Recommendation system product demand analysis guidance!

Based on his own practice and what he has learned, the author has disassembled the specific design process of the recommendation system from the perspective of product demand analysis. I hope it will be helpful to you.

Ten years ago, perhaps most of us had not heard of recommended products; ten years later, recommended products have spread to every corner of our lives.

When we want to buy clothes, we can often see products that we may be interested in on the homepage of the e-commerce website; when we want to listen to music, we often open the [Daily Recommendations] in the music app; when we want to read the news, we also often open the [Recommendations] list in the news app.

The more we use it, the easier it is to find content that interests us, and thus, we will become more dependent on this product and spend more time on it.

The biggest difference between recommended products and other products is that recommended products are typical "small front-end, large back-end". Although they look like a very simple list, the back-end is actually a large system that is a bit "invisible and intangible".

So, how can we make a recommendation system that people like?

If one day, the boss suddenly sends a message to you: Xiao C, our platform has accumulated so much content now, can we make a personalized recommendation feed flow product?

When we receive the requirements, we may be confused at first, but if we continue to improve according to specific methods, it will become clear step by step.

For example, if we want to make a feed flow product for programmers, how should we start?

01 Requirements Collection

Purpose: To collect demands from all parties and preliminarily determine the scope of demand.

Analysis of the interests and needs of product-related parties:

Boss: We want to make a recommendation system. After the product is launched, what daily active users and retention rates can we achieve?
Content provider (user): When will my article be included in the feed? What is the recommended daily intake? What content providers want most is the sense of accomplishment and superiority brought by the huge number of recommendations, readings, likes, etc.
Content consumers (users): "I only care about Java-related stuff, so just recommend Java-related stuff", "I only care about front-end technology", "I don't care about articles for beginners, they are too simple", "Just give me the articles I want to read..."
Operations: What are the characteristics of our users? What kind of articles do users want to see... Can you leave a place for manual recommendation for me? The data effect should be wow~
Business: After brand content cooperation, how to display it in the product?
…

02 Demand Value——User Analysis

Goal: Use scenario analysis to solve user problems.

We often use the 5W2H method, for example:

1. WHO?

Who will use our product?

If they are programmers, what are their characteristics?

Programmers here are both content consumers and content providers.

2. Why?

Why do programmers use this product and what problems can it help them solve?

For content contributors, the realization of personal value brought by sharing can greatly satisfy their inner sense of honor and accomplishment.

For content consumers, for example, if it is for programmers, due to the rapid updates and iterations in the technology field, they hope to learn new technologies in a timely manner and understand the technical sharing experiences of their peers, so as to solve their learning and growth needs in a timely manner, break the bottleneck in the workplace, and bring themselves better future career development and more objective salary.

3. WHAT?

Just do what?

Define the product, such as making a personalized feed for programmers, so that they can happily share their insights and smoothly learn the content that interests them.

4. WHERE?

Do apps, PCs, mini programs, and h5 all provide recommendation lists? The factors to be considered are mainly:

User experience fluency;
The entrance is easy to find;
Easy for users to share;
Workload.

An analysis needs to be made based on the company’s specific circumstances. If the company already has an app, has a user base, and meets the app’s positioning, then adding the recommendation list to the app is a relatively simple operation and you can consider adding it to the app.

5. WHEN?

So when will users use this product?

If it is a product for programmers, you can learn about programmers’ favorite learning time through the following methods:

This can be done through questionnaires;
Colleagues who are programmers around me;
Active time in similar communities;
User interviews;
Analysis of historical behavioral data.

6. HOW?

The recommendation system first needs sufficient content. Regardless of whether the content is PGC or UGC, there needs to be a corresponding content creation incentive mechanism (this is not the focus of this article, so I will not go into details).
At the same time, the created content must be conveyed to the right users through a personalized recommendation system. So how do we deliver the right content to the right users?

7. HOW MUCH?

Regular users can access the content for free.

03 Demand Value——Market Analysis

Purpose: Analyze the market value of the product, how often users use it, and how large is the user base?

It mainly includes PEST analysis (external environment), SWOT analysis (product itself) and competitor analysis.

1. PEST Analysis

It is a macro-environment analysis model of an enterprise, which is the external environment of the enterprise and is generally not controlled by the enterprise. It is mainly for:

Find out the external factors that currently affect the company's business;
Identify external factors that may affect the company's business in the future and may change;
Exploit change (opportunity) or defend against it (threat) better than competitors.

P: refers to the impact that politics will have on corporate regulation, spending power and other activities.

It mainly includes: government policies, political stability, tax policies, labor laws and other laws and regulations.

Since 2018, the country has attached great importance to high technologies such as artificial intelligence and has issued a series of documents. We can analyze and predict the impact of these policies on this recommended product, whether it is positive or negative? For example, will it lead to an increase in user scale?

E: Economic factors may include: industrial structure, economic growth, disposable income, international exchange rates, inflation rates, tax levels, industry growth, unemployment rates, etc.

These factors can have a direct or indirect long-term impact on the business as it affects the spending power of users in the potential market and can change the demand/supply pattern in the market.
Here is a picture of the salary for school staff recruitment in an Internet company. Let's see if you want to change careers:

Although everyone is not optimistic about the national economic situation in 2019, according to specific statistics, the salary increase rate for programmers who changed jobs was the highest in the first half of the year, reaching about 14%.

S: Social factors refer to the influence of social ethics, culture, population changes, education and values on enterprises and markets.

Social factors may include: demographic factors, age distribution, health awareness, social mobility, consumer psychology, etc.

As society becomes more intelligent, programming will become a skill that everyone can master.

If you don’t believe it, just take a look at how popular programming is among children nowadays. So will the number of programmers increase?

Moreover, people are now more and more interested in lifelong learning. So will users’ learning frequency and learning cycle be improved?

T: The overall level of social technology and its changing trends, technological changes, and the impact of technological breakthroughs on enterprises.

The more new technologies emerge, the more beneficial it is for the platform for technology sharing and exchange, whether it is providing recommended products with good user experience or users' rush to new technologies.
At the same time, will the emergence of new technologies facilitate the rapid development of the company's products?

2. SWOT Analysis:

Analysis of the company's strengths, weaknesses, opportunities and threats.

Advantages:

The main consideration is what are our advantages in doing this?
From what the boss said above, "a lot of content has been accumulated", this is an obvious advantage of building a recommendation system.

Disadvantages:

What are the disadvantages?
Is the product launched later than competitors?
If you are making content products, then not having enough high-quality authors may be a huge disadvantage, and not having the key technology for the recommendation system will also be a huge disadvantage.

Chance:

Will the rise of new technologies such as artificial intelligence and blockchain drive the development of the entire technology field, including the increase in users' learning and communication needs, and will it drive the prosperity of learning technology communities in vertical fields?

threaten:

There are already many feed streams on the market, and many of them include personalized feed streams in the technical field. For example, Toutiao and Zhihu both have feed streams in the IT field. How can we form our own advantages and characteristics and keep users attached to our products?

3. Competitive analysis: Porter’s five forces

Mainly analyze the situation of competitors in the market.

Threats from existing competitors: Toutiao, Zhihu, and Jianshu are IT vertical content feeds, and Nuggets is an emerging IT learning community, which will consume users' learning time;
Alternatives: public accounts of technical experts, WeChat communication groups/QQ groups in a single technical field, etc.
Threat from potential competitors: infoQ etc.
Bargaining power of suppliers: The recommendation system requires content, so there is a steady stream of content providers. Content providers are divided into PGC and UGC. Where do these two groups of people come from and how much does it cost?
Bargaining power of buyers: Although the recommendation feed is free for programmers, if you want to monetize through advertising, you also need to consider the prices that technology vendors can offer.

By analyzing from the user and market perspectives, we can basically conclude whether this direction is worth pursuing, and ultimately output a list of business goals and user needs.

04 Demand Management

The first step is to break down the requirements and classify them:

Classification by source: commercial realization needs, product core functions, product experience, operation activities, data, product stability, and product security;
According to the KANO model: it includes excitement needs, expectation needs, basic functions, indifferent functions, and reverse functions.

The second step is scheduling: usually operating urgent and important four-quadrant tools, but in general:

You can give high priority to your boss's needs. After all, your boss will get more information, which will be more valuable for reference.
Secondly, give priority to ensuring the needs of users for core functions and basic functions, and avoid providing undifferentiated functions and reverse functions.
Business and operational needs can be evaluated and prioritized based on specific needs;

05 R&D and launch

Execution is as important as demand analysis. After all, “talk without practice is just empty talk” and “real knowledge comes from practice”.

Nonviolent communication helps us apply for resources faster and allows the team to work happily;
Data analysis and sensitivity to data can help us optimize our products as quickly as possible.

Due to time and space constraints, this article focuses on the demand analysis phase. If you are interested, please leave a message and follow me. I will continue to output some experience in the following articles: