Menu
Menu
How pizza businesses are using predictive analytics to optimise promotions

How pizza businesses are using predictive analytics to optimise promotions

Two pizza outlets are analysing data to predict which promotions will provide the highest yield

Many retailers analyse the results of their promotions, but what if they could predict which ones will produce the best response before spending time and money creating and sending them out?

That's a question Mohamad Khatib, senior project manager at Nielsen, was able to answer for pizza businesses, which often get a good chuck of their sales from promotions.

Nielsen is a global information and measurement company, which conducts analytics in the consumer sector. Khatib said pizza is a US$40 billion business in the United States, with US$4 billion a year spent on promotions.

“As you can imagine, pizza providers are very interested in optimising that spend to increase their benefit,” he said at Predictive Analytics World in San Francisco.

It is calculated that there are on average 350 slices of pizza being made per second in the US, with the bulk of sales coming in during weekends and Halloween, according to Khatib.

Many pizza businesses still use traditional mediums for promotions such as postcards, door hangers and magazine advertisements, Khatib said. As these businesses are not large corporates with lots of play money, knowing which promotions are best to invest in helps ensure money and effort do not go to waste.

“When business managers see the results they know how much additional resources they need to plan for the promotion, how much additional inventory they need to have in stock so they can anticipate for business and the operational issues for that,” Khatib said.

Khatib worked with two pizza stores in the US, with similar market share and operational challenges, but different demographics.

Historical sales data from 2011 was gathered to first map the responses to promotions, which can be traced through coupon numbers. He then assessed the response rate to promotions over an eight-week period, and assigned a response rate to each type of promotion.

He then came up with a simplified predictive model, which integrated demographics analysis. After using the model to make predictions, he compared the predictions to actual sales results.

“Obviously there is room for enhancing the model for each store ... but the results came very close to the predictions,” Khatib said.

The insights that were found in the model was that store A had overall better response rate to postcards than store B. Both stores has high response rates for door hangers, with store A having faster response rates to this than store B. For magazine ads, store B had good response rates, compared to store A, which did not do well with this type of promotion.

Having these forecast means these businesses can know where more or less money and resources should be allocated to maximise sales and revenue, Khatib said.

Visualising the results against the predictions is also important, not just using the model without communicating to the business on how it is contributing to sales, Khatib added.

“We work with business managers who are not necessarily technology savvy, but interested to support the project. And in order to enable us to succeed in working with them, visualisation plays a key role in doing that. It helps us to empower the business managers when they make a decision, they like to see it.”

Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.

Join the CIO newsletter!

Error: Please check your email address.

Tags analyticspredictive analyticspredictive modellingpizza analytics

More about Nielsen

Show Comments