My cofounder and I have been building products for the better part of 15 years now. From starting a consulting agency at the age of 18, to helping early stage startups build and launch their first products, to jumping aboard fast-growing rocket ships and taking them to an exit, we've seen our fair share of successful (and unsuccessful) products.
Great products happen when the feedback loop between the people building the product and the people using the product is as tight as possible. The best companies that we've been fortunate enough to be a part of understood this—without listening to your customers, your product is doomed to fail.
We started Viable to help teams get closer to their users, to truly understand their needs, and to help them make better decisions. We started by helping early stage companies to find product market fit using the engine that Rahul Vohra and the team at Superhuman built, but that was just the beginning.
One of the tech industry's greatest innovations has nothing to do with software or hardware, but with process: the tech industry has embraced customer feedback in a way that no other industry has. Our industry thrives on fast, tight feedback loops: we measure product usage at a granular level. We track NPS scores. We keep a constant eye on conversion, retention, engagement, and active user metrics. And we put processes in place to make sure that we use this data to make better decisions, faster: A/B tests and feature flags help us choose the right messaging and release the right features, sprint planning helps us remain responsive to the feedback we're getting, and data and analytics teams help us answer tough questions about how our users use our products.
This focus on tight customer feedback loops has helped catapult the tech industry to the top of the market—a fact that's made even more impressive by the realization that we've been missing half the picture.
We have an obsession with quantitative metrics in this industry. And rightly so—quantitative metrics are the best way to understand how people use your product. What quantitative metrics can't tell you is why people love your product, why they're frustrated by that new redesign, or why they're confused after onboarding. They can't tell you how you can improve your product; they can only show you what's being used and what isn't.
In order to dig into the why behind your users' actions, you have to go straight to the source. You have to ask them. User researchers have known this for decades. But there's a problem with this approach: you can't do it at the same scale that you can with quantitative metrics. This approach takes work.
You have to decide what questions to include and which users to ask. You have to wait weeks for the results to trickle in. And once you've got all those responses, you have to sift through all of that open ended text yourself—manually tagging topics, segmenting respondents, and tracking sentiment. Even after all of that work, you still have to analyze and summarize the newly structured data for the other teams in your company to use. This is, by definition, not a tight feedback loop.
But what if it was? What if your company already had all of the data it needed to shorten that feedback loop from weeks to seconds? It turns out most companies do: customer feedback is siloed in support tickets, NPS responses, live chat logs, social media mentions, app reviews, and hundreds of other sources. Teams in almost every company generate customer feedback as a side effect of getting their work done. Companies just aren't using it yet. This is the problem that we set out to solve with Viable.
In order to enable companies to get the most out of the feedback that they already have, we needed to enable them to aggregate and structure that qualitative data, to mine it for insights with the ease of asking a question, and to surface the big picture of the data by quantifying the qualitative.
First, we needed to gather and structure all that data from a disparate set of sources, so we set out to build a no-code integration system that was as easy to use as possible. Our integrations don't require any engineering time to implement and can be done in just a few clicks. Once we had a way to import data from a ton of different services, we needed a way to structure the data and for that, we turned to the most intelligent language model in existence: GPT-3.
It's only recently become possible for computers to understand human language well enough and fast enough to analyze feedback at the scale modern companies require. Up until a few years ago, it would have been impossible. So what changed? NLP (Natural Language Processing) researchers made a breakthrough: they found a way to use a type of deep learning neural network called a transformer model to accurately model human language. This was discovered in 2017, and just three years later, OpenAI released GPT-3. This transformer model could keep track of 175 billion parameters: 10x larger than any transformer model before it. It turns out, this completely changed the game. Not only was it capable of understanding human language, it was capable of writing it.
Viable was an early beta partner with OpenAI and we worked closely with them to develop a system that could understand customer feedback well enough to extract topics and assign each of them a sentiment and an emotion. Using this data, we're able to tell which topics customers think are positive or negative, and which topics are confusing, frustrating, or delightful.
Now that we had a system that could aggregate this data and structure it in a way that makes it easy to analyze, we needed a way to let anyone interact with this dataset. If our users had to be researchers or analysts in order to mine their data for insights, our customers' feedback loops wouldn't be as tight as they could be—teams would have to wait for the analyst to get around to their question. So we built a system using GPT-3 and other in-house models to allow our users to ask questions of their data in plain English. Our system would take in a question like "How can we improve our calendar feature?" and translate that into a complex search query to pull back all of the feedback data about the calendar. Then we used GPT-3 and in-house models to generate a summarized, actionable answer for the question that reads like a friendly and knowledgeable colleague wrote it. You no longer need to read through thousands of pieces of feedback just to get your answer!
All of that alone would be useful, but it wouldn't be as actionable without some quantitative data to back it up. So we built a system to quantify the topics in the answer, allowing our users to see just how big of an issue each topic really is. Then we took that same tool and zoomed it out for a company's entire dataset so that our customers could track the trends in their feedback by topic, sentiment, and emotion.
With this new product we can now take in what may look like 200,000 rows of customer feedback data coming in per month for your company, and show you the relevant pieces of feedback for any question you ask. We can also plot the bird's eye view of the dataset onto a set of easy-to-read graphs. Using this system, your company doesn't have to wait weeks to learn from your customer feedback. You can get your answers in seconds.
We want to see a world in which CX, operations, and product teams don't have to waste time sifting through rows and rows of user feedback across spreadsheets. We want all customers to have a voice, and we want the smallest of teams to be able to hear it. We can't wait for all of the things you'll learn along the way.
Now qualitative data can finally be a part of our feedback loops. Now we can finally understand the why behind our users' actions.