Reforming Retail

The Massive Challenges for Labor Tools and What A Good Solution Requires

There’s perhaps no more hot-button issue for hourly jobs than labor. As unemployment rates cratered toward sub 5% it was virtually impossible for retail establishments to keep people on the payrolls.

“We’d have people scheduled to work and they wouldn’t even show up.” <- basically every operator’s frustration.

Robotics is obviously taking a massive bite out of these problems, as is the online ordering/kiosk channel which requires fewer employee-to-customer interactions. Thank you, technology.

But it will still be another decade before robots et al remove a significant percentage of hourly labor, and even then there will still be managers and people hired to do more than flip a burger (sorry burger flippers).

Solving labor and staffing is not a difficult challenge, but it is hard. If you’re asking what the difference between the two adjectives is, difficult are things that are cerebrally rigorous (e.g. give any of these unsolved math problems a whirl) whereas hard are things that take brute force (e.g. ultra marathons or Navy SEAL hell week).

One might quibble that some of what we’ll propose is difficult, but it’s not that difficult. Yes, data science is required, but there are plenty of people who specialize in that these days should you be willing to invest in it.

The three tenets are:

Predict demand

The hallmark of any good labor solution should be forecasting demand. In our early years we had to build all these models from scratch, now you can rent an algorithm on AWS, Azure, and even Facebook. Still, the best algorithms are built by people with lots of data and lots of time to train said data, which is why we’re starting to see algorithm-as-a-service models emerge. These models are modestly more accurate, but do matter when you’re talking about orders of magnitude, like if you’re a hedge fund, airline, or even if you’re someone the size of McDonald’s.

And we don’t care what some bundled back office or POS provider tells you: their algorithms are not as good as a company that focuses on algorithms all day.

Handle compliance

When legislators get bored they write laws. This is their favorite past time because it makes them important: you then need them to interpret the garbage they wrote, guaranteeing their demand and therefore income. #lifegoals.

There exist no more absurd laws than the ones around predictive scheduling. Seriously, if you want a good laugh read any of the coastal nonsense that poor businesses must adhere to that make running a low margin operation even that much more difficult. If you wonder why business owners are turning to machines, this stuff will clear it right up.

Staying in compliance is hard. It’s a lot of tedious work, coding the right shift breaks, labor rules, etc. that keep the operator in compliance. And it’s not as if the laws are stagnant, so it’s foolish to dismiss as one-and-done. The penalties for non-compliance are so severe that it makes many businesses give the whole idea of staying in business a second guess: is the risk worth the reward? Is my $30,000 in annual profits worth a $100,000 non-compliance fine?

Schedule optimization

If any part of this whole process could be defined as difficult, this would be it. Here we’re talking about using data science (that you’re not renting from a third party) to look at individual and team performance to determine who should work what shifts. Furthermore, a good system should even recommend different shifts for top performers by interpolating performance based on a number of factors.

We’d also prefer systems to quantify the value of such moves: if Bob works Tuesday, how much more money could he make the business? If Suzy wants to swap shifts, what’s the economic toll on your financials? We gave up on the compliance piece of the puzzle and settled on this as part of our labor prescriptions, though we’re not building schedules anymore because, as we say, you can’t do everything well (just don’t tell a POS company that).

A thoughtful labor system will do all three of these things, and it’s not cheap. There’s a real moat to get to this level of precision. We expect future systems to even build these schedules automatically, understanding the ratios of manager to employee type A and employee type B.

Data science is very useful, but if you’re not willing to pay for it you’ll never get the benefits. In a tight market with thin margins, it can make the difference between a profit and loss. Think about it.

Add comment

Archives

Categories

Your Header Sidebar area is currently empty. Hurry up and add some widgets.