How We’re Shaping the Next Era of AI Fitness Operations
"I've had the pleasure of keynoting at many conferences this year, from Active Uprising to Perform X and EHFF. Not surprisingly, the topic organizers wished me to cover was the rise of AI or some version thereof, as operators' interest level is at an all-time high. I have found audiences highly engaged and receptive to learning. Unlike in previous years (I've been banging the AI drum on the industry conference circuit for five years now), many more have started or are looking at starting implementations.
I've been involved in predictive analytics for more than two decades. I've built AI features on SaaS platforms for the last ten years. I've witnessed the impact this technology has on key business fundamentals, and I expect much of the productivity dividend being talked about by analysts and economists will indeed come to fruition as AI not only uncovers new levels of insight but also gives us the power to take consistent action to leverage these new insights. However, this low-hanging fruit is only the beginning; it will not just be a few new insights delivered. The potential for revolutionizing the majority of our existing workflows to levels of previously unimaginable productivity is within grasp, and at Keepme, this has been the work our development group is readying for release.
The ‘OG’ AI in Fitness Operations
Keepme has had AI at its core since its inception. I hypothesized in 2017 that the data within a fitness club could be highly predictive of member attrition using machine learning (ML). It wasn't a guess; I knew through statistical modeling that was the case, but there is a material difference between the 73-75% I could get through modeling and the 90%+ accuracy I expected I could get through using ML. However, I quickly found that my challenge was not just the accuracy. I needed the signal detected as early as possible to have the runway to change any potential adverse outcomes. That was where the real work came in, and a level of detail I suspect those now attempting to follow in our footsteps will have missed. It was at the end of 2018, nine months after we had first started that investigation, that we were confident that the 91-95% accuracy could be sustained from the first three months of an annual membership. Thus, the Keepme Score™, now used by hundreds of clubs worldwide today, was launched. Every week, we learn more, and every week, we fine-tune based on the new data and outcomes experienced and entirely new data offerings to understand the impact on both the accuracy and time horizon, such as wearables and NPS platforms.
This iterative process enables us to continuously improve the accuracy and timeliness of our insights. It's this level of detail that has some of our customers operating with 98% accuracy. You see, it's not just a product feature for us, it's part of what we do and our stated reason for existing, which is simple. We build products that allow our customers to increase member sales and retain them for longer, and that guides everything we do.
Member retention is not the only area where machine learning can be applied. There are many others, effectively any area of the business where there is a lack of transparency on future performance where a highly accurate prediction could be of value. Another area where we find great success is in lead scoring within the member acquisition process. We work with customers who have thousands of new leads every month, the acquisition of which Keepme plays an integral part. However, not all leads are equal, and visibility into which ones are likely to convert or produce the highest LTV can ensure resources are allocated to the right areas and inform decisions around acquisition channels and fine-tune to optimize the cost of each lead.
Turning Insights into Action: The Power of Automation in Fitness Operations
We learned the hard way that the real value isn't in providing insights alone. Despite being aware of issues, many operators struggle to take consistent and timely action. Most have an idea of what to do, but ensuring action is taken can be challenging. This is where automations come into play. I estimate that automations have consumed the majority of our engineering time over the past five years. We've focused on providing customers with the capability to deliver personalized, consistent digital engagements required for large-scale interventions. Let's face it: an insight without action is worthless. This is why we've concentrated on this challenging area and become automations leaders. We provide our customers with the ability to deliver a range of interactions, from simple daily engagements like welcome emails or birthday recognitions to complex multi-step member journeys involving hundreds of steps and variables.
This is an area in which AI will once again give us an efficiency and productivity boost. Building automations presents several challenges. First, you need to know what data items you can use. With many solutions, this isn't too challenging as they only provide a few options. However, with Keepme, our customers have anywhere from thirty to a couple hundred different filters and actions to choose from, and that provides both impressive flexibility and a need to understand what each data source represents.
Where We’re Going
We've seen these challenges firsthand, and they create barriers that prevent operators from deploying advanced, highly personalized engagements. To solve this, we've turned to large language models (LLMs) with a beautifully simple methodology. Gone are the needs for filter selections, condition choices, and an understanding of what's available. Now, you simply enter the automation you want to build in plain language, and Keepme does the rest.
For example, I could type, "New members in the last ninety days whose average attendance has dropped in the last thirty days but have attended in the last 7 days," and voila, the automation is ready to go. Keepme not only understands the instruction but also validates it against available filters and data items to build the correct automation or campaign within seconds. This massively increases productivity while eliminating errors. But there's more. What if we know the problem we want to solve but need ideas on how to solve it? You can say to Keepme, "What would an automation look like that could engage all members on rolling contracts, who have been active in the last 30 days, participate in Body Pump classes, and have either not provided an NPS or given a score below 8 in the last 90 days?" Again, Keepme will interpret the requirement and deliver the optimal automation or campaign setup, along with the next feature evolution. Through insights like the Keepme Score™, you can identify potential problems, and with the automation/campaign feature, you can target the audience to engage. But what do you engage them with? This is a very real challenge for operators, especially as members expect engagements to be both relevant and highly personalized. Let's face it: good content creation is hard! We started tackling this problem in 2023 with the release of Keepme Creator to our customers. Importantly, Creator is nothing like ChatGPT iterations, instead it leverages large language models (LLMs) and learns the brand's voice by analyzing its website, blog posts, and previous communications. It then allows customers to craft engaging content with just a few words of prompting. Whether it's emails, blog posts, or web copy, Keepme Creator produces output matching the brand's tone-of-voice, complete with appropriate vocabulary and the nuances unique to each brand. The feature has been well-received, but we're taking it to the next level.
We've been training a Large Language Model (LLM) on best practices for membership sales and retention. This training has involved thousands of academic papers, blog posts, and content from thought leaders, not only within our sector but also in wider membership/subscription-related industries, from streaming companies to hotel and airline loyalty programs. This means that Keepme customers can now leverage this enormous amount of knowledge in their campaigns.
So, Keepme can not only provide the correct prospect/member targeting but also suggest approaches and content based on millions of data points. I've been calling this the Sales and Membership Oracle, though I'm sure our marketing team will improve on that name. The intent is not only to help at the point of execution but to make this capability accessible throughout the Keepme product. Our customers can use it to brainstorm ideas, validate approaches, and generally have access to the most extensive set of best practices and approaches available anywhere.
But we're not done, not even close…
One area that isn't exactly sexy but is ripe for an overhaul is reporting. Hardly a week goes by when I don't hear an operator comment about the lack of transparency or availability of key reporting elements. We've come a long way in the last decade, and some advanced tools, such as Tableau, allow you to present just about any data in any format. However, that still requires an additional product and a new learning curve. We've tackled this by using the power of Large Language Models (LLMs) and natural language to develop a reporting feature that allows you to request what you want and have it presented quickly and elegantly. Gone are the days of being limited to a report library or structured date ranges. Now, if you want a report of all new members in the last seven days, by lead source and by site, you just type it (or even say it), and that's what you'll get. It's not just the ease of accessing the reports you need; it's a whole new use case. No more last-minute hustle is required to gather the necessary insights for a meeting or end-of-month reporting. In real-time, any question can be asked, and answers provided. It's like having a 24/7 business analyst available. We're just getting started in this area, and you'll see us deliver new AI team members to assist your business in many other areas, such as membership services, in the coming months. To be clear, this is not about staff replacement — far from it. This is about adding consistency and capability to specific areas of the business that will, in turn, free up the existing team to provide value to customers in ways that only they can do.
👉 You can learn more about our product developments here.The Road Ahead
I hope this post clearly demonstrates the capabilities that each of you can take advantage of in the coming months. We speak from a place of deep experience, having been the first and continuing to be the most innovative and capable provider of Fitness CRM solutions. If it were an AI feature war, we would win hands down, but that's not my message.
What I want to emphasize is that the AI revolution isn't about this feature or that. AI will transform how tools look, what they're capable of, and dramatically improve the workflows they participate in. From a technology perspective, we live in an era of unprecedented change, presenting enormous opportunities to deliver business-changing capabilities. At Keepme, we accepted this challenge five years ago. We hope that other technology suppliers will now grasp not only the opportunity but also the necessity to appreciate that the paradigm has changed. Software that doesn't evolve rapidly will not only limit our sector's capabilities but, quite frankly, cease to be relevant in a couple of years. From what I see, there are currently a few candidates for extinction."
Keepme Roadmap
Here we explore Keepme's pioneering journey and unveil our exciting product roadmap, featuring cutting-edge developments in AI-powered solutions set to further transform fitness operations.
