Conversion optimization: 7 recommendations for using predictive analytics

Predictive analytics is a technology that relies on Big Data, data on people's behavior, to predict how they will behave in the future, and to optimize business processes using this knowledge. Have you ever wanted to know in advance about what products your customers will most likely buy? It would be great if you could predict the maximum price the customer is willing to pay for the product. But what if you could optimize the client service and solve all the problems before the user had them? Most likely, this knowledge would help you increase your profits in the field of e-Commerce and increase conversion.

Predictive analytics offers solutions not only in the above areas, but also in many others. Below are 7 tips for optimizing conversion using predictive analytics methods.

1. Increase customer interest and increase revenue

There are different types of customers, and their perception of e-commerce segment sites differs significantly from each other, each needs its own approach, each can be attracted in some particular way. Predictive analytics considers all possible options for perception in order to arouse the desired interest of each customer. This may be an offer to subscribe to the newsletter, click on the "share" button or some other way to attract customers.
')
There are several products that help retailers create models for tracking and analyzing user behavior - for example, Alteryx , Attivio , Lattice , SAS . Such models can later serve as benchmarks in business, helping to develop in the right direction.

Lattice explored how leading companies Amazon and Netflix used predictive analytics to better understand user behavior and develop a solution that will help sales professionals to more accurately determine leads.



Private capitalists invested more than $ 160 million in 2014 in forecasting tools that help marketers understand how best to make online and offline sales.

Investors in this area understand market opportunities - they see that Lattice type forecasting tools help sales managers better calculate leads using publicly available information about them than simply comparing them with an existing base of company customers. Below is an illustration of the amount of investment in various forecast analytics programs.


A source

Forrester calls the combination of predictive analytics and customer engagement as prognostic analysis applications, considering them to be part of an era in which the entire Internet segment approaches the extremes expected for it: the world of hyper-individualized experience. It was thanks to such tools that Dell managed to almost double the sales figures, although the number of leads that the marketing department sent to the sales department was reduced by 50%, using predictive analytics, only the most promising ones were selected.

In this example, PredictiveAnalyticsWorld.com uses a predictive advertising system in an unnamed educational portal that is used by every third high school student, which helps to better match the promotional offers to the existing traffic. As a result, user response increased by 25%, which is equal to advertising revenue of about $ 1 million in 19 months.

2. Launch advertising campaigns targeting your customers.


An example of a personalized advertising company

Advertising in retail is a must-have, but it’s not so easy to direct your advertising campaign in the right direction.

According to an Oracle study, 98% of fast-growing trading companies understand that segmentation and targeting are the most important part of their online merchandising strategy, but more than half of them are not satisfied with the tools they use for advertising campaigns.


a source

Predictive analytics can change this by comparing data from multiple sources, which helps to create personalized promotional offers that will be as effective as possible for a particular customer or segment.

Macy's have already managed to convince themselves of the advantages of predictive analytics, using the solution proposed by SAP, which resulted in more effective targeting of users already registered on the site. For three months, Macy's observed an 8-12% increase in online sales, which was achieved by analyzing users' transitions by product category and sending targeted emails to each segment of potential customers.

StitchFix is another retailer using a unique sales model. Users are invited to take a survey about the style of clothing, and then with the help of predictive analytics it is determined which clothes will be more pleasant for each of the clients. If the buyer does not like the clothes, he can return them back without paying for the delivery.


a source

Another example: Turkcell , the largest mobile operator in Turkey, uses more than 150 parameters characterizing its users, such as usage patterns, device preferences, location data. All this information is used to send customers the most appropriate advertising offers for them in real time and thereby reduce the outflow of users.

It is important that you understand that predictive analytics tools do not work on the principle of Plug & Play: you upload data and get instant revenue growth. According to a study conducted by Ventana, only 13% of the 2,600 enterprises consider forward-looking analytics to be an essential element of their business strategy.


A source

David Menninger, former director of research at Ventana, argues this as follows:

“Predictive analytics remains a tool for professionals. Personally, it seems to me that it is very difficult to do this, and the complexity of calculations in this area is beyond the capabilities and knowledge of most people . ”

Robert T. Mitchell found out in the process of interviews with expert analysts of the installation firms that the most common mistakes in the implementation of predictive analytics programs are related to the fact that entrepreneurs cannot form the final goals of their activities, do not delete outdated data (due to lack of understanding) and the spirit of conservatism and rejection of change reigns in their company.

For example, Dean Abbott from Abbot Analytics shared a story about a collection agency that wanted to develop the most effective sequence of actions for debt collection, but followed strict rules, due to which collectors had to perform the same actions each time.

“Data analysis is the art of making comparisons, and for this you need examples from history. Fortunately, most experts are of the opinion that, although even defective forecasting models are rarely fatal and can always be improved, the general opinion is that building quality models requires a lot of effort and can take a lot of time. For the client, this means that he is wasting money and time and does not receive instant returns, or worse, he is wasting resources altogether. John Elder says it can take a year to bring the prediction model to mind, and for this very reason, even if technically 90% of the models created for clients are successful, only 65% ​​of them actually work . ”

3. Optimize prices to maximize profit.


An example of testing different price tags

Traditionally, retailers have used A / B or Multivariate testing to set prices for different product categories and determine the optimal cost that will maximize profits. The problem is that each price is set manually and very much depends on the human factor, which means that the probability of error is high.

Predictive analytics uses a different approach to developing a real-time pricing model, which is formed on the basis of information from such sources as:
• Historically accepted price of the product;
• customer activity;
• order history, customer preferences in the past;
• prices for similar products from competitors;
• desired markup;
• available stock of the product;
• other.

This video shows how Uber & AirBnB got rid of the difficulties associated with setting prices for various categories of goods by equalizing supply and demand, and how they were helped by predictive analytics.



You need to constantly monitor the pricing process to avoid automatic price changes that will cause questions among retailers.
The benefits of predictive analytics for pricing management have long been underlined by Accenture. Their report says that it is never too early for a retailer to start experimenting with pricing using predictive analytics. The sooner a company starts doing this, the sooner it can achieve successful analytical forecasts. This is hardly somehow connected, but after the publication of the report in 2011, the demand for the work of analysts increased significantly.



4. Inventory management: replenish them on time, but avoid oversupply

Walmart made a revolution in inventory management by asking suppliers to provide real-time support in this area, the system was named VMI (vendor managed inventory - supplier-managed inventory).

Predictive analytics improves this solution by reducing the required / critical stock level of the product, if the forecast model does not foresee large orders for the near future. This helps retailers distribute their funds so that they can buy products that are in high demand and potentially more profitable.

Last year, a useful study was published on how big data (Big Data) and predictive analytics are able to change the strategy of inventory management.

Researchers from Sam M. Walton College of Business and Weber State University have determined that the lack of technical skills to use such technologies is the biggest obstacle to a wider implementation of predictive analytics analysis, but this situation is changing with the help of suppliers who offer easy-to-implement integrated solutions for inventory management using predictive analytics.

The graph below shows how often in recent years Google has begun to search Google for the phrase “predictive analytics” - this indicates that the gaps in knowledge are decreasing every day.


a source

One of these examples was the Southern States - it tells how agricultural cooperatives used Alteryx to maintain sales at the same level, while storing 31% of the goods in warehouses.



For farmers, this meant 31% less spoiled products, but for a business owner in the e-Commerce segment, this means less money spent on storage, fewer goods at customs and other unnecessary spending.

5. Reduce fraud risk by actively identifying it.

Unfortunately, fraud is a frequent story in modern retail, including online, annual losses from it are calculated in billions of dollars.
Any technology that can reduce losses from fraud is like a breath of fresh air for any retailer. Solutions offered by predictive analytics, like those found in the IBM's SPSS suite , allow an entrepreneur to analyze user behavior patterns, payment methods, and product purchases in order to detect and prevent possible fraud. Some retailers even experiment using self-learning predictive analytics programs to automatically identify patterns that can be used to detect and prevent fraud.

This is really necessary, because fraudsters are becoming more and more inventive every day.

In the work carried out by Aberdeen, various types of fraud were analyzed, as well as readiness to deal with them. Below is a graph showing the degree of readiness to combat various types of fraud, in percentage vertical - the degree of readiness of retailers to protect against each type of fraud, horizontally - the prevalence of this type over the past year.


A source

The report also stresses that predictive analytics as a method of dealing with fraudsters still have a long time to develop: only 16% of respondents said that they used analytical tools primarily to detect fraud.



Walmart showed a serious attitude towards predictive analytics as a way to combat detection of fraud when they acquired a promising start-up in the field of predictive analytics Inkiru last year.

Other retailers have also been using various algorithms that calculate fraudsters over the past few years. But now such models are expanding and improving, starting to use predictive analytics to recognize and prevent fraud before it happens.

6. Offer better customer service for less

Customer service is one of those areas with which retailers have a lot of questions, here are some of them:

• Should there be customer support only in electronic form or is a call center also needed?
• If a telephone support service is also needed, how many managers are needed?
• Maybe, besides a call center, you need an online consultant on the site?
• What should be the optimal time to wait for a response when a customer calls support?
• How to prioritize loyal and valuable customer issues?

Find answers to these and other questions by building a model, a unique customer service of each particular retailer. Over time, this model will be improved and will provide the most accurate forecasts that help to improve the quality of customer service.

Linux distributor Red Hat uses predictive analytics to improve customer service by increasing the parameter they call “stickiness to the subscriber.” According to the study, the company managed to anticipate user questions and solve their problems even before they appeared.

Hotel chains, such as Marriott, are another great example of a business that pays a lot of attention to predictive analytics, which helps exceed customer expectations before, during and after a hotel stay.

Premium segment hotel chains, such as the Four Seasons and Ritz Carlton, always try to anticipate the desires of customers - and all this happens thanks to predictive analytics! This is just one example of when the Ritz Carlton got out of hand to exceed customer expectations using a bit of predictive analytics. The son of one of the hotel guests forgot his favorite toy there - the giraffe Josie, the father came up with a funny story that the giraffe stayed at the hotel and had a sun bath. On the same day, the hotel staff called saying that they had found a giraffe, and when their father asked them to take one funny photo of a giraffe in a lounge chair to demonstrate the truth of their invention, the hotel staff made a real vacation photo session for Josie, a couple of photos below.



7. Analyze information and make decisions in real time

Streaming analytics is the ability to generate ideas in real time, which helps retailers to make decisions "here and now."

The retail segment is developing very quickly, so it is meaningless to use predictive analytics, relying on outdated data. Decisions made in real time help to choose the most successful day for launching an advertising campaign, determine the products that will be sold best, popular products that will sell well, correctly target specific campaigns, etc.

Netflix is ​​a well-known example of effective streaming analytics: they record and analyze each element of interaction with clients, including at what point the client lingered, how many times she did it, record, what movie color attracts more attention of clients, etc. . All this helps them to give useful recommendations in real time.

The technology platform Granify releases predictive analytics of the same level, using the accumulated data to adjust the website in real time. For example, if the behavior of a site visitor matches the behavior of someone from past customers who were worried about clothing sizes, his attention will be drawn to the size chart, and as soon as it becomes clear that the customer wants to know about the delivery, the system will immediately switch his attention to delivery terms in any way.



Source - http://conversionxl.com/predictive-analytics-changing-world-retail/?hvid=352IDw# .