Prescriptive & Predictive analytics to the rescue -Out of Stock / Excess Stock situation.

As a part of our engineering process at Deskera, we are always brainstorming and innovating to evolve the Deskera Books Platform to give users useful features with the best user experience. We will discuss today one of such useful future feature and its implementation approach.

Problem Statement for future feature

Deskera Books platform has been widely adopted by mid-size companies that need inventory tracking. As part of an application, users are able to add and track inventory, and it is auto managed. What we want to provide users with is the ability to know if products are going out of stock well in advance, so their business goes smoothly.

Proposed Solution

Now how can we achieve this? Well, we can solve this problem by monitoring the product consumption trend (using historical inventory data) and by doing sophisticated predictive analysis or prescriptive analytics with the use of statistics and modeling.

So, for example, if a user has 100 Umbrellas in his stock before the start of the rainy season and usually they sell 200 in the same season (say we have inventory data for the last couple of seasons), then the predictive & prescriptive data analysis will predict to a user that they can go Out of Stock and are short on stock for Umbrella and will prescribe to stock more!

Similarly, there could be the case if a user has 100 Sun caps in his inventory before the start of the rainy season, and it unlikely they will sell those in the upcoming season, so that’s an Excess Stock situation.

Now that we are clear about what to do, the next question is how to do it? This can be achieved either by Prescriptive Analytics or Predictive Analytics. The difference between predictive analytics and prescriptive analytics is the result or outcome of the analysis. Predictive analytics provides you with the results which can be further used for making informed decisions, while prescriptive analytics provides you with data-backed decision options that you can weigh against one another. First, let’s take a look at the popular Predictive Models that are available; there are around five analytical models that are popular for Predictive Analysis.

1. Classification Model

The classification model is the simplest one out of the lot, and this model basically categorizes data into two different buckets based on what it learns from the historical data. The classification model is generally used where you want an answer in Yes or No, providing broad analysis to take the decisive action in any business.

Some of the examples could be -

In the retailer world -Will this Customer churn?

Banking sector -Is this transaction fraudulent?

So, as the classification model has the breadth, and with ease, it can be trained for new data sets and used in a vast majority of the industries.

2. Clustering Model

As the name suggests, clustering models divide data into different clusters of separate, nested smart groups based on similar attributes.

Sample use cases for this predictive analytics technique could be, In a banking and finance industry, if a company is looking to implement targeted marketing campaigns for their customers, they can go through the individual records (which can be 100s or 1000s of customers) and come up with a tailored strategy for individual customers. But this will not be an efficient use of their time. Instead, they can use a clustering model to separate customers into small buckets depending on the similar attributes and then devise a marketing strategy for these individual groups at larger scales.

3. Forecasting Model

The forecasting model is one of the most widely used predictive analytics models, this model does metric value prediction, providing new estimated numeric value for new data based on learnings from historical data. Forecast models can be used wherever historical numeric data is available. Some of the use cases could be -

• A company providing software services can estimate how many customers they are likely to convert within a given week.
• A Mall can predict the footfall of customers per hour.
• Retail Industry can calculate how much inventory they should keep on hand in order to meet demand during a particular sales period.

The forecast model considers multiple input parameters. If a restaurant owner wants to predict the number of customers, he is likely to receive in the following month. The model will take into account factors that could have an impact on this. For example — Is there an event close by? What is the weather forecast? Is there a pandemic going around?

4. Outliers Model

The outliers model is oriented around data entries within a dataset that are deviating from what is standard, normal, or expected. The model can identify deviating figures either by themselves or in conjunction with other numbers and categories. Some use cases could be -

• In a call center spike in support calls, could indicate a product failure that might lead to a recall
• In banking transactions finding anomalous data, or in insurance claims, to identify fraud.

The outlier model can particularly be useful for predictive analytics in the retail or finance sectors. For example, when identifying fraudulent transactions, the model can assess not only the amount but also location, time, purchase history and the nature of the purchase, e.g., an $800 purchase on electronics is not as likely to be fraudulent as a purchase of the same amount on books store.

5. Time Series Model

The Time series model is used to predict future values based on previously observed values. It consists of a sequence of data points captured, using time as the input parameter. The model uses last year’s data to develop a numerical metric and predicts the next three to six weeks of data using that metric.

Use cases for this model include the number of daily calls received in the past three months for a Call center, Sales for the past ten quarters, or the number of patients who showed up at a given hospital in the past eight weeks. It is a powerful means of understanding the way a singular metric is developing over time with a level of accuracy beyond simple averages. The model also can take into account other attributes like seasons of the year or events that could impact the metric.

If the owner of a retail mall wants to predict how many people are likely to visit his mall, he might turn to the orthodox way of averaging the total number of visitors over the past say 120 days. But as we know, growth is not always static or linear, and the time series model can better model exponential growth and better align the model to a company’s trend. Time series can also forecast for multiple projects or multiple regions at the same time instead of just one at a time.

Conclusion

For the use case we discussed, we will try both Forecasting & Time Series models, and depending on the results, we will choose one. In the next post, we will talk more about prescriptive analytics.