TimeForge’s sales forecasting system is pretty amazing, and for good reason. We leverage heavy-lifting artificial intelligence (AI) and machine learning algorithms to build the best sales forecasts possible. It sounds complicated, and it’s certainly complex, but the results are straightforward. Sales and labor metrics are presented in easy-to-read charts and graphs that managers can view from a computer or even from a mobile device.
Most often, we forecast sales in order to help managers build better labor schedules much faster than they could by themselves, but the engine is quite robust. It can also forecast almost any metric that a business might care about, just as long as there are historical data available. This includes things like customer counts, transactions, and even online orders.
Below, let’s look at what sales forecasting is and some advantages to using sales forecasts. Then, we’ll take a look at what makes TimeForge’s forecasting system stand out. Fair warning: I’m going to nerd out a little bit, but I’ll try to keep things fairly simple.
What is sales forecasting?
Sales forecasting is the process of estimating future sales, usually by predicting how much product or service you’ll sell over a given period of time. There are actually many ways to forecast sales, but they all have one thing in common. They all rely on having information about the past (historical data), which can be used to make predictions about the future.
Why is sales forecasting important?
Accurate sales forecasts are important because they allow businesses to make useful predictions about the future. Every transaction contains useful data points, and if you don’t take advantage of these data points, you ultimately lose out. When you can’t make accurate predictions about the future, your plans may not always align with available resources.
In the retail industry, we see this most often in the day-to-day operations of store managers. Without accurate data, managers can’t easily predict how many employees and what positions to schedule in advance. Schedule too few employees, and you end up with long lines and frustrated customers. Too many employees, and you waste valuable labor dollars on staff you don’t need.
Over time, the results of both overstaffing and understaffing can add up to noticeable and significant losses in revenue and profit. This is why sales forecasting is so important to our customers – and why we work so hard to engineer the best forecasting system possible.
Forecasting in Areas with Secure Scheduling Laws
First, because these laws require employers to post work schedules weeks in advance. And second, because the laws penalize employers for making changes to posted schedules. Managers in these situations must therefore accurately predict their labor needs 2-3 weeks out or else get “nickeled and dimed” by fees. Over the course of a year, these fees can add up across a business. Think: the cost of a luxury sports car or your kids’ college education. They’re no joke.
How is TimeForge’s sales forecasting different?
Here at TimeForge, we have the benefit of over 14+ years’ worth of experience in both retail and hospitality labor management. We also employ some pretty smart folks. Our forecasting system uses a wide variety of machine learning and AI methods to calculate sales on a per-week, per-day, per-hour, per-15-minute, and per-department basis! Many forecasting applications can’t get that granular and precise; TimeForge can.
3 Benefits of TimeForge’s Improved Forecasting System
In the past few years, we released an even more powerful version of our forecasting engine, which has three major advantages:
- Improved accuracy during holidays, including the days leading up to and after. Our system can accurately account for and predict the effects that holidays will have on sales based on previous years’ performance during the same time period.
- Increased speed. Our system can run a full forecast on the fly in only 20 minutes. This leaves more time for addressing issues ahead of schedule.
- Improved accuracy of department sales distributions across the workday. Our system has become even better at predicting the distribution of sales across the workday within a department (e.g. produce, meat, etc.).
Because our improved forecasting engine uses AI modeling, the more sales data you have, the more accurate your forecasts will be. This is true for all systems that use best-fit statistical models to predict future trends.
How does TimeForge’s sales forecasting engine work?
Okay, so we can’t give you the full recipe to our special sauce, but I can tell you a few things about how the system works:
- Using API calls, the forecasting engine retrieves each store’s sales data from TimeForge. (Sales are typically pulled into TimeForge through one of our many POS integrations.)
- The engine filters and pre-processes the data to make sure everything is good to go. We wouldn’t want to run forecasts on bad or incomplete data!
- The engine then selects a statistical model to create the best fit for the data across time. This involves cross-validation and fine-tuning to achieve the best fitting model for each store.
- The engine looks at the data for things like sales trends and seasonality. The model extends these forward in time to achieve a statistically sound forecast.
- The forecast is then sent to TimeForge for use in generating amazing schedules and reports.
Ow, My Brain Hurts
Too much jargon? That’s fair. Let’s look at a specific use case: predicting how much extra labor you will need for an upcoming holiday. Holidays are super important times for business in retail and food service, so it’s important to get them right.
For example, let’s say we’re building a schedule for the week of Thanksgiving. TimeForge’s forecasting system looks at all the Thanksgivings and calculates the effects that the holiday had on sales. It will then account for those effects into the forecast. Then, you can either use those forecasted sales to build a work schedule, or TimeForge can auto-generate the schedule for you with all the staff you’ll need. The key, of course, is having sales data for at least one Thanksgiving in the past (and more is better). Otherwise, any system would have a hard time accounting for it.
Actual Sales and TimeForge Forecasted Sales for the Week of Thanksgiving
How do sales forecasts compare to manager projections?
This is one of those instances where a picture (or in this case a diagram) is worth a thousand words. Often, manager projections are based on simple rolling averages that look at the past 6 weeks’ worth of sales. Let’s look at our Thanksgiving example again, except let’s compare manager projections to TimeForge’s latest forecasting engine.
In the diagram below, we can see that the manager would have underpredicted sales leading up to the holiday (11/22). They also would have overpredicted sales on the holiday and days to follow. This makes sense, as shoppers are usually in a rush to buy their turkeys and pie fixins before Thanksgiving, but they’re generally going to be coasting on leftovers a few days after that. A rolling 6-week average wouldn’t be able to account for the effects that the Thanksgiving holiday has on sales. Thus, a schedule based on the average would have resulted in first an understaffing situation and then an overstaffing situation.
Sales for the Week of Thanksgiving: Manager Projections vs. TimeForge Forecasts
Now let’s take a look at accuracy. On the days leading up to and directly following Thanksgiving (11/22), the rolling 6-week average becomes less and less accurate, dipping to under 20% on the Thanksgiving holiday. TimeForge’s forecasting system, on the other hand, maintains a high accuracy throughout the week. Holidays can be one of the trickiest times of the year to predict, and yet TimeForge can do it with incredible accuracy.
Accuracy of Manager Projections vs. Sales Forecasts for the Week of Thanksgiving
And there you have it! TimeForge can accurately predict holiday sales, which means TimeForge can build a better holiday schedule. And not just for the day of Thanksgiving but for the days leading up to and after the holiday.
One of the amazing things about our forecasting system is that it’s independent of platform. Meaning, it’s POS-agnostic. It doesn’t matter what point of sale solution you use. As long as we can integrate it and pull sales, we can make our forecasting engine work hard for you.