
  • And, how we eventually doubled our sales conversion rate!

2020: A dichotomy

As we look back at the last year, we see a growing dichotomy.

A dichotomy between growth and uncertainty

On one hand, we were lucky to see tremendous growth. Our product crossed over a million registered users, our team doubled in size, and we raised funds from some of the best global investors.

However, on the other hand, 2020 brought with it tremendous uncertainty. Not unlike most PLG companies, we'd historically relied on our product to be the sole driver of revenue growth (through self-serve premium plans). Despite our conversion rate holding up through the year, 2020 surfaced several questions that challenged some of our key assumptions.

Would freelancers continue to add credit cards on file through a rocky market? Would media agencies churn from team plans post layoffs? And would influencers continue to pay for software as they lost sponsors?

And there began our effort to increase the predictability of our growth.

Sales: A lever of predictability

Our first experiment in predictability began in November 2020. We noticed that our customer support personnel had been organically converting free users to paid plans through their interactions with users. Buoyed by this signal, we hypothesized that if support efforts could convert users, would a dedicated sales effort perform better?  

To begin with, we started small. A 3-4 member team was set up by reallocating in-house support resources to sales. The team was split three ways (to serve our key geographies) and tasked with contacting users who had signed up over the last 24 hours in a first come first serve (FCFS) manner.

Despite its scrappy nature, the experiment worked! Like clockwork, the team started converting users to paid plans. However, our FCFS strategy proved to be inefficient. The team consistently provided feedback that a majority of users they were reaching out to did not want to convert to paying users (students, casual users, etc.) and that their bandwidth was not being used optimally. We framed this as a two-fold problem statement:

  1. Which subset of users should we be reaching out to?
  2. What order do we reach out to them in? i.e., who is most / least likely to convert?

The problem of plenty

As we tackled this problem statement, we were faced with an issue all too common amongst PLG companies.

We had too many users (millions) but were limited by our sales bandwidth

Our product saw several hundreds of thousands of new users every month. However, the sales team had the bandwidth to only reach out to a small subset of those every day. We needed to prioritize!

(Hyper)prioritization based on user behavior

An InVideo user can carry out several unique actions within our product. Our hypothesis was to build a prioritization algorithm based on these user behaviors. Unsurprisingly, making sense of the noise from 100s of millions of such actions being taken was more difficult than we imagined.

Not only were the software engineering and data science resources required to set up a data warehouse and run complex analytics on it too expensive, but this pursuit would've also taken several months if not quarters. Time we didn't have!

That's when we decided to start with a small experiment. Intuition suggested that users who came from certain key countries, had viewed our pricing page, and had taken some high intent actions in the product were more likely to buy one of our premium plans. We quickly iterated on a simple machine learning model that was built on these limited signals - bringing down the complexity of the problem severalfold. The model then calculated each user's likelihood to buy (propensity score). Our sales team then went about contacting these target users in decreasing order of propensity score.

Within a week or two, the results were in. Our conversion rate had increased - but the increment was modest.

The experiment showed promise but needed more work. The next unlock was clearly to scale up the experiment to run on the entire set of actions a user could possibly perform - an exercise that would take months if not longer.

That's when we came across the Toplyne team. Toplyne promised to solve the above problem for us out of the box. They pitched us the following three-step solution:

  1. Building a 360° view of our users by stitching together user actions (product usage), demographics, and billing data all in one place
  2. Helping us segment our users - who are most likely to convert, buy more, churn, etc.
  3. Integrate into our workflows by syncing target users and related information into the platforms our sales team knows and loves.

Step 1: Building a 360° view of all users

Toplyne helped us stitch together information for every one of our millions of users from 3 key databases:

  1. Product analytics
  2. Billing
  3. CRM

To enable this we provided Toplyne with SSO logins / API keys to the above platforms and within a few hours, Toplyne created a unified 360° view of all our users right out of the box.

360° view of all our users (data and logos representational)

Step 2: Segmenting our user base

Now that we had an exhaustive view of each user, the next step was to figure out which user segments were ready for a sales touch.

Using the 360° view, we leverage Toplyne's built-in look-alike model to gauge which of our free tier users behave most like our paying customers. The platform provided us the output of this model as a simple to understand list of behavioral, demographic, and financial indicators, ordered from strongest to weakest.

Our team then created target user segments basis the above indicators that looked somewhat like the image below:

Our segmentation logic expressed in Toplyne (data representational)

Step 3: Integrating into our workflows

Once we had our target segments ready we needed to sync the users who qualified into our CRM and communications channel along with relevant information (demographic, behavioral and conversion likelihood) that would help our sales teams personalize their sales pitches.

We used Toplyne's scheduler to enable the syncing of target segments at the right time into Salesforce and Slack.

Our workflows (data and logos representational)

Results

Since we started using Toplyne in early June, we have seen a 1.8-2x increase in sales as compared to May ‘21. All without the addition of new sales capacity.

Sanket Shah, Founder, InVideo

Toplyne has also enabled us to be more personalized in our sales approach by providing salespeople with relevant information right in our CRM. As a consequence, we have created intricate playbooks for each customer persona, which has dramatically improved the buyer experience!


InVideo is the world's easiest video editor and a one-stop-shop for creating professional looking gold-standard videos in minutes, even if you’ve never made a video before. They publish weekly blogs and tutorials on their channels to help marketers and creators bring their ideas to life via videos. You can visit invideo.io and sign-up for a free account today!