Laurie Schaefer, advisor to SAS on interaction with retailers: "Analytics is meaningless only for the sake of analytics"
Big data analytics in retail is one of the hottest topics today and in the near future. Retail chains collect huge amounts of data about customer behavior, demand changes, and each transaction. But what really works and what can help predict the future? We are talking to Laurie Schaefer, SAS' retail advisor, about how the retail analyst is evolving.
– What's the place of the analyst in the retailer's office? What tasks help to solve analytical data?
– Retail is becoming more and more omnipresent. In North America, there are not many offline retailers left who never dared to try the online sales channel. That's why there aren't many of them left - most of them either went online or went out of business.
Based on our experience, today there are several main tasks in which the retailer needs analytical data.
The first group of tasks is solved with the help of commodity analytics, sales analytics:
- Optimization of assortment
- Price optimization
- Demand forecast
- Optimization of the dimensional grid (highly relevant for Fashion retailers)
- Warehouse optimization, distribution and replenishment
Next, we highlight marketing analytics that help us to understand customer behavior. Whereas we used to talk about the behavior of online users, now, thanks to the Internet of Things, we can study the behavior of customers in offline stores. Retailers place a variety of sensors in the retail space, analyze the flow of customers, the interest of visitors to specific areas of the store. And all this happens in real time.
We also have an analyst who builds around optimizing marketing campaigns and client base analysis. Here we solve the problem of studying consumer behavior all along the way: after all, the buyer can go to the site, then come to the offline store, read the newsletter or study the mobile application and then return to the online store.
A major challenge for analysts is to predict demand and optimize the supply chain. Again, thanks to sensors such as RFID tags, a retailer can know in real time how many items are currently in each point in the supply chain.
Well, of course, analytical solutions are actively used in e-commerce. Thanks to the available information about customer preferences, we are able to perfectly personalize the content of the website, the content of individual pages, individually for each visitor, based on an understanding of their behavior, knowledge of their interests.
There is one more thing - analytical solutions for top management. They include mechanisms of machine learning, which are able to answer specific questions about the state of sales, give financial forecasts for all indicators.
Every process - be it marketing, merchandising, supply chain and warehouse management, online trading or management reports - is based on data analysis.
And finally, because most retailers are switching to an omnikal model, they need to look at demand forecasts in a new way, combining all sales channels, rather than looking at the results separately as before.
– Out of all this variety of solutions, what is the most popular among retailers? What is the most common problem to be addressed?
– It depends on the type of retailer. In the food segment, the most frequent need is to improve demand forecasting. This has a noticeable impact on the profitability of the retailer. Another issue for food retailers is the optimization of prices and blotting campaigns. The seller needs to know that the stock of promotional goods will be sufficient for everyone. And also not too much to reduce prices, so as not to frighten those who are willing to buy more expensive, and not to spoil those who benefit from discounts.
For Fashion-segment the most common task is to optimize the assortment. And, both in the online store and offline. Fashion-retailers optimize not only the availability of goods in a particular store, but also the size range, laying out more than the sizes that a particular store more often buy.
– Do Russian retailers have any specifics?
– In general, the problems around the world are about the same. But Russian retail really has its own specifics - some chains in Russia are very large, with a huge number of stores and tens of thousands of SKU in each of them. Because of such sizes of a retailer, forecasting of demand in its network requires much more complex, large-scale solutions, more complex analytics. Here we need scalable technologies, and it is for such cases that we develop high-performance systems for working with Big Data. The data in them is stored in high-performance storages, for example, on the basis of Hadoop, and it allows systems to work quickly. It is modern technology that allows us to create flexible solutions that scale for each retailer and help us to perform accurate analytics.
– Are there any new approaches or is the analytics still working on the algorithms created in the 60s?
– The algorithms themselves can be the same. And in many ways they are the same. The speed, quality of analytics, i.e. its accuracy, and availability of analytics for business users have changed. Modern technologies allow models to work faster and on much larger volumes of data. Of course, there are also new solutions. Now, when retailers receive data from many new sources - the Internet, shops, social media - we need new algorithms that allow us to isolate the necessary information about what is happening at each stage of the buying process: whether the demand from the buyer has changed, whether his behavior has changed.
- Here's a good example. Take the same Amazon. Prices change every second. In addition to the fact that suppliers change the prices of goods, it is also important to keep track of price changes from competitors. You need an algorithm that calculates the ideal price. And these algorithms must work in real time.
We're always looking for these new algorithms. We always find new ways to use mathematics.
– Retailers are putting a lot of effort into collecting data. But what does it really take? Maybe we collect a lot of information, but we don't use it all?
– Retailers learn every day. They have so much data! This is exactly what our algorithms do - from all the information they receive, they allocate exactly the data that is relevant to the retailer. Depending on what kind of business problem you're trying to solve, you only need a small piece of all this data, but the algorithms have to process the entire amount of data to highlight the right one. Statistically, 98% of the data collected is not used.
– Are the costs of collecting this data commensurate with the result?
– Yes and no. In cases that we have perfected, such as price optimization, demand forecasting, customer behavior analysis, we are very good at pulling the right data out of huge data streams. But there are quite new areas, such as social media analysis, text analytics, video analytics and voice analytics. The whole market is still studying here, and of course the costs are not commensurate with the result.
– What awaits the analyst in the near future? Perhaps there will be some breakthrough, will there be completely new technologies?
– There are always breakthroughs, I think. For example, IoT or real-time analytics, when data is collected from specific devices and processed in streaming mode. These trends will increase and deepen. But with the current dynamics, of course, we would like to see new breakthroughs.
Artificial Intelligence, for example, or machine learning, is a very big breakthrough, a revolution in the marketplace. Although for us it is a well-developed area: SAS has been working with machine learning technologies for many years. But new technologies, combined with new approaches to implementing user interfaces such as Amazon Alexa, allow us to filter large amounts of data very quickly. When a person talks, the system quickly understands what he wants to buy, for example. Retail has a great need for artificial intelligence technology. People don't want to see endless reports, and they don't want to deal with the results. They need to get answers to specific questions quickly. Imagine asking "Alexa, what's our current sales situation?" or "What's the sales forecast for this category tomorrow? Instead of looking at bundles of paper and trying to interpret the reports yourself.
In any case, the analyst should serve a specific purpose, whether it be profit growth, sales increase, improvement of customer service. If this is not the case, then analytics is meaningless for the sake of analytics.
Maria Sysoikina talked to Laurie Schaefer, SAS' advisor on interaction with retailers.
The original news is posted on the websitе: new-retail.ru