At a time when the success or failure of online sales determines the performance of a company, effective technology in the marketing market is highly valued.
In this article, we will explain the outline and advantages and disadvantages of recommendation engines that are used in BtoC business on the Internet such as EC and are attracting attention.
[What is a recommendation engine? ]
● About the recommendation engine
The word “recommendation” means “recommend” or “recommend”. A recommendation engine is an IT technology that automatically recommends products and services to customers based on certain rules in
web marketing activities .
[Overview of recommendation engine mechanism and functions: filtering and recommendations]
●What is filtering?
As mentioned above, the recommendation engine is a function that automatically presents “recommendations” based on the customer’s needs. The presentation is presented based on certain rules (called algorithms) .
The ability to incorporate algorithms to make automatic suggestions to customers is called filtering . There are four
of them, so let’s look at them one by one.
●Types of filtering
・Rule-based *Analog “recommendations” that reflect the company’s sales policy and season
(example)
- Seasonal sales such as Christmas and Valentine’s Day
- PR for the start of sales of the expected new product with the company’s fortunes on the line
Appeal to users on the website based on the expectations and intentions of the seller .
It is humans who decide “what, when, and how they want to sell”, and we make recommendations based on their intentions. It is an old analog PR method.
In addition to the rule base, the following three filtering techniques make it a useful recommendation engine.
Let’s take a look at each of the remaining three non-rule-based filtering.
・Collaborative filtering
Collaborative filtering is an algorithm that makes recommendations based on the behavior history and purchase history of many customers visiting a website .
-Regarding the collaborative filtering method “Recommendation”
Collaborative filtering has two execution methods called recommendations .
- Item-based recommendation
In general, it refers to recommending items on the website that are highly related to purchase at the same time that the product is purchased.
Example: When I put a laptop in the “cart (shopping basket)”, an extended display that can be used for telework is displayed as “Did you forget to buy it?” and “Recommended”
- User-based recommendation
This is a method of recommending “products that other people who like this product are likely to be interested in” based on the behavior history and purchase history data of other users who have purchased the same product .
Example: When visiting a website, a product of a genre that I had not thought of at all was displayed as a recommendation, saying, “Other people have also purchased this kind of product.”
-Features of collaborative filtering that change depending on recommendations
The advantage of collaborative filtering by item-based recommendation is that it
recommends products that are closely related to the product that the visitor picked up, so it can be recommended with accurate content . As a demerit, it is difficult to obtain the effect of giving hints for novel ideas .
The advantage of collaborative filtering by user-based recommendation is that it can present “recommendations” that are not in the mind of the purchaser
based on the behavior history of similar users .
On the other hand, a newly launched website has the weakness of not being able to derive and present “recommendations” when the number of visitor behavior history data is small .
・Content-based filtering
Content-based filtering is filtering that recommends highly similar products based on product item attributes without relying on visitor history data .
While it is possible to present recommendations with a high degree of accuracy, there is a problem in that the content of the proposals is often similar to the product that the visitor picks up and does not feel novel .
Another point to consider is that it is necessary for the company to perform preliminary work to group the products by attributes in advance.
・Hybrid filtering
Hybrid filtering takes into consideration the strengths and weaknesses of collaborative filtering and content-based filtering, and combines the strengths of each to enhance its capabilities .
A typical example is Netflix’s recommendation engine, which uses hybrid filtering. It is expected that this method will spread throughout the world by increasing its accuracy.
[Advantages/disadvantages of the recommendation engine]
Advantages of the recommendation engine
・It will be an opportunity to improve LTV
By having the recommendation engine propose attractive “recommendations” to users, customers will feel great value and satisfaction in CX (customer experience), just like in actual customer service.
As a result, they will become fans of your company and produce the following effects.
- Expansion through cross-selling and up-selling
- Increased purchase frequency
- Increase in repeat customers
Disadvantages of recommendation engines
・Problem with recommendation data accuracy when customer data is small
As mentioned above, I explained the merits and demerits of the recommendation method in collaborative filtering.
- When there is little data on the website at the beginning of the launch
- If there is no clear conclusion as to which filtering is appropriate for your company
In such cases, there is a risk that the recommendation engine may not be able to demonstrate its merits immediately.
・When there are few products handled by the company
On a business website that does not offer a wide variety of products, the recommendations that can be presented are uniform and do not attract the interest of customers .
In the above case, it is necessary to carefully consider whether it is optimal to adopt a recommendation engine in the first place.
[Usage scene of recommendation engine]
●Marketing
・Improve opening rate and revisit rate of e-mail magazine
This is a method that utilizes the analysis results obtained by the recommendation engine not only during web access, but also for e-mail magazines and SMS transmission services that will be sent at a later date.
We will improve CX by regularly making user base recommendations, and expect to improve LTV by not missing regular replenishment of consumables from purchase history.
・Increased video selection rate from recommendations on video distribution sites
There are many video distribution services, but many of them display “Popular works ranking by video genre” or “Those who watched XX also watched this work” on the menu screen. increase.
This “recommendation” is used to increase the time spent on video distribution sites and have a positive impact on the continuation rate of subscription contracts.
● EC site
・Cross-selling proposals based on product browsing history
Based on the user’s website browsing history, related products and data obtained from other users will be used to make recommendations such as “People who bought this product also viewed this product.”
・Recommendation from information in the cart (cross-selling while shopping, prevention of forgetting to buy)
This is a method that can be expected to cross-sell by recommending related products based on the behavior history of other users who have purchased the same product .
・Switch the product display according to the customer’s loyalty level
This is a method of selecting recommended content based on the relationship between the customer and the company.
For example, depending on whether the customer is a first-time visitor or a customer who has improved LTV and is positioned as a regular customer, we will switch the display of the website and present information that is easy to attract interest.
If you are a new visiting customer ,
you can introduce popular rankings or best-selling products .
・As a product selection support tool for businesses with an extremely large number of products
such as wine selection and outdoor products,
- There are many products and I can’t choose which one is best for me
- I would like a detailed guide for beginners
A recommendation engine is useful even in such a case.
Some customers find it difficult to teach from scratch, and some customers find it difficult to talk to staff members they meet for the first time. The recommendation engine can be said to be a tool that provides a very good CX (customer experience) as an entrance for heavy users .
[Points to note when introducing a recommendation engine]
●Recommendation engine contract format
・About ASP and open source
-ASP
ASP is a service that uses a cloud server that does not require the customer to prepare a server.
- No initial investment required
- You can receive consulting and support even if you do not have enough technical knowledge
There are advantages such as
-Open Source
Open source is a method in which a company prepares its own server and recommendation engine tools and builds a recommendation engine by itself.
The merit is
- Customized to incorporate your company’s needs
- Information leaks are less likely to occur
There is a point.
However, most of the recommendation engines are made overseas, and it is necessary to use English. Currently, the recommendation engine introduction market in Japan is still in a growth stage. Again, we recommend using ASP from a software vendor with expert support.
● What you need to prepare before introducing the recommendation engine
・Understanding of preparations before introduction
When introducing a recommendation engine from a software vendor, be sure to understand the preliminary preparations that you should do in your company.
- Confirmation of compatibility with the company’s website and related systems
- Do you have the optimal recommendation method and result reporting function for the goals set by KPI?
- To what extent will the support range and external partners support implementation?
Don’t forget to perform a preliminary check in order to simulate the actual operation as described above.
-Set quantitative targets
First of all, it is important to set KPIs by introducing a recommendation engine yourself, such as “what is the goal” and “what is the guideline for the result”.
What would you choose as your top priority for your company’s goals?
Whether you want to improve the LTV of existing customers or increase the contract rate of first-time customers,
setting specific goals and plans to achieve them is an important task.
– Utilization of output by recommendation engine
It is essential to measure the effectiveness of the recommendation engine after it is introduced. Therefore, each recommendation engine tool should have a “report function”.
It analyzes the current situation and improvement results numerically, so you can measure the results to see if the results match the initial policy.
[Effective use of the recommendation engine will significantly change your company’s marketing]
The recommendation engine makes it possible to enrich and improve the CX (customer experience) on web shopping and video distribution sites. As a result, it leads to improvement of LTV (lifetime value), and it can be said that it is a tool that can contribute to the improvement of corporate performance.
If you are considering a recommendation engine
- Have a clear “KPI setting by introducing a recommendation engine”
- Use the reporting function to meticulously measure effectiveness
- Make thorough preparations before installation and understand if there are any fatal problems before installation is important.
A recommendation engine that is likely to hold the key to DX conversion in your web marketing. Why don’t you check the whole thing?
