.jpeg)
If you’re even kind of acquainted with tech Twitter, you’d know about Roon’s (at this point almost seminal) blog post for Scale AI from September 2022 called ‘Text Is the Universal Interface’.
Given what we’ve seen since then, he’s starting to look increasingly prescient.
Since the introduction of the Macintosh and its GUI in 1984, ‘clicks’ have been essentially synonymous with interacting with computers.
They quickly became a staple in user interfaces, and in retrospect, were an inevitable shift from the complexities of command-line interfaces.
Arguably Jobs’ biggest contribution to technology, GUIs represent an epochal shift in how we interact with computers.
We are currently at the inflection point of the next one-Natural Language.
An increase in abstraction at the back end enables a decrease in abstraction for users.
As users of analytics tools themselves, product builders are also users, and can greatly benefit.
To delve deeper into the transformative power of NLP in product analytics, let's compare the workflow for traditional click-based platforms against an NLP-enhanced one, focusing on four critical phases: Data Access, Data Exploration, Insight Generation, and Taking Action.
In conventional platforms, access to data relies heavily on structured query language (SQL) or other similar architectures. Despite their robustness, these methods necessitate comprehensive data schema knowledge, meticulous query construction, and multiple clicks for data filtration and segmentation.
This often necessitates a linear and rigid process, hampering efficiency.
Conversely, NLP-enabled platforms provide an intuitive interface where analysts can pose queries in natural language. Similar to a Google search, one can just request ‘Show me user engagement metrics segmented by region for Q2 2023’.
The AI then translates this into an equivalent SQL statement, circumventing layers of interaction and dramatically accelerating data retrieval, while also enabling those without the required technical know how to procure this data.
In addition to making current PMs jobs easier, this drastically reduces the learning curve required to interact with analytics tools as well, getting rid of the data bottlenecks that exist as a result of the UX of current analytics tools.
Traditional, click-based interfaces guide users down pre-established paths of data exploration, often limiting the visibility of potential insights outside these paths.
NLP provides a much-needed deviation from this norm. Analysts can ask open-ended questions like, 'Identify anomalies in user behavior over the past month,' or 'Rank features by their utilization rate.' These inquiries guide the AI to perform complex analyses, such as multivariate analysis or outlier detection, offering insights previously hidden in click-based interfaces.
In essence, NLP allows users to carry out exploration in a more narrative or conversational way, enabling users to have a 'conversation' with their data, asking questions, exploring hypotheses, and deepening their understanding iteratively.
By breaking down the barriers to data exploration, NLP can help unlock the full potential of your data, and in turn, drive more informed and effective decision-making processes in product analytics.
Guiding strategy is basically what data’s role in product building boils down to.
Data’s the crude oil, insights are the petrol, and the product development is the vehicle.
To that end, the way in which NLP can augment product builders capabilities is even more significant than the addition to speed and convenience that they get.
Consider the daily workflow of a product analyst, Alex. Tasked with investigating user drop-off rates for a recently introduced feature, Alex begins by downloading user interaction data.
He segments the data based on user demographics and painstakingly creates visualizations to spot patterns. After numerous cycles of segmentation, cross-tabulation, and visual analysis, he identifies a specific demographic segment that exhibits high drop-off rates.
This process, while ultimately effective, demands extensive time and relies on Alex's in-depth expertise.
Now, imagine an alternate scenario in which Alex works with an NLP-powered platform. He wants to explore the same issue but approaches it differently. He enters a query: 'Identify demographic segments with high drop-off rates for the new feature in the past month.'
Behind the scenes, tools like named entity recognition, dependency parsing, and topic modeling can comprehend Alex’s query (the TLDR is that they help computers understand language). It translates his natural language request into a structured query language, processes the vast user interaction dataset, and swiftly spotlights the demographic segment with unusual drop-off rates.
Let's imagine a scenario in which Alex is trying to analyze the churn associated with a new product. Traditionally, identifying the reasons behind user drop-off rates for this feature would require significant data wrangling and extensive queries. However, with an NLP-enabled system, Alex can type a question such as, 'What factors are contributing to the high drop-off rates for the new feature?'
In another instance, Alex wants to improve user engagement with another feature. Instead of pouring over past data and trends, Alex asks the system, ‘What strategies have led to increased user engagement with similar features in the past?’ The system analyses historical data, factoring in implemented strategies, their timing, and the subsequent change in user engagement. It might suggest, ‘Offering a tutorial on feature usage has led to a 15% increase in user engagement in the past.’
NLP not only fast-tracks the journey from raw data to actionable insights but also simplifies these insights, reducing cognitive load. It minimizes potential errors and biases intrinsic to human analysis, yielding more accurate and objective insights.
In addition to these improvements in efficiency, NLP also empowers product analysts like Alex to evolve their roles. With NLP shouldering the burdensome task of data extraction and interpretation, analysts can pivot to more strategic, impactful roles. They can focus on deriving business strategies from insights, designing better products, improving user experience, and driving growth – an elevation that marks a substantial stride forward in the field of product analytics.
As this comparative journey illustrates, NLP has the potential to radically transform our interaction with product analytics platforms. By introducing more intuitive, dynamic, and efficient workflows, NLP transitions us from a static, rigid and unintuitive experience to one that mirrors natural human communication. This shift significantly enhances the speed and quality of insights derived from raw data, ultimately fast-tracking our journey from raw data to actionable insights.
