Crunchtime is an operations management software platform for the restaurant and food-services industries used in over 100,000 locations in more than 100 countries. The software is used for inventory management, staff scheduling, learning and development, food safety, operational tasks, and audits. Their top-line goal is using data to make decisions not only faster, but also smarter—all of which starts with a two-way relationship with data.
Director of Analytics Adam Armagost sat down with Sigma to talk through how they use a two-way relationship with data.
Sigma: How does data analytics work at Crunchtime right now?
The big outcome that our analytics team is driving towards is fast, profitable decisions without our customers waiting on us—or our processes. Our customers want—and need—the ability to drive change quickly and efficiently. The nature of the industry we're dealing with is complicated and nuanced. Data should be leveraged to test and validate assumptions or hypotheses by our customers in order to drive the best possible outcome, and our goal is providing that mechanism in the easiest, fastest, and most consistent way.
The standard operating procedure for the restaurant industry has been very reactive. The vast majority of our customer base is used to a more analog way of tracking data. I’m talking about things like pen and paper to track restaurant timing and inventory. If we can take our customers’ ability from being reactive in addressing issues they know or think exist to a more proactive manner, that means we're providing the right results. It’s not just about delivering data, it’s about delivering insights. We have data on all of the primary cost drivers for our customers.
How does working in analytics for restaurant operators tie up to bigger-level problems?
One of the biggest challenges we’re trying to solve is food waste. People don’t realize how much dine-in and carryout restaurants are trying to tackle this problem and be more efficient. Nobody wants to throw food away. The latest numbers are that something like 22-33 billion pounds of food must be tossed every year in our dine-in/carry-out experiences just in the U.S. So having actionable data there is one of the most important things we’re doing.
How does a two-way relationship with data help companies like yours make decisions faster?
Analytics can sometimes be viewed as almost modern-day black magic. Where you give us, the analytics team, data—and we give you answers. But that shouldn’t ever be the case, right? Analytics should be treated as a partnership with the customer (or end users) of the outcome we are helping solve for. Our customers have a lifetime of experiences, ideas, heuristics, and processes that we can and should leverage to solve challenges that are important to them. This also benefits our time to insights as it allows us to focus on testing and validating things they're already doing, vs. fishing for answers in vast seas of data.
As an analytics team, if we’re asked to do something, and we go heads down for six weeks and we pop up at the end and say, “Here's this amazing analysis that we've done.” And we're not able to explain that or we haven't taken our users or our stakeholders along for that journey—they're very unlikely to actually use that, right? It's spinning our wheels to a large degree. That's not a two-way relationship with data.
How is machine learning and AI affecting your priorities?
When it comes to business outcomes and making predictions, I could give you a model that is 90% accurate. But if I can't explain to you why the model is that accurate, and if the business user never looks at it again, that's the worst-case scenario. That's a slow march of death for any analytics program. If we are providing a prediction or a root cause to a customer, then we need to also provide our reasoning behind how we are drawing that conclusion in order to build buy-in.
How can companies move from simply having data to a two-way relationship with their data?
By focusing on solving customer outcomes first, and treating analytics as a partnership with customers. We could do all the amazing statistical findings and analysis, diagnostic modeling, and stuff like that. But if it is not fundamentally something useful for our end users, we shouldn’t be doing it.
We send task completion rates to customers, usually restaurants, so they can see over time how they're trending. But what would be more helpful for them would be: Why do we see task completion rates trend in one direction or another?
If our customer is seeing task completion rate drop, we then develop hypotheses on why that might be happening. Then we're comparing task completion rates as an outcome to different input factors. Data without context is a tough place to plan at the end of the day, because we have to be able to know the additional caveats.
How far are we in an analytics space from being able to actually be more predictive in the sense that like a restaurant could say, I know, with 80% confidence, if I switch inventory from here to here and decrease labor from here to here, it's gonna affect revenue this much next quarter?
We could easily equate the amount of food waste in a given restaurant to dollars. If we're talking tons of inventory, we're talking hundreds of millions of dollars easily a year. So if we're able to help them save 10% of that, from an operational cost… that's massive, that's huge for them.
And we're uniquely positioning ourselves to help do that. We've got the architecture built. That's an easier solve than going and writing a more complicated machine learning model, right? Simpler is always better in my opinion. There's a lot of untapped potential.