How do leading e-commerce platforms achieve high ROI conversion from 0 to 1?

If I were to add a metaphor to the conversion rate, I would think that "conversion rate is like a practice, always in the process of understanding and continuous optimization." For any Internet platform, conversion rate (CR for short) should be one of the core goals. The gaming business considers the conversion from active players to paying players; the advertising business considers the conversion from exposure to visits; and the e-commerce business considers the conversion from consumer visits to successful orders.

The essence of conversion lies in commercial traffic, how to efficiently transform user operation behavior into commercial behavior, and continuously shorten user decision-making time to generate more sales. The development process of any project has phased nodes, and the same is true for the improvement of conversion rate. The user base, retention rate, churn rate, platform stay time, etc. are all factors that determine order conversion. These indicators will also be reflected differently as the stage cycle of business development changes.

Next, we will divide the conversion rate solution into two parts: project case analysis + improving core methodology (the project case will share the business development process with you, so as to extract efficient methodology and provide evidence review!)

1. Periodic project development process, starting from the node to consider resource allocation

(Project case analysis)

We (Walmart) only started to participate in the new retail home delivery project in 2018. The severity of the situation and the challenges for us are self-evident, but our advantage lies in the fact that we have nearly 500 strong offline supermarket stores resources and traffic advantages to support us. But how to quickly digitize traffic and effectively convert it has become our top priority.

At the beginning of the project, we began to formulate a cycle for store launches. Each cycle was highly targeted, whether it was the CB (internal testing) stage in the first two months or the large-scale launch stage in the later stage. At the same time, there are differentiated strategies in attracting new customers and conversions. (See the figure below for details)

1. The conversion rate is the highest during the internal testing phase, and high profits stimulate employees to place orders

The strong offline resource advantage requires the replication of an effective conversion strategy model. Therefore, after formulating a series of operation process models, we repeatedly conduct A&B Tests in the CB (internal testing) stage, from promotions to coupons, products to ordering processes, and try to get the highest ROI strategy.

The project was officially launched in August 2018 and the testing lasted for nearly two months. The core purpose of this phase was to understand the overall product process and conduct an investigation and analysis on the promotion acceptance. This stage is convenient for testing the promotion system. We encourage internal employees to place orders and receive benefits. We have launched various types of store discount coupons, so the conversion rate is the highest. Finally, we selected three coupon types with the highest ROI (29-10, 79-25, 100-30) from the order conversion data to facilitate our reference in the next stage of delivery.

2. Small-scale public testing to analyze real user conversion data

Around October, we started to conduct OB (public beta) in a store closest to our headquarters. During this stage, we introduced 3k-4k users. The number of users at this stage is still in its infancy, and the conversion strategy is still inclined towards investments in coupons, discounts, flash sales, etc.

After the first round of internal employee CB, we launched coupon packages in stores (59-15, 79-25), and the ROI can reach the level of 4-4.5. For flash sales and discounts, we will select the TOP50 SKUs with the highest sales rate and offer 15%-10% off promotions. For new users, promotions and coupons can be combined. Secondly, there will be limited free shipping benefits. The overall real conversion rate is still very high. At this stage, we can maintain it at 30%-35%, but high conversion does not mean health, so we need to move on to the next stage.

3. The large-scale public beta test of 100 stores began to refine operations and analyze data health

The launch of 100 stores is a very important milestone for us. Before this stage comes, we started to build a data labeling system and user recommendation algorithm in the second stage. This set of data BI began to empower services for this stage. Our user base has reached 500,000-1,000,000, and most of them are new users. It is relatively difficult to convert this part. Although there are high-profit incentives, the user profiles of all new customers come from offline, and their age range is 40-50 years old. Therefore, we need to change this part of the shopping habits, and the psychological threshold limitations are very large. Therefore, the conversion rate at this stage is the lowest, approximately 10%-15%.

As mentioned above, on the one hand, we built a data labeling system and began to stratify users. Based on the frequency of shopping category preferences and previous shopping attributes of offline QR code scanning, we divided users into ID packages and pushed contacts according to the RFM model and portrait model. Secondly, in order to improve conversion, after we distinguish the user types, we will also actively BI users to receive coupons center. For example, if we recognize that you are a mother and baby, we will give priority to recommending category coupons with a value of 168-30 or more. Including the flash sale channel in the prime position on the homepage, category rotation recommendations, etc., all of them are biased towards maternal and child categories and related categories, which are more direct and effective in driving conversions.

It is worth mentioning that in terms of product experience, the coupons we choose are from the WeChat ecosystem's own coupon system. After the user has not used them, WeChat will actively send a service notification to remind the user, and there is also an obvious reminder in the WeChat card wallet, which is a good buff for user return. For flash sales, we will also set up a one-click subscription reminder function, and can also reach return through public account event appointments and service notifications. This product function is well closed-loop with the operation strategy.

4. Smoothly launch all stores nationwide with multiple formats to boost conversions

The fourth stage is a stable cycle. The reason why it is called “stable” is that it requires a smooth launch of stores across the country, a stable balance of business profits and expenditures, a stable conversion rate CR output, etc.

The launch of stores nationwide means that the user base has also doubled to 2-3 million, and is still increasing. We need to gradually increase the conversion rate through multi-dimensional means.

Home page resource management: The display of resources needs to start targeting two groups of people: users and brand owners. When it comes to users, we need to provide different faces for different groups of people in corresponding categories, and when it comes to brands, we need to monetize resources. In order to realize monetization, we need to let users generate sales, so we have formulated a series of reward and punishment measures. We will reward high-volume brands with more resources, and we will punish low-volume brands by removing them from the shelves in advance.

Refined user management: User stratification is also multi-dimensional. The RFM model, category model, and AARRR model are all stratification principles we have derived based on big data analysis. The larger the amount of data, the smaller the granularity of user segmentation, and the greater the possibility of conversion.

Algorithm recommendation promotion: The positions and modules on the homepage are flexible and have big data allocation. The layout/content of the modules will change as the user shopping data accumulates. At this stage, the algorithm recommendation model has gradually matured, and the data converted on the homepage is the highest in the entire site and closest to the platform average.

The digitization of offline scenes of people, goods and places repeatedly reminds users that offline scenes are rich and complex, and there are many scenes that can increase the possibility of conversion.

Taking "people" as the carrier, offline supermarket units are nothing more than store guides/pickers/cashiers/operators, etc. This team actively promotes to users while sticking to their posts. We increase distribution channels. Guides can guide users to scan codes to place orders, and then they can receive online coupons and pick up the goods in stores, realizing the convenience of buying and going immediately. We will also give these guides a commission of 5-7 yuan as a reward, so they are very motivated;

The unit of "goods" is the SKU of the entire store. We create online product details for the department's high-frequency SKUs. Users can scan the QR code on the shelf to view more product information, user reviews, and traceability information online. Of course, the most important thing is that they can get online coupons for the corresponding categories. Users can jump to the platform with one click to increase the possibility of conversion;

The unit with "field" as the carrier is the supermarket itself, the store signage/digital advertising screen, the user starts from entering the store, from the store entrance, the stacking area, the shelf pallets, the shelf baffles, the hanging tags and flags, the oncoming columns, the wall posters, the store brochures, the store exit and other user-necessary routes, adding online QR code advertising + online promotional interest points. The key point is that as long as we can successfully attract users to our platform, we are one step closer to conversion!

After one and a half years of development, the project has a relatively stable conversion rate of around 20%, and is still accumulating user data and improving refined strategies. Next, I will analyze some thoughts and methodological extractions during the project case process, hoping to help you in actual combat!

2. Understand user psychology and accelerate the link from demand stimulation to decision-making

(Core methodology for improving project conversion rate)

When users have shopping needs, they will always be influenced by various advertisements, promotional information, product screening, quality considerations, price comparisons, delivery time and other factors. After a series of behavioral operations and ups and downs in mentality, they will finally decide to press the payment password. We might as well break this process down into three steps: the demand generation layer, the target recognition layer and the final decision-making layer. We will use a mind map to sort out the 3 steps for you.

1. The motivation layer is also the demand generation layer. This layer is generated by the subconscious mind after being stimulated by internal/external stimuli. After being generated, a mapping relationship is established with a certain category, thereby solving rigid or non-rigid needs. Therefore, there are several points we can start with as entry points: internal and external stimuli, the mapping relationship between categories, and rigid and non-rigid needs.

• Internal and external stimulation

This link can be associated with the "perception layer" in our figure. Internal stimulation comes from human food, clothing, housing and transportation, and the living conditions are stimulated from the inside out. External stimuli come from external stimulation input from outside through advertisements, things, and environment. The core points of conversion rate lie in user attribute analysis, user shopping frequency and refined user outreach.

For example: Pregnant mothers have to buy milk powder once a month at a fixed time. We will push a milk powder coupon and reminder 2 days before the purchase to accurately reach them. The accurate timing and demand conversion will be much higher.

• Category mapping relationship

This is related to the "cognitive layer" in the table. After having the impression of the product, the next step is to preliminarily clarify what this "thing" is! Why can you handle my needs! The key point in conversion rate lies in precision marketing. A text message or a push notification may make the difference. The content should not be too complicated. 1-2 sentences that are close to the user's pain points will suffice!

For example: 9.9 grab a 9.9 odorless and non-toxic mosquito coil for a peaceful sleep and a whole night of rest.

• Just need and non-just need

Human beings are always greedy, with needs for survival, enjoyment, development, aesthetics, social symbolic needs, and so on. At this time, we need to quickly grasp the corresponding needs of certain types of people, so we need to continuously refine the user model and break it down using the AARRR model + user stratification/incentive methodology.

2. The thinking layer is also the target cognition layer we just talked about. This layer is the process of passively or actively receiving countless target objects through users generating needs and initially clarifying the target objects. In this process, we can completely identify the objects and generate shopping intentions

Information sources are divided into active and passive, active information search and passive information input. The conversion rate of actively seeking information must be higher than that of passive input, and the products actively sought are easy to satisfy.

For example, orders placed through the search box on a certain JD.com shopping app account for more than 40% of the entire site. Search quality is particularly critical here. Keyword mapping, category association, promotion mapping, and keyword fuzzy search are all factors that improve search conversions.

Passive input information is more inclined towards product marketing. When we start to capture these potential customers and find that they have browsed before, big data capture and user reach are very critical at this time. We monitor the user's purchasing performance based on the bounce rate of the product details page, compare the bounce rate trend horizontally, and at the same time, we need to reversely infer the rationality of the product information, and do secondary optimization based on the user's focus points corresponding to the product, and repeat the test after one week to see the effect.

3. The action layer is the last layer, and it is also the most difficult choice layer for users. The amount of information at this time must be the largest, and it is also a process that users need to filter and integrate. Therefore, it is most important to tailor information communication to each individual.

Scenario A: A hot mom buys retail food for her children. During the selection process, she repeatedly observes and understands the production date, freshness, safety and price of the food.

Scenario B: A group of students are shopping for retail products. They focus more on the appearance, whether it is an online celebrity model, brand awareness, and of course the price (after all, their purchasing power is limited).

The purchasing groups in the two scenarios are different, and the focus of the information communication we present to users will also be different. Let me share a detailed case. When we are operating a private domain traffic community, after we classify and manage the users, the messages, activity content, and promotional content we promote will vary depending on the group of people.

The importance of emotional cognition lies more in touching the user's mentality, and the conversion between rigid needs and non-rigid needs will also be more obvious. Idol promotion, packaging of popular products by internet celebrities, content marketing, viral videos/print marketing can all be good entry points.

For example: We use the online media matrix of Mengniu spokesperson TFboys for live streaming and offline store meet-and-greets. We embed exclusive QR code pages for the event online and offline. Users scan the code to enter the page to receive coupons and place orders. Fan hosts invite fans to enter the page to vote for the top spot. If the ranking reaches 1,000, not only can the user get free milk, but it can also drive many new users to generate sales.

3. Reasonable operation means gradually increase the conversion rate CR

When talking about the relationship between operations and conversion rate, it must be multi-dimensional, and the conversion forms in different application scenarios will also be different. Let’s not assume for the moment that after a user enters the platform, a valid UV can achieve a valid CR. We break it down into six major dimensions: resource management, promotion system management, product management, precision marketing, content management and path diagnosis.

1. Resource management, every inch of land is valuable to ensure CR

Resources are always a battleground for e-commerce platforms, so standardized management and standardized indicator considerations are very important. Here are a few principles:

• Resource position design: simple and clear layout, clear and organized products, 1 main benefit copy + 1 subtitle, main button to guide jump
• CR data of resource positions: Take the two CR values ​​of the pit position during the off-peak period and the peak period respectively, and take the average as the standard. It is best to observe the data trend for one week and take a reasonable value.
• Compare the overall CR differentiation: The CR of the first screen resource position on the homepage must be greater than the overall CR. At the same time, compare the CR of each sub-page to differentiate the carrying capacity of each module.
• Resource position reward and punishment system: 1 month sedimentation analysis, rewards and punishments according to resource cycles, and selects high-value resources. For example, a brand puts a full-width resource on the homepage during the 618 promotion. One day later, if the CR of the activity is much lower than the previous CR, the resource position will be quickly adjusted or even removed. On the contrary, if the production performance is excellent, the reward days will be released.

2. Promotion system, bundling sales to increase CR

The principle of product bundling sales is to guide individual UVs to place multiple products or multiple orders out of interest. At the same time, this user is encouraged to attract additional users to place orders, thereby improving the overall CR:

• Brand/category discounts: When a brand or category reaches a fixed price, you can enjoy discounts and instant discounts
• Full discount/full discount/free shipping: You can enjoy discounts, instant discounts or free shipping when the total amount of the shopping cart reaches a fixed price
• Free shipping/discount for N orders on the same day: free shipping for the second order on the same day, or 9.5% discount for the third order, etc.
• Invite a valid user to place an order and get cash back: Share and invite friends to place an order, and get 10%-15% cash back
• Group CR: community group buying/multi-person group buying/etc.
• Virtual currency system: full-site/category/brand coupons, sign-in points exchange coupons for multiple scenarios

3. Product management: expose the right products to the right people

• Core category impression: Establish a core platform to penetrate and penetrate, focus on user minds, and select 5-8 core categories for vertical platforms
• Products are divided by activity: the product library is divided into three categories according to activity rate: high activity rate product library/medium activity rate product library/low activity rate product library; and the three categories are displayed on the homepage according to the cycle frequency
• Products are divided by price: Products are divided by price, and products with high dynamic effects and no more than 20 yuan are selected as popular products to attract traffic and establish a seed product library
• Hot items in each category are exposed: 20 SKUs are selected from each category to increase shopping efficiency on the homepage
• Thousands of personalized product recommendations: based on user category preferences/category frequency exposure and intended purchase categories

4. Precision marketing, user segmentation, and improving conversion rate starting from users

• Regarding users, I have written two detailed articles to introduce them. Students who are interested can review the official account. Here I will share with you a few key points:
• Differentiation between new and old users: new users can receive a benefits package (50% off coupons for the first order, special prices for 10 flagship products, free shipping coupons) in a fixed module on the homepage, and repurchase coupons for old users, etc.
• Category user differentiation: Use private domain traffic to group users by category (social group/public account fans/SMS package/user package), and push different category packages (coupon + product + flash sale benefits) to different users
• User attribute segmentation: According to the basic information such as user order frequency/order customer order segmentation/user age group, the user can be segmented and captured, and precision marketing can also greatly improve the conversion rate.

5. Content management, from product content to marketing content to improve user conversion

• Product title: Brand name + product name + keyword interest points (preferably within 15 words), subtitle adds some promotional interest points, product core interest points
• Product header image: two product images from different angles + three product application scenario images + one product image with packaging (6-7 images are best)
• Product details page: The overall style must be consistent with the product's tone, such as green/orange for food and gray/plain colors for dairy products; secondly, the layout should be concise and capable, with 30% product display + 50% product application scenarios + 20% product after-sales service and relevant qualification certification information, etc.
• Content marketing is also part of the platform's efforts to improve conversion rates. We try to build a "Lifestyle Museum" on the platform. We will embed a large amount of UGC and PGC content in the channel, and push some popular science information, seasonal food recommendations and other information through pictures/texts/short videos. The conversion effect of this part is not particularly obvious, and the viscosity is relatively high

6. Path diagnosis, starting from UV to diagnose the effect of each step

• From UV to platform stay time: 1 valid UV enters the platform, the first step is to consider the length of time on the platform. From the data of our early business, on average, each valid UV stays on the platform for 50 seconds, so users who stay for less than 50 seconds and do not make purchases are what we need to pay attention to, such as user ratio/number of people, etc.
• Home page resource position - activity landing page: After browsing the platform, the user jumps to the landing page by clicking on the resource position on the home page. This part of the jump rate is what we need to pay attention to in the second step. Usually, the jump rate on the first screen of the home page will not be lower than 20%, and the peak values ​​of the full bar or startup picture can reach 40%. Therefore, if the data is lower than this part, we need to reflect on whether the resource position material is not well prepared or the entry interest point does not attract users to click.
• Event landing page - product details page: After the user enters the event landing page, the next step is to click on the product details. This part is the time to test the product content. The jump rate through the landing page header can reach 15%-20%.
• Business details page - add car: The efficiency of adding car is the indicator we consider in this step. It should be noted that the lower the add car rate is, the lower the conversion rate is. The higher the rate is, the better the conversion rate is. Because most potential users will buy directly without adding a car after they have the desire to buy, we use the car adding rate as a reference indicator and do not make it a core assessment content.
• Product details page - order: This part of the page is the terminal behavior indicator and is particularly important. The conversion from business to order also requires screening 1-2 months of valid data to make judgments and define the middle value. In the early stage of our business, this threshold can reach 10%, and the overall platform can also reach 13%-15%, and the peak can reach 20%. This data is already above average for mini program e-commerce

4. Transaction Funnel Model Path to See Conversion Rate

Operations are always inseparable from "models", and the same is true for conversion rates. Let's take the first shopping experience of an unregistered new user as an example. We list several necessary steps for users from search to final transaction. The user's direction in the process must be gradually reduced. What we need to do is to reduce the user churn rate in each step of the process!

1. Model process visualization

The model is based on a visual mind map, which is easy to analyze and disassemble (for example: the golden process from user search to final order)

2. Model process disassembly

After the model is output, we need to constantly challenge ourselves and answer ten questions and ten answers. At least list three or more challenge points for each key process (from the user side, operation side, etc.). This collection of challenge points is packaged into our attribution factor.

3. Granularization of attribution factors

Attribution factors always involve a lot of details. At this time, we should grasp the core of "which is the factor that affects the final transaction" and continue to polish and verify with big data.

5. Notes on the relationship between the funnel model and product experience

The funnel model is gradually disassembled based on the user's operation steps, so the entire product experience is closely related at this time. Here are a few principles to grasp:

1. When disassembling the funnel model process, it is necessary to consider shortening the model

For a successful product, the conversion path must be concise and simple, and the entire core golden process of the product can even be completed in 2-4 steps. The same is true for new retail e-commerce. While we are disassembling the model, we need to continuously shorten the product process.

For example: users can go directly to the add-to-cart page with one click in the home page resource product showcase, go directly to the flash sale settlement page with one click in the flash sale channel, and the shopping cart settlement page automatically calculates the best combination of matching coupons, etc.

2. If the order of nodes in the funnel model is changed, it will not affect the core order conversion

The golden process is definitely feasible in an ideal user operation process, but the actual user operations are ever-changing. We need to repeatedly consider the process nodes. Except for necessary nodes, we assume that the remaining process operation nodes can be freely combined. We also need to ensure the smoothness of the overall process.

For example: After searching for a product, a user does not complete the purchase through the search page, but instead completes the purchase through the recommended products on the shopping cart page. This is a scenario where the deal is closed even though the ideal process is not followed.

3. User churn traces during the funnel model process need to be recorded and analyzed

User churn is a huge waste of initial new user acquisition costs for operations, so during the entire funnel model process, we must use button embedding with parameters to record at which step the user churns, and analyze the reasons for churn to determine whether they can be optimized or awakened through active contact.

For example: The user's bounce rate on the product details page is abnormally low. At this time, the user browsing trace data of the product details page is very critical: page dwelling time/page sliding traces/, attribution problem points.

6. Heatmap Analysis and Verification Model Validity

I believe everyone is familiar with heat maps. Heat maps are usually used in product layout/functions/jumps, etc., and analyze user operation traces through the distribution of user click data.

(1) Extract the heat map of each page corresponding to the model and verify the effectiveness of the process based on the heat distribution.

As in the example above, assuming that users have different sub-pages for each step from searching to placing an order, we try to extract each sub-page, arrange 4-5 pages in order, and then look at the heat distribution to see whether the high-click areas meet the expectations of the entire model process.

For example: home page heat map + search page heat map + search landing page heat map + product details page heat map + shopping cart checkout page heat map, and connect them together to form a process, then look at the heat distribution, it is intuitive and obvious!

(2) The heat map is divided into 1, 2, 3, and 4 levels. Levels 1-2 are classified into the ideal process and the reverse verification model steps.

The distribution of the heat map uses color blocks to distinguish the degree of clicks. We regard the heat distribution of levels 1-2 as the ideal process, and reversely infer from the description in point 1 above to see whether the distribution of levels 1-2 meets our golden process. Only with positive verification + reverse verification can the model be valuable.

(3) 3-4 levels are classified as abnormal processes to explore potential user groups.

We cannot blindly regard the heat distribution of levels 3-4 as invalid clicks. Similarly, extract several key pages, circle the key points of each heat distribution of levels 3-4, and then try to piece together several pages. At this time, you will find that perhaps the user operation is also feasible, or it ultimately failed and was stuck at a certain step! Please remember that these users must be high-potential users, and the ID packages for active contact and awakening are very clear!

For example: On the product details page, the user did not add items to the cart and place an order. Instead, the user was redirected to the flash sale channel through related category recommendations and subscribed to the flash sale channel, but did not place an order. This user must have high potential value in the category and is price-oriented. At this time, you can also capture the ID user layer and try to push it.

7. Reasonableness of the homepage layout

Whether it is new retail or pure e-commerce, the conversion rate of the homepage must be the highest and best represent the core culture of the entire platform. It is prudent and critical to do a good job in homepage conversion!

1. The overall functional layout and resource display need to meet the overall platform promotion rhythm

Any brand/category/single product hopes to display its own resource position on the homepage. We need to judge which suitable promotions to put on the homepage based on the product turnover rate and user tag attributes. The arranged resource positions/products need to be verified after going online to see whether they have reached the platform's average conversion rate level, and set up reward and punishment measures to encourage and stimulate high-yield content delivery.

2. Reasonable setting of multi-dimensional shopping guide columns

The number and richness of shopping guide columns is not the more the better, but of course they cannot be too few either. A balance is needed to ensure that users do not get lost. Here are some tips for distinguishing:

• Differentiate by user groups: allocate resources from head to long-tail according to the proportion of different core user groups, and connect to the display of thousands of people with thousands of faces, and increase personalized PGC content, such as Xiaohongshu/Mogujie.
• Differentiate by promotion form: The promotion forms of each module on the homepage are different, such as activities/game play/discounts/group purchases/flash sales, etc. This form is the most common and should be the most widespread in comprehensive e-commerce, such as Taobao/JD.
• Differentiate by product category: Whether it is the homepage or the category page, it is straightforward to differentiate by product category, with clear communication and no other complicated content, such as Pinduoduo/Daily Youxian.

3. The homepage is responsible for conversion and needs to be efficient in traffic diversion

On the premise of assuming production, the homepage also has a very important function of diversion. What needs to be noted here is the depth of the path from the homepage to the product details page. Make the exposure of products reasonably shallow, and carefully consider the means of diverting traffic from each page to the next level page.

For example: If there is a brand of milk on the homepage, try to encourage people to click directly to the milk; if the milk is just for decoration and guide the click to a sub-channel page, then please make sure that the product is displayed on the first screen of the channel page after the jump. Give users expectations that are not disappointed!

4. Improving visit depth indirectly creates favorable conditions for conversion rate

Regarding the improvement of visit depth, here is a scenario example: if you observe carefully, offline stores usually put high-frequency fresh fish, shrimp, crab meat at the innermost part of the store, and the upper and lower handrails of the elevator entrances are always one in front and one behind, one on the left and one on the right. The reason is simple, the store hopes to increase the user's shopping time and scenarios! If we copy this principle and apply it to the online world, is it feasible to hide high-frequency categories at the deepest part of the path? It goes without saying that this is definitely not possible, you will die faster!

This is a big difference between online and offline store visits. The essence of online user operations is to accelerate user purchasing decisions. The significance of increasing visit depth does not lie in increasing the time spent shopping before purchasing, but in having more visit possibilities after making a quick decision, thereby increasing exposure for additional conversions.

Therefore, what we usually do is to put some interesting gameplay and promotions at the bottom of the shopping cart/payment result page, such as the decision-making quasi-completion page, such as a roulette wheel/lottery coupon/related promotional categories/gift package benefits, etc.

No matter how deep the practice is, there are certain rules and patterns to follow, and the same is true for conversion rate/penetration rate. Build a rule model and break it down step by step. Verify and optimize from several different angles, including users/colleagues/boss/mom, etc. Even the most novice user will always have some unexpected surprises!

Author: leon

Source: Leon Operation Notes (Leonnote123)