Why Retailers Are Flocking To Predictive Analytics

It is impossible to miss all of the media attention that “big data” is receiving lately – the Cambridge Analytica controversy, GDPR and the countless brands who have been fined by regulators for breaching data regulations all illustrate the growing importance of data, its uses and applications. For some time now, retailers, too have been reaping rewards from data-driven strategies and are increasingly devising their marketing strategies to include insights gained from large datasets. Predictive analytics is just one part of this drive, and this article will focus on what predictive analytics is and how it can be used to give retailers a competitive edge in their day-to-day operations as well as in their marketing strategies.

 

 

 

Predictive analytics is the process of gaining insights into future events, using data. Data is used to examine historical trends, patterns and dispositions in order to be better placed to predict what events are more likely to occur in the future. A simple example might be in terms of a retailer selling goods. A customer who has made 3 prior purchases can be seen as more likely to make further purchases than one who has never made any purchases at all. As such, a retailer might decide that the customer with the history of repeat purchases is more valuable to them because they are more likely to make a further purchase. As a result that “more valuable” customer might be targeted with more marketing communications. In The Art of Statistics, renouned statistician Sir David Spiegelhalter argues that analysis of statistics could have been used to identify serial killer Harold Shipman years before he was eventually caught, due to the disproportionate numbers of deaths linked to his GP practice. In the 1990s, Spiegelhalter was hired by the UK government to look at patterns related to outcomes in critical heart care and make deductions about how competent different hospitals and staff were. Furthermore, banks use predictive analytics to fight cyber crime and identify suspicious transactions and account holders. UK banking compliance rules require banks to analyse certain patterns in financial transactions, like repeated transfers of large sums of cash, in order to identify and prevent money laundering and fraud. Another example of predictive analytics at work is the idea of credit scoring, which seeks to predict how likely people will be to make repayments reliably under future credit agreements.

 

Predictive analytics encompasses machine learning, data modeling, AI and data mining amongst other techniques, which are aimed at making accurate predictions about future events, based on data. Predictive analytics is different from forecasting because it “learns” from past experience, and typically uses a wider variety of data points and values comparing to forecasting. For example predictive analytics can analyse data from external sources, in addition to data from internal sources to make valid predictions. One retailer used predictive analytics encompassing weather reports and past customer behaviour to target a sales campaign for iced tea, using the rationale that when the weather was hot, more people were likely to be interested in making iced tea purchases.

 

 

 

Starbucks has been using data to decide what new products to create, based on data from their own baristas about how customers ask for their coffee to be prepared. The brand commissioned several studies into at-home tea and coffee consumption, and combined this with information gathered from their own in-store baristas. Information and trends gleaned from this research, for example the fact that 25% of customers surveyed never added milk to their iced coffees when prepared at home, were used to create new products for sale in Starbucks. As a direct result of this research, new products – Mango Green Iced Tea, Peachy Black Tea and blacked iced coffee were created. Additionally, Starbucks has been using this research to decide how to customise their drinks and what customisations to offer along with their products, in order to maximise sales. 

 

Starbucks has also used data analysis to identify and tap so-called “mega trends”, with their “pumpkin spice” flavouring providing a good example. Starbucks began to notice how popular its pumpkin spice latte was back in 2003. Since then it has added the flavouring to a wide variety of different “pumpkin spice” products, driving additional footfall through its chains, for every new product added. The current pumpkin spice range incorporates pumpkin spice lattes, instant pumpkin spice latte packets, iced pumpkin spice espresso and pumpkin spice frappuccinos, and the breath of the pumpkin spice range has led some commentators to label pumpkin spice as the Starbucks “golden goose”.

 

There are lots of ways in which data can be used to maximise current and identify new business opportunities. Furthermore, both internal and external datasets are valuable in identifying opportunities, as they arise. A toy manufacturer for example might notice that a particular toy is selling particularly well. That retailer might then decide to promote that particular toy very intensively, as well as look to promote toys that are similar because they might also sell well based on the similarity. Otherwise, a business selling a particular product or service might decide, based on data to create a product or service similar to one that’s already selling very well.

 

 

 

Retailers are flocking to predictive data because it can be used to predict future customer behaviour, which in business terms, means it can be used to predict which customers are more likely to purchases cross-sells or up-sells, preferred purchase channels, which customers are more likely to make repeat purchases, which customers are more likely to be influenced by particular campaigns for example “green” or environmental causes, and which customers are more likely to spend large sums of money and when. Understanding each of these issues allows businesses to create more cost-effective marketing campaigns that are more likely to sell what is being advertised.

 

 

 

Data analytics also allows for more personalised marketing campaigns to be created to target certain customers in certain ways. It allows businesses to identify and target what are referred to as “high value customers”. Customers can be “high value” for a number of reasons, but typical examples of high value customers are those who remain loyal, those who make high value purchases and those who make repeat purchases.

 

Personalisation based on data analysis can also assist businesses is terms of what product information they provide, and how. A product description can be built based on what information is most frequently sought online by customers shopping and looking for new products, and as such the product can be described more helpfully and in a way that makes it of optimal appeal to prospective customers.

 

Target, for example uses records of customers’ past purchasing decisions to decide what promotional email campaigns to run for particular customers. It has a “pregnancy prediction formula” which helps identify customers who may require baby products in the near future. This has helped it maximise sales in relevant product segments.

 

 

 

Some retailers with their own distribution networks are turning to analytics to help them predict shipping problems. Data can be gathered relating to the age, and functioning of a fleet of vehicles and this can help retailers prevent logistical problems before they occur.

 

 

 

Many make the mistake of thinking that predictive data used in retail is all about marketing. In fact, many retailers are now using predictive analytics to operate their supply chains optimally. Take the seemingly simple problem of stock. When retailers run out of stock, this represents a major problem because it means that customers will go elsewhere to make their purchase. Moreover, by running out of stock, retailers risk creating a negative impression in the eyes of the customer whose time has been wasted by looking for a product, only to find it isn’t in stock. According to Which? research, which considered what factors customers consider when selecting where to shop, convenience is high on the list of what can attract and what can repel customers. For example Homebase, with its high incidence of “out of stock” items was named “worst online retailer” of 2018. The research suggested that customers were frustrated by Homebase’s slow restocking and failure to match price reductions that were available elsewhere on certain products.

 

Furthermore, by using predictive analytics retailers can predict ebs and flows in their sales of popular items, and restock accordingly, making it less likely that they will run out of stock of popular items, thus avoiding losses involved and the damage to their brand reputation that occurs when retailers are out of stock of particular products.

 

Predictive data can also help retailers plan around pricing, for example it can be used to identify when products sold are most in demand and when certain products sell more frequently. This can be used to make better profit from particular products, or to reduce prices to sell a larger quantity, making an overall profit based on sales volume.

 

Deep insights into customer behaviour can assist retailers in taking critical decisions, faster. Decisions to stop selling a particular product, to reduce price, or increase the price of a given product can all translate into profits or savings, depending on the situation. Take the example of a doll that isn’t selling as well as expected. Predictive data can alert the retailer to this and the retailer can take remedial action by not allocating as much space in their stores for the unpopular product, and not allowing the unpopular product to take up valuable space in their nearest warehouses, making way for the retailer to use that space more effectively by replenishing with products that are selling better.

 

Additionally, predictive analytics can be used to help retailers decide how to and in what ways to expand their product itineraries. Having the right products on sale in the first place places retailers at a significant advantage, over non-data driven counterparts, who have to use their instinct and “best guess” about future demand and to decide what products to focus on.

 

 

 

Decisions about where and when to open a new branch of an existing business, or any new business for that matter can be greatly assisted by predictive analytics, which can be used to identify and understand customer demand and how it may grow. This information can then be used to build a business that caters to demand, thereby reducing the risk that a new premises won’t succeed or grow as expected.

 

 

 

Predictive data can be used by retailers to detect and prevent fraud and theft. Theft and fraud will almost always leave a “data footprint”, and this information can be used to identify when stock is being depleted as a result of fraud and theft. Analysis of data can save brands considerable sums of money by detecting these patterns sooner rather than later, thus enabling retailers to take remedial and preventative measures more quickly.

 

 

 

E-commerce advertising can be very costly, so retailers want to be sure that they get the best return for their investment in e-commerce strategy. Predictive data can be used to achieve this. By understanding customer behaviour, retailers can predict when customers are most likely to visit their site and, marketing and promotional strategy can be adjusted accordingly.

 

 

 

Very deep analysis of how customers interact with apps and websites can be vital to a retailer, and the correct actions based on analysis of this data can be pivotal to the success of a business. Businesses can analyse “click-through” rates to devine what promotions are most popular and when. Analysis of drop-off rates and basket abandonment can be used to set pricing optimally and understand what the highest possible price customers are willing to pay for given products at given times, thereby ensuring that retailers get the best return possible for their product offering.

 

Customer engagement with a website can be used to create more accurate product recommendations, and as such tailor a marketing campaign to the individual preferences and interests of customers, in real time. So, if a customer goes shopping on a website and places a number of items into their basket, only to abandon their purchase before they checkout, this information can be used to target that individual customer with adverts relating to the products they were the most interested in. Shoppers will often fill an online shopping basket with items they are interested in buying at a future date, so the right basket abandonment strategy, informed by predictive analytics can prove critical in maximising sales.

 

The value of using data this way is illustrated by the case of Macy’s, who implemented a predictive analytics solution that proactively targeted registered users of its website. The solution analysed browsing patterns and followed up with targeted advertising. Just 3 months later Macy’s reported that their online conversions had increased by 8-12%.

 

 

 

Data isn’t always a panacea for business. In some cases data storage can be hugely costly and reap little or no benefits. Data doesn’t have a value in and of itself. It needs to be manipulated and interpreted. Moreover, data storage attracts considerable responsibility and cost, and it can cause businesses to be targeted by cyber criminals and scammers. In almost every jurisdiction nowadays, the storage of data is governed by government regulation. If this is ignored or not applied appropriately, companies can be fined considerable sums of money. This makes it necessary for every business handling data to hire staff to safeguard their data and ensure it doesn’t fall into the wrong hands, get lost or misused.

 

Often, to gain real value and insights from data, a retailer will have to hire trained data professionals. A cheaper alternative is to use software that interprets data. Furthermore, implementation of a data driven strategy can be time consuming, but this problem can be alleviated by using software that is designed to make data easier to understand.

 

 

 

There are so many reasons why businesses are becoming more data-driven, it is impossible to list them all. However predictive analytics is something that few successful retailers overlook. Data can be used to inform smart decision making about what to sell, where, when and to who. It can also be used to inform smart stock and itinerary planning and to ensure that decisions about opening new premises are fruitful. Data can be used to ensure that customers’ needs are identified early, and catered for to a high standard. Additionally, data can be used remedially for example it can be used to prevent fraud and theft through early detection. Some retailers are even using data to ensure delivery fleets are operating optimally with as little downtime as possible. There are few disadvantages to a data driven strategy, but retailers need to be aware that handling data irresponsibly can result in fines for data regulation breaches. As such, any data strategy should be carefully evaluated before it is implemented.