User segmentation and positioning for brand marketing!

I think digitalization is the right time, the right place, and the right people.

"Time" mainly refers to the organizational structure: because data naturally has barriers on the business side, each business team accumulates a lot of data during the operation process, but the business team needs organizational drive to obtain the data. Our marketing department at that time was affiliated with the e-commerce team, and my leader was the e-commerce director, so I had a natural advantage in control over numbers.

"Geographical advantage" mainly refers to the process: in the first five years, the two closed loops of operation and data are naturally separated. Generally, operations have their own exclusive closed loop, such as commodities, sales, forecasting, listing and final channel distribution, which constitute a complete business closed loop. On the data side, you can only see the final presentation results of the business. If you drive the business from the results perspective, there will inevitably be a disconnected state.

Therefore, in the last two years, the company made an adjustment in its processes. Previously, the business team usually came up with strategies, but as the amount of data increased, we hoped that the data team would come up with strategies. For example, when approving a budget, the data team decides whether to spend the money.

"Renhe" has high requirements for data teams: first, they must understand data and be sensitive to data; second, although they are not in the business circle, they must know how the business system runs and operates. Therefore, this aspect places extremely high demands on people.

When we first started brand marketing , our team did three things:

1. Collect and organize data, just like every team, and lay a good data foundation;
2. Data insight analysis, that is, how to gain insights from the collected data, how to empower the business and support business decision-making;
3. Data marketing application and the three-step brand marketing method focus on the third thing. Based on data and user behavior analysis, how to do brand marketing is easier to get started because everything is based on internal methods.

There were three main challenges at that time:

1. Data is scattered in different databases, which is not conducive to unified management, analysis and reuse;
2. The systems of the two service providers are both SaaS versions, that is, standardized products, and it is difficult to meet new functional requirements;
3. There are numerous online data reports, and each department spends a lot of time preparing them. There is a lot of manual work involved, which is inefficient and prone to errors.

Our consumer e-commerce touchpoints cover a wide range of e-commerce platforms on the market, such as Taobao, Pinduoduo, Youzan, etc. The corresponding data acquisition tools include JD Jiushu, Brand Data Bank, ERP, etc. Data storage will also be placed on Alibaba servers, JD servers, Jushi Tower, etc. Overall, the data is relatively scattered.

At the same time, looking at the data from a brand perspective, it is very inefficient when doing data utilization and efficiency analysis. The business department will give us many requirements, and almost all of these requirements need to be done manually, but the business has very high requirements for timeliness.

Low efficiency will directly lead to a question: how to improve the efficiency of team data analysis?

At that time, the company had order and behavior data of over 40 million consumers, but if this data did not generate value, it would be a cost. Our boss challenges us every day around these questions: How can data empower the business? How to create value? If the data just exists there, storage, hardware and other costs will have to be paid every year.

So, we were faced with four problems:

Data efficiency:

1. The amount of data from each brand is huge, and the existing system has a low computing speed, which affects the accuracy of the data;
2. The data demands and processing of each department are frequent and error-prone, so there is a need to improve the efficiency of data processing and report production. Basically, the reports are produced manually. The entire business and the data team have about three to four thousand reports.

Data dispersion:

1. Data from each sales platform, system, and business line are isolated and scattered across different systems, making it difficult to manage, integrate, analyze, and apply;
2. The formats of each platform are not unified, which hinders global brand and user analysis. At the same time, from a brand perspective, the data is very fragmented. For example, when looking at user portraits, there are Tmall user portraits and WeChat user portraits. As a result, when the marketing department is placing advertisements, the portraits with a full brand perspective are less useful as a reference. In addition, there is a large deviation in the cycle of the consumer's entire life cycle.

Data value:

The amount of users and data is large, but the value of the data is not deeply mined. Much of the data is dormant data assets with low utilization rates.

Data Application:

1. Existing tools fail to meet new user classification and marketing needs;
2. There are various user marketing channels, and it is necessary to integrate and plan user marketing methods to optimize investment and efficiency. Let data truly empower business and drive business growth.

In response to the above problems and challenges, the following mainly shares the data application practice part, which I abstract into three steps:

1. The first step is to segment users. Brand retail must segment users, and then look at the differences in the segments and the differences in user mindsets.
2. The second step is to develop a communication strategy, that is, to promote different products to different users at different times;
3. The third step is to use scientific statistical evaluation methods, that is, to speak with facts.

Specifically:

1. User stratification

For brand fast-moving consumer goods, the most important thing is the repurchase cycle, which determines the user mind.

1. Repurchase Cycle RC (Repurchase Cycle):

The repurchase cycle refers to the time period from the first purchase of a normal (average) user to the next purchase.

There are two ways to calculate the repurchase period:

1) Normal user calculation method

1. Normal (average) purchase amount
2. Normal (average) daily usage
3. RC = Normal (average) purchase amount / Normal (average) daily usage amount

2) Existing data induction method

1. According to the brand consumer purchase information, the average repurchase interval of all repurchase users is made into a distribution bar chart
2. The average repurchase interval of 90% of users is 3RC
3. RC = Average repurchase interval of 90% of users/3

Here is a small example:

The company's dishwasher detergent brand was introduced in China not long ago, so there is not much data accumulated yet. However, we can look at the business data. The PCT purchased by consumers at a time is 150 yuan, which can be used about 50 times (according to our estimates). Assuming that the dishes are washed once every two days on average, the number of times dishes are washed per day is 0.5 times. We can calculate that the daily user repurchase cycle is about 100 days.

Then we look at the 20,000 or so data in the current database and use existing data analysis methods. 90% of the users in its data center complete repurchase within 450 days, so RC = 150 days. However, for this brand, user mindset has not yet been formed, so our two methods are selective.

For immature brands, when the user mind has not yet been formed, the brand hopes to input a strong mind and reach it at the right time. We recommend that when it comes to immature brands, we use the calculation method we infer for normal users to define the repurchase cycle.

For mature brands, the amount of data and user accumulation has reached a certain scale. We still hope to follow the user mindset and use statistical induction methods to formulate a repurchase cycle. So what is the use of stratification after calculating the repurchase cycle? The following introduction to the classification of user stratification is very clear.

2. User stratification

1) User stratification: first-level classification

According to the customer's most recent purchase time, the first level of population classification is carried out. Here, the repurchase cycle number RC is used for the first level of division, and can be divided into potential groups, active users, dormant users, and lost users (Note: each customer will only exist in 1 category).

• Potential group: the target consumer group of the brand/category, who has not yet purchased the products of our brand/category;
• Active users (R ≤ 2RC): the last purchase occurred within 2 RCs;
• Sleeping users (2RC < R ≤ 3RC): the last purchase occurred between 2RC and 3RC;
• Churned users (R > 3RC): The last purchase occurred more than 3 RCs ago.

These four categories are still relatively extensive for brand retail, so we will conduct a second subdivision.

2) User stratification: double segmentation

We will further classify customers based on their R values, and from the perspective of potential, they will be divided into people with quality potential and people with competitive brands. The former are the target consumers of the brand/category and have not yet purchased products of the category; the latter are the target consumers of the brand/category and are customers who currently purchase products of competing brands.

• Active users: These are the users that need to be focused on. They can be divided into highly awake people (R ≤ 1RC), moderately awake people (1RC < R ≤ 1.5RC), and people on the verge of falling asleep (1.5RC < R ≤ 2RC) based on the time of their last purchase.
• Sleeping users: can be divided into light sleepers (2RC < R ≤ 2.5RC) and moderate sleepers (2.5RC < R ≤ 3RC);
• Lost users: can be divided into brand lost users and category lost users. The former are those whose last purchase time is more than 3 RCs and have switched to competing brands; the latter are those whose last purchase time is more than 3 RCs and have not purchased category products subsequently.

3) User stratification: triple segmentation

New customers, old customers, and loyal fans are the users who contribute to our GMV. Triple segmentation is another layer we divide from active users. When dividing further, in addition to purchase time, we also need to segment based on purchase frequency, contributed average order value, etc.

If our old customers make two or more purchases for the first time, they are considered old customers. For such fast-moving consumer goods brands, especially in scenarios with a strong e-commerce promotion atmosphere, many users have a purchasing power of only 9.9, which may not be considered old customers. The definition can also have requirements for the average order value. In this way, I will redefine new customers, old customers, and loyal fans to identify who are old customers and who are loyal fans (Note: PCT2: the average order value of the user's second purchase; V: a fixed amount, which varies by brand).

• New customers: users with F = 1 among active users;
• Old customers: active users with F = 2 and purchase PCT2 < V;
• Loyal fans: active users with F ≥ 3 or F = 2 and second purchase PCT2 ≥ V.

The above are general definitions of old customers and loyal fans, and each brand has its own determination method. Some use the average order value, some use the number of product categories, and some use the number of cycle usage. Each brand can also define old customers and loyal fans according to its actual situation. For example, A's V = 200, which is twice the store's average customer spending. The average customer spending of A's flagship store is 100. B is the amount spent on 4 bottles of disinfectant/sterilizer, V = 200.

After the sorting, the marketing goals of the entire department will become clearer, and our user marketing target KPI will become to improve the quality of active users and increase the number of active users.

First, we open up new sources of revenue to attract new customers; second, we reduce costs to consolidate, prevent sleep, awaken, and activate customers, so that our entire active user pool will become larger and of higher quality. The number is the sum of new customers, old customers, and loyal fans, and the quality of the entire active user pool is the ratio of old customers and loyal fans.

In the specific use process, there is actually a concept of labeling in it. After the most basic layering is completed, various labels will be added during the marketing process, covering basic attributes, crowd preferences, behavioral characteristics, marketing labels, RFM model labels, etc., and they will change in real time. The RFM model is an important tool and means to measure customer value through customer purchasing behavior.

• R: Recency, the time of the most recent consumption;
• F：Frequency, consumption frequency;
• M：Monetary, consumption amount.

We hope to integrate this label grid into the layering, as shown below:

For potential customers in category 1a, four grids will be affixed according to labels, such as brand potential population, category scenario population, and brand high-similar portrait population. The user mindsets of each group of people are different, so the content, materials, and communication timing provided will be different. Each small four-square grid can be superimposed with labels. The advantage is that after we have been doing marketing activities for a year, we can understand how much money has been spent on new customers, attracting new customers, promoting activations, and other events as a whole.

I have communicated with many industry brands before, and they are usually scattered when stacking labels. They may use the label "average order value" today and the label "browsing in store" tomorrow. However, I hope to make a more detailed segmentation on the underlying labels and then stack the labels on top, so that we can see it more clearly.

Similarly, active users can also stack tags, and even develop into specific applications such as eight-grid and sixteen-grid tags.

In addition, we need to do one more thing. The users have been segmented, but resources are limited, so we also need to make priority decisions. If we only have a budget of 5 million this time, which user segment should this 5 million budget be allocated to?

I suggest that the assessment can be conducted from two aspects: one is the importance assessment; the other is the changeability assessment. Some customers are very important, but their changeability is not that high. At this time, you need to make a judgment based on the actual situation.

Our experience in the past has been to focus our resources on preventing sleep first. For example, for customers who are on the verge of falling asleep, because they play a connecting role and are not asleep yet, but are on the verge of falling asleep, we will spend a lot of resources and energy to maintain them and prevent them from falling asleep and being lost.

2. Communication Strategy

After user stratification and priority are determined, a specific communication strategy needs to be formulated. Here I will talk about a framework model of the method - the 6W3M model.

When we were making this framework, the first thing we did was to sort out our logic, and the second was to turn this framework and logic into a capability iteration of the product, and then empower some new employees.

1. 6W

6Ws stand for Why, Who, When, What, Which, and Where.

• "Why" refers to why I want to do this. Every marketing must be clear about its communication purpose, which is to attract new fans, consolidate fans, and gain loyal fans. Each purpose is different, and the labels chosen later, the corresponding groups to be reached, and the way you reach them are all different.
• "Who" needs to filter different tags. There are five user stratifications for attracting new users. You need to define when to reach which circle, whether it is during an event, daily sales, or when performing a certain action.
• "What" refers to what content I give him. As I just mentioned, the mindset of each user segment is different. For users who are more sensitive to prices or like promotions, I will give him information about price discounts. For example, in the Finish just now, there will be high-end people who are concerned about the earth and environmental protection. At this time, there will be brand motivation and emotional communication given to these users, and there will be some promotional copy and content.
• "Which" refers to attracting new products, which must have strong conversion power and competitiveness. Flagship stores may be more brand-recognizable, while specialty stores may be more event-recognizable. The people corresponding to different channel positioning and product positioning will also be different.
• “Where” refers to what tools you use to reach it. Common ones are WeChat and text messages, and it will also link to in-site advertising promotion tools.

2. 3M

3M stands for Money, Measurement, and Monitoring.

• How much does “Money” refer to the total project budget? (Main investment budget/cooperative investment budget)
• "Measurement" refers to some evaluation indicators, such as click-through rate, interaction rate, conversion rate, per capita PCT, investment amount, sales, PC amount, etc.
• “Monitoring” refers to the real-time adjustment of communication strategies: AB testing of delivery to different groups of people, and delivery to different groups of people in different time periods.

We will focus on the selected groups of people to communicate with, invest most resources in important groups, and set reasonable evaluation and observation indicators. Here are a few examples:

1) 2c Daily sleep prevention communication strategy

For people who are on the verge of falling asleep, it is very important to prevent them from falling asleep, and we hope to keep it in the pool of active users at all costs. Therefore, we will design a series of marketing strategies to influence it. For example, we will have a five-level sleep-blocking communication mechanism in about a month, on Day 1, Day 7, Day 14, Day 28, and Day 35 respectively.

For different communication nodes, we will provide different content, interest points, channels, etc., so that we can continuously optimize our copywriting, channels, and interest points. Ultimately, we will do everything possible to keep users who are on the verge of falling asleep in the pool.

Here, someone once asked, what is our cost control level? At this time, we actually have a calculation logic to calculate a value based on the entire user's life cycle.

2) 1a Daily communication strategy for attracting new customers

After preventing sleep, there will be some daily actions to attract new users, which will require more cooperation with the business team.

Because labels and population packages are selected by the data team, and the final execution and reach are operated by the business team, when attracting new customers on a daily basis, we will give a lot of suggestions on labels and population packages to achieve precise reach. For example, for customers who have purchased a dishwasher but not Finished, the communication strategy should be: Function > Emotion > Motivation, and a discount on the startup kit should be given, and it can be placed in JD’s public domain SMS or Diamond Exhibition.

3. Statistical Evaluation

Statistical evaluation, generally speaking, includes two methods: one is the evaluation of a single data marketing activity, and the other is the evaluation of long-term data marketing goals.

Because user operation does not have a quick sales explosion like Juhuasuan and flash sales, user operation is a long-term "silent and subtle" process that requires you to continuously care for and serve consumers, which will have some mental impact on your products and brands. So at this time we will have some long-term goal assessments. How are these two assessments done?

1. Single data marketing evaluation

The effectiveness and efficiency of a single data marketing activity is evaluated through immediate or short-term indicators to evaluate the investment efficiency. The three elements of a single data marketing evaluation are: a large population base, good conversion effect, and low investment cost.

1) Population base

The population base refers to the total number of users successfully reached using the communication channel (where), such as the number of people who receive text messages, the number of people who are exposed to (click on) Diamond Ads, etc.

2) Conversion effect

The conversion effect is the conversion rate, conversion rate = number of transactions/population base.

3) Input costs

Input cost refers to the cost of communicating with users in data marketing activities, such as SMS fees, Diamond Ads fees, sampling fees, etc.

In addition, intangible assets can also be evaluated. In addition to indicators such as ROI for short-term evaluation of marketing results, it can also reflect the population value accumulated by the brand, the connection between the brand and consumers, and the brand building situation; traffic belongs to the platform, while consumer assets belong to the brand.

The indicator can be brand acquisition, which is to cultivate high-potential consumers into existing customers or interested customers, rather than just reaching more people with advertising.

I have been doing data marketing activities for a year to evaluate long-term KPIs, such as the number and quality of active users, which is developing and operating in a positive direction for the brand.

We will evaluate its long-term effects. As I just mentioned, the KPI is the number of active users. At this time, we will look at the overall number of new customers, old customers and loyal fans, whether there is a trend of increasing numbers compared to the previous numbers, and which marketing nodes will be of greater help to marketing.

In addition, the quality of active users, the proportion of old customers and loyal fans, the higher the quality ratio, the higher the quality of our active customers, that is, the greater the GMV proportion of subsequent old customers and loyal fans will be.

2. Evaluation of long-term goals of data marketing

Track and analyze consumer indicators with a long-term and developmental perspective, and evaluate data marketing direction and long-term effects.

Specifically, there are three relevant KPIs:

1. The rate of attracting new users is high and the number of active users increases. That is, the conversion rate of category potential and brand potential into our pool of active users.
2. The churn rate is low and the number of active users is increasing. That is, when 2C is converted into 3A, the proportion of our users who are on the verge of falling asleep and switch to light sleep is low.
3. The activity level is high, and the ratio of old customers to loyal fans is large. That is, user interaction should be greater.

Just looking at these few may be rather one-sided, so we also need to look at trends and compare with competitors as a method of long-term data evaluation.

Look at the trend:

• What: Long-term tracking and evaluation of core indicators, such as sales, brand consumer assets, number of active users, etc.
• How: It is necessary to clarify the changes of indicators in the time dimension, year-on-year, month-on-month, and forecast models;
• Why: Analyze and explore the reasons for the changes.

Compared to opponents:

• What: Set a reference system, such as industry market or competitor brand indicator comparison;
• How: Identify the gap between you and your competitors, the level of indicators, and the rate of change of indicators;
• Why: Analyze and explore the reasons for the changes.

Finally, the above introduction is easy to understand and use. It can be summarized into three points:

1. Lay a solid foundation, perform user segmentation, and gain a more detailed understanding of users. This should be assisted by tags, repurchase cycles, etc., and a rough user segmentation should be performed based on brand characteristics.
2. After that, we formulate communication strategies for different customer segments to continuously ensure that resources and budgets are under relatively good control;
3. After I communicated, I used scientific methods to evaluate it. In addition to short-term evaluation methods, it is still a long-term thing. There may be some long-term evaluation methods. At this time, we must have a better overall view and clear key points.

These three steps are not necessarily absolute and can be continuously optimized during the process. For example, we continuously optimize labels in communication strategies, continuously optimize strategies in the statistical process, etc., using human wisdom, system intelligence and business feedback to make these three things smoother and more efficient.

Author: Zhou Wenjun

Source: Sensors Data Decimal Point