The PowerMetrics AI experience - Learn why metrics provide rich context for better AI

Metrics provide rich context for better AI: Introducing the new PowerMetrics AI experience

Published 2024-06-21, updated 2024-07-08

klipfolio image

Summary - Today we’re launching two exciting features that will change the way you interact with your metric data! Our new generative AI experience, called PowerMetrics AI, lives in Explorer and is powered by our metrics platform and OpenAI’s GPT-4o. Our new account-wide search capability increases productivity by helping you find assets quickly. These features work together, along with the context provided by your metric catalog, to provide visual answers to questions about your data.

Generative AI is a transformative technology that’s having profound impacts on the way users interact and learn across all industries. The promise, if applied thoughtfully and with the right context, is that AI will elevate our understanding of meaning and help us focus on the most important parts of our data based on the information available to us.

For many years, our PowerMetrics team has been considering the potential impact of AI and how it could transform business analytics. It seems almost every month, new large language models (LLMs) are being announced by leading vendors in the data space, most recently by Snowflake (with Arctic and Cortex), Databricks (with DBRX), and Google (with Vertex AI and Gemini). It’s clearly a fast-moving area of technology, and although it’s exciting to see the performance of each model, it’s unclear what the quality of the output will be. Side note - We recently launched full integrations with Databricks, Snowflake, and Cube not only to enable our customers to leverage their data warehouse and semantic layer capabilities but also because of their leadership on the AI front.

Most solutions to date have been focused on data engineers and developers, with innovations such as faster and more optimized SQL queries, code generation, automatic data cleansing, and other discrete functions. This is all great, but how do we unlock the potential of this technology for non-data team members? Up until now, no one has really delivered on the “next generation” of advancements for business users.

Our business users have often asked us to help them answer questions about their data since we understand their fundamental measures, customer segments, and core data sources. For our third generative AI application (MetricHQ Metric Assistant and PowerMetrics Assistant being our first two), our goal was to deliver a generative AI feature that could turn insight into action for business users. This became our north star and we’ve been investing in the solution for the past 6 months. You might ask, what’s been holding us back from releasing it until today?

Why Context Matters

The key learning for us, like many of the best AI narratives, begins with context. We needed to find a way to provide our next-gen AI with the solid contextual foundation our customers built using their business data. Modeling and understanding the fundamentals of any organization by looking at the vast amount of information they have can feel overwhelming. However, it reminded us why we created the concept of metrics as first-class objects in our platform. We capture every measure, dimension, and data set being used and represent them in our metric catalog, where the data has clear business meaning. This establishes the fundamental business relationships within the organization and provides a comprehensive view of the entire business.

We also spent a lot of time making sure prompts were being properly interpreted and that they related directly to the user’s metrics. This effort helped ensure that relevant, accurate answers were provided more often. For example, if the question asked to the AI was ambiguous or it could be answered with multiple metrics or in multiple ways (either potentially correct), we wanted the AI to ask the user for additional details or clarification before proceeding. It took time to make our solution compelling but it was worth it!

Today, we’re excited to introduce PowerMetrics AI - an OpenAI powered add-on to our metric catalog and analytics platform. Our goal is to enable true analytics self-service for business stakeholders. With this new capability, we’re demonstrating that our vision for AI goes well beyond chatbots. PowerMetrics AI helps teams have a real conversation with their data, to build understanding and drive better decisions. Questions that used to take you days to ask your data team and hours for them to find in the data can now be answered in seconds. This new capability will help teams transition from navigating a sea of numbers to exploring the “why” and “what’s next” with their data.

What is PowerMetrics AI?

PowerMetrics AI uses the latest OpenAI GPT-4o model and other related technologies to provide an incredibly fast, trustworthy, and interactive experience. Any user can now ask natural language questions of their data, relying on their own metrics to provide a rich context. Business users can get answers to these data questions instantly - eliminating their reliance on a time-strapped. resource-constrained analyst or data team. At the same time, the data team remains in complete control of the metrics and data being used by the AI for consistency, trust, and governance purposes.

Launch 4

For our initial launch, we’ve exposed PowerMetrics AI within the Explorer - the location in PowerMetrics currently reserved for free-form analysis of single and multiple metrics. The Explorer already lets you manually add metrics to visualize them in a single chart but now you can prompt PowerMetrics AI to add them for you. Regardless of how you explore your data, you’re free to investigate and dig in without affecting the metric’s underlying data or its default display for others. When you discover something of interest, save the visualization to a dashboard for easy sharing with others on your team. 

PowerMetrics AI enables:

  • Natural language questions to be asked within the Explorer with the answer being a visual representation of a metric (or set of metrics) that currently exists in your metric catalog.
  • The creation of a dynamically generated example prompt based on the metrics in your account. This is a great way to get started exploring your metrics.
  • Any question to be added into the prompt entry field, with a limit of up to 200 characters.
  • Users to see the history of the prompts they previously entered. You can go back to any step, and see the visualization as it was displayed at that point, by clicking the entry in the History list.
  • The History log to be downloaded as a file with the text for each prompt and the time and date it was entered - ideal for record keeping purposes.
  • Specific metrics to be referenced quickly, using a prompt beginning with the“$” symbol. Doing this displays a pop-up list of metrics matching your criteria.
  • Integration with our advanced analyses, such as outlier detection and forecasting, to provide better answers to your questions such as “How are my sales looking for the next 30 days?”
  • Investigation of your data and automatic creation of visualizations. It can be used in combination with manual interactions in the Explorer or it can automate the entire experience.
Level up data-driven decision making

Make metric analysis easy for everyone.

Get PowerMetrics Free

Is My Data Secure? What about AI Hallucinations?

We are not training our AI model with your data. With PowerMetrics AI, no customer data is sent to OpenAI or made public. Only non-confidential information is used when interacting with OpenAI apps. Your data and metrics are safe within the PowerMetrics platform.

The real differentiation in our PowerMetrics AI solution, besides using the full metric catalog to provide a very rich context, is the fact that it will never hallucinate. There have been numerous discussions on the Internet about how often chatbots come up with a completely fabricated, nonsensical, or inaccurate answer that is not based on any fact - a.k.a. an AI hallucination. We do not use OpenAI to calculate the resulting visual chart - we only use our metric platform and the underlying models within the metrics themselves that are based on your customer data so the result is always accurate. OpenAI is only used to extract appropriate context from your question and the metrics within your catalog that it identifies as relevant. As a result, the only thing that might occur is a slight misinterpretation of the question you’re asking. The visual, metric, and data used to answer your questions will always be 100% accurate.

What are our New Search Capabilities?

Our new account-wide search is leveraged by PowerMetrics AI and also accessible via the Search icon on the side nav. It uses a natural-language optimized model to provide a fast way to find, view details for, and go to metrics and dashboards in your account. Using a familiar experience, similar to Spotlight on the Mac, you now have one location to describe and find exactly what you’re looking for within PowerMetrics. 

Launch 5

Account-wide search enables:

  • A search term or string, for example, a metric name, dashboard name, or service name to be entered into the search field. The search refers to text in the metric or dashboard title, description, dimension names, and service name(s). 
  • Search operators such as + (AND operation), | (OR operation), - (negates a single token), “ (Wrap a number of tokens to signify a phrase for searching), or use “(“ and “)” to indicate precedence.
  • Users to only find assets they created or that have been shared with them.
  • The results window to auto-display up to 5 assets for each category (metrics and dashboards). If your search returns more than 5 results, you can see the rest of your metrics and dashboards by clicking Show more metrics or Show more dashboards.
  • Certified metrics to appear in the results list. Hovering over the certified symbol displays certification details.
  • A details pane to be revealed when hovering over an asset in the results list. This shows additional information, such as, the owner, type of metric, and when the data was last refreshed.
  • An asset, such as a metric or dashboard, to be opened in the current window or in a new tab from the results list.
  • Questions to be asked to our AI-powered PowerMetrics Assistant, which searches our Knowledge Base articles for an answer. You can also access the Knowledge Base and the What’s new, and contact our Support Team, all from the PowerMetrics Assistant UI.

Get Started Today

PowerMetrics AI is included in the 30-day premium trial for all new customers. When the trial ends, you can continue to use this feature by purchasing the PowerMetrics AI Add-On for our Professional plan or by contacting our Success Team to learn more about our Enterprise plan. Our account-wide search is available for all users today.

This announcement is only the beginning of where we plan to take PowerMetrics AI in the coming months. We’re also working to bring deeper contextual awareness by leveraging all of the value-added features in the PowerMetrics platform such as metric goals and the relationships between different sets of metrics using unaligned data sources. We’ll also be working toward a better understanding of the APIs and data sources themselves, multiplying the relevancy, accuracy, and applicability of the feature to an even greater level. Stay tuned for more and we welcome your feedback!

Join us for a live webinar and demo Chat with your Data - with PowerMetrics AI on June 25th at 1 PM EDT / 10 AM PDT / 7 PM CEST. Learn about AI and account-wide search capabilities and more from product experts at Klipfolio and our partner Eyeful on these new AI and account-wide search capabilities and more. Register now! 

Related Articles

Level up data-driven decision makingMake metric analysis easy for everyone.Get PowerMetrics Free