How to Make Better Business Decisions by Applying Next-Level Data

In today's Confab - our LinkedIn live event - our Stoneridge Architects talked about their predictions for the business application space in 2021. Their #1 prediction was that 2021 would be the year of data. I agree with them and I had just written this article for our blog to talk about the type of BI businesses need to run their business. Here's the blog - let me know if this will be an initiative for you in 2021.

Everyone agrees that reports and business intelligence is fundamental to running your business. It's the same way that stats are important in baseball or any sport. The book (2003) and movie (2011), Moneyball, told the story of the value of the next level of data. For 100 years, baseball executives looked at batting average (BA), home runs (HRs) and runs batted in (RBIs) and assessed players based on that data. Starting with the writings of Bill James in the late 1980s and highlighted by the A's of the mid-2000s, it became pretty clear that if you want to win in baseball, you need to know more than how many RBIs someone had. Today's baseball teams have taken it a step further and now most teams have data analytics departments who dig up every scrap of data that may lead to a competitive advantage.

You don't need to hire a 20 person data analytics department to know your business well but I would argue you can't run your business well with the most surface-level statistics. Take Stoneridge, we are a consulting company so the most important metrics to us are utilization and backlog. Utilization is the percentage of billable hours our consultants compared to total work time and backlog is a forecast of the amount of expected hours of work our consultants will have over the next several weeks. There is great value in the basic utilization number and the backlog number. They are appropriate metrics for us to use to look at our business at the highest level. When you measure utilization at an individual level or over certain periods of time, it isn't that useful. When we look at our utilization over the week of Christmas, it's awfully bad compared to our standard expectation. That's completely driven by the fact we have most of our team on vacation and planned holidays that interrupt our ability to do billable work. 

Expected vs Actual

The whole reason business owners make budgets is to have a target to compare your performance. If you said you made $5m in profit last year to me, that sounds pretty good, I guess? If you made $5m in profit against a target of $100m, that is pretty bad. If you made $5m in profit against a target of $500,000, that is really good. You need to use the same concept to metrics within your business. About 4 years into Stoneridge, I finally realized that the weekly utilization number doesn't tell me that much unless I have an "expected" utilization number for each week. Every time we would come to the week of Christmas and see 36% utilization, I wouldn't know if that was good or bad. If our target is 67% utilization for the year, that certainly seems bad, but after adding the expected utilization metric I started to see that 36% was higher than the 32% I expected for that week. On the flipside, January is generally a very busy month in the our world, so sometimes we need to hit 72% utilization during those weeks to make our annual target. If we were humming along at 67% in January without an expected utilization target, I might think we were on target for the year. Then when summer hits and things slow down, we'd be below target because we didn't bank up enough hours earlier in the year.

To have an effective "expected" utilization, we had to develop a separate system for tracking this. We ended up using our model-driven Power Apps environment to set expectations for every billable consultant in our organization on a quarterly basis. We estimated their hours after subtracting out holidays, expected time off, training hours and admin time. We also needed to factor in the weekly impact of holidays and time off so we needed to track that as well. Once we had systems for each of these data points, we could build a metric of the expected hours by person all the way up to by company.

The challenge with this level of reporting is it's not always easy to get expected data out of your system. If you can get that data from your current systems, great, but in most cases you need to use Excel or Power Apps and then look to combine the data to create a report. That's what we had to do to get there but now we have a very meaningful metric that is critical to our ability to track how we're doing.

Leading vs Lagging Indicators

The backlog metric is really a leading indicator on future utilization. It's a metric that tells you how many billable hours you expect to have in the future, factoring ongoing client work and future consulting sales. It is never perfect but it is critical for our business to know when we need to hire another consultant to keep up with the demand.

Leading indicators are metrics that will tell you what your future metrics are likely to show. Lagging indicators reflect something that has already happened. We can't predict next week's utilization based on last week's utilization - it might give you some hints but it isn't predictive. It's the same as predicting a baseball player's home runs next year based on the fact he hit 20 this year. In baseball, leading indicators would include 1) the player's age - they typically perform better up to age 27 and slowly drop off after that, 2) injury history - was he injured last year? Is he injured now?, 3) the amount of playing time they got last year versus the coming year, 4) is he playing for the same team in the same ballpark and 5) a host of deeper metrics tracking their strikeout rate and exit velocity on the balls they hit. As you can see, historical statistics make up one variable that helps predict the future but there are far more variables to build into the equation.

For manufacturers, you need to be tracking expected demand to know if you're building enough product. Historical demand is helpful but there are far more variables that affect future demand. Every situation is different but the more variables you can bring into your projections, the more accurate your forecast will be. It's also an opportunity to take advantage of machine learning to build a model based on your historical activity across the different variables. Scott Frappier recorded a video on our YouTube channel about how this works: Machine Learning and Dynamics 365 Finance and Operations

To better predict future performance for your metrics, you need to identify those leading indicators and build a mechanism to track them. Like your "expected" numbers, you may have to track data that's not in your ERP or CRM system to put it all together.

Questions to Ask Yourself

As you think about what metrics you have today, I'd encourage you to ask yourself a few questions:

•  What "expected" metrics do you track today? 
• Do you have the data to help you build the expected values?
• Do you have metrics that help you forecast?
• Are you tracking all the variables that impact your forecast?
• Are you using machine learning to help build a model for your forecast?

I'd love to hear your thoughts in the comments about what expected metrics you're tracking today.