Blog

Is it Time to Use Artificial Intelligence in My Retail Sales Forecasting?

Posted by Sam Lowder on Jul 8, 2019 8:00:00 AM

AI in retail real estate site selection

You’ve probably heard about it, but what exactly is artificial intelligence? Artificial intelligence or “AI” is the process by which machines simulate human intelligence to solve problems and make decisions. There are many different outlooks on AI, but whether you believe that AI will be a positive force for society or the downfall of humanity, it is impossible to ignore the ramifications of this technology. 

AI in retail real estate site selection

If you are in the business of retail real estate selection and evaluation, you have undoubtedly heard the claims about artificial intelligence (AI), aka machine learning, improving sales forecasting accuracy and algorithms that “learn” and evolve as your business grows and changes. You may be wondering if it’s time to abandon the methods you employ today or whether adding AI to your current process is worth the investment.

Are AI models better than traditional models?

Before we attempt to answer that question, let’s address the hype.  Is the accuracy and flexibility of an AI model an improvement over a traditional statistical model? 

In most cases, the answer is a resounding yes.  Although there is value in employing models and processes that do not use AI (particularly for companies that do not have the quantity or quality of data necessary for a robust forecasting model) most chains that meet the requirements for employing statistical sales forecasting will see improvements with AI.

So how do you know if you are ready to start realizing the benefits of AI?  Here are several questions to ask yourself:

1. Where are we in our lifecycle?

Although you don’t need thousands of retail stores to employ AI, in general you will get better results the more data the model has to learn from.  Thus, a company with five locations might be able to collect thousands of data points about their customer and use AI to predict customer behavior, the company will be unlikely to build an accurate sales forecasting model based off of five data points (stores).

2. How willing am I to invest in data?

The quality and quantity of your data isn’t just a byproduct of growth.  It is an investment…of time to gather site attributes for each of your existing locations and an investment in licensing quality third-party consumer demographics, spending data, or market data. The results you get out of your AI model are directly related to the data you put into it. 

3. Are you comfortable with the trade-offs that come with an AI model?

Although accuracy is a major goal of any forecasting model, it is often difficult for companies to stand behind forecasts that are not transparent. With most statistical models, it is fairly easily to deconstruct an algorithm to see how it was calculated.  Although AI has become more transparent recently, there is still a “black box” component to these models.  While you can easily see what information a model uses to make decisions, it’s not always possible to understand why your model reaches the conclusion that it does.

4. How fast do you need your forecast models?

AI models can take significantly longer to run than conventional models. Often, the model is gathering and evaluating much more data. And in the case of an “ensemble” model, it is building and running the equivalent of hundreds or even thousands of models or simulations to arrive at a forecast. For a single site, this may mean that a forecast that took seconds to calculate previously might take a minute or several minutes with AI.  The better result is worth the wait, but for more complex problems and very large datasets, AI models can take hours or even days to complete.  Depending on the needs of the business, AI might be impractical. It might be necessary to choose AI approaches that factor in both accuracy and efficiency.

5. Should I replace my existing process with AI?

Despite the shortcomings discussed above, we are big proponents of AI.  We believe that most companies that have historically relied on conventional statistical models or more low-tech approaches to site selection and evaluation, will see measurable improvement with AI. 

However, AI is simply another tool – albeit a powerful tool – to help the decision-maker make better decisions.  It doesn’t replace the decision-maker.

We believe that your site selection and sales forecasting process should provide the confidence you need to make decisions you can stand behind.  For most companies, that means employing approaches like AI that make sense of what data is known about a site, alongside approaches that help you deal with the uncertainty of the unknown. 

Still unsure what site selection modeling approach is right for you?  Contact us today to learn more about our predictive modeling services. 

 

Topics: Retail Predictions, Real Estate Analytics Tool, Retail Industry, Artificial Intelligence, Analytics Solution, Machine Learning