Real Time Data For AI: Why I Invested in FeedbackNow
Where do models go in the next 5 years?
Happy Sunday and welcome to Investing In AI! If you are new to the newsletter be sure to check out our AI Innovator’s Podcast, and in particular the latest episode with Jaclyn Rice Nelson, CEO of Tribe.ai. We talk about enterprise AI adoption and where it is really happening.
Today I want to highlight an investment I recently made, and talk about the non-obvious way it fits into my thesis on where AI is going next. I participated in an investment group that bought FeedbackNow out of Forrester. Before I talk about why this is such a critical technology for AI, I want to give a little history on the company and use cases.
You probably know FeedbackNow as those buttons in airport bathrooms that you press to tell someone they need to be cleaned. That’s where the business started over a decade ago. Now it’s a real time operations platform for physical spaces that integrates with over 100 different types of sensors. Airports, stadiums, hospitals, and convenience stores use it to optimize revenue and labor costs.
For example, in airports, FeedbackNow’s platform saves labor costs by predicting when cleaning crews need to go to certain areas of the terminal, and how to staff better at TSA. The same algorithms optimize revenue by ingesting sales data from airport stores beyond the security gates and correlating it to wait times. So they can tell you something like “people will wait in line for 10 minutes at TSA but for every minute longer, on average, you lose $1700 in sales on the other side.”
In hospitals, the value prop is even more powerful. FeedbackNow’s platform saved one major hospital $2M in labor costs by improving utilization of nursing staff. And it improved revenue by increasing patient satisfaction (which is often a component of reimbursement) to levels never seen before in the hospitals that use it.
I invested because I think businesses like this have a ton of AI potential for a few reasons. First, the business benefits from data network effects. Take the simple example of an airport bathroom.
Step 1: Install buttons in the bathroom so that instead of sending a cleaning crew every 4 hours, you send a cleaning crew once 3 people press the “needs cleaning” option.
Step 2: Add a people counter at the door so you know how many people go in and out. Now correlate that to the button presses. Now you learn that on average, after 120 people use the bathroom it needs to be cleaned, so, when 110 people have been in the bathroom you can queue up a cleaning crew to go soon. (Data network effects because the people counter data makes the push button data more valueable when it is added together)
Step 3: Ingest flight and gate data so that you know how many people are near a certain bathroom. Now you know if there are 700 people in the gate area over a 2 hour period, roughly 100 of them will use the bathroom, and that means you can predict when it will need to be cleaned based on gate and flight info, and can update that prediction in real-time as gate data changes with flight delays, cancellations, and early arrivals. (The gate data makes all the previous date more valuable)
Data network effects in the real world are difficult to find but this business definitely has them.
The second reason I invested is that at the moment, AI models don’t take in much real-time data. Large foundation models are trained on data at some cutoff point weeks or months before because the training itself takes so long. As these models grow and perform more tasks, they are going to become hungry for real-time data, and FeedbackNow will be well positioned to provide some of it.
The third reason I invested is related to the real-time data aspect again. Businesses have real-time data around many workflows like their supply chain, or website visitors or ad performance. But most don’t have data and algorithms about what is happening in physical spaces as people move around. How long to they stand in line? What are the noise levels like? How do noise, line length, crowds, cleanliness, etc impact revenue? The real-time data from FeedbackNow, combined with predictive algorithms on that data, let businesses run their physical spaces like they would run their websites, constantly testing and optimizing.
And then, since this was a corporate carveout, AI theses aside, there was just a lot of low hanging fruit to fix by getting it out of the parent company and making it an independent business optimized for its own needs rather than those of the parent company.
Early/Growth stage investing is all about making bets on the future, and I really believe over the next 5 years models are going to be hungry for real-time data of all types. If you run an AI business dealing with real-time data I’d love to chat about it and learn more. And if FeedbackNow sounds cool as an investment, the company will raise a growth round next year so please reach out if you have an interest.
Thanks for reading.