More and more enterprises or teams like to use funnel model To observe business data changes, for example, Twitter uses funnel analysis to improve user participation, Baidu takeout uses funnel model to analyze flow subsidy trend, and Little Blue Cup Rising Coffee uses funnel model to analyze the attractiveness of coupons to new users.
Danny Jon, who once worked for Facebook, Twitter and Quora, a large company, has experienced this once. In his Growth hacker In my career, I encountered such a problem: traffic from search engines accounted for 60% of website traffic, but less than 1% of them turned into registered users, which means that 99 out of every 100 visitors eventually lost. The conversion rate of 1% made him feel that he and his team were inefficient. Danny's team spent three months studying the optimization of landing page, and finally successfully improved the conversion rate to more than 10.5%. In other words, they increased the conversion rate by more than 10 times, created new value for the company, and made more and more people pay attention to him and his team.
What he uses is the "funnel model", so it's so useful Data analysis What is thinking? How should it be integrated into the business? Today we will talk about the funnel model from the following aspects, hoping it will be beneficial to your work.
(1) What is the funnel model?
(2) How did the funnel model evolve?
(3) How can e-commerce websites use the funnel model?
(4) How does the funnel model guide the optimization of landing pages?
1、 What is funnel model
The funnel model is a set of process type Data analysis , which can scientifically reflect the user behavior state and the user conversion rate at each stage from the start to the end, is an important analysis model. The funnel analysis model has been widely used in traffic monitoring, CRM system, SEO optimization, and products for website and APP user behavior analysis Marketing And sales and other daily data operate And data analysis.
Funnel analysis is the most commonly used conversion rate and loss rate two complementary indicators. Let's use a simple example to illustrate that if 100 people visit an e-commerce website, 30 people click to register, and 10 people successfully register. There are three steps in this process. The conversion rate from the first step to the second step is 30%, and the loss rate is 70%. The conversion rate from the second step to the third step is 33%, and the loss rate is 67%; The conversion rate of the whole process is 10%, and the loss rate is 90%. This model is the classic funnel analysis model.
2、 Evolution of funnel model
The concept of funnel model was first introduced by St Elmo Lewis (a famous American advertiser) put forward in 1898, which is called the purchase funnel, also called customer funnel, sales/marketing funnel. It is a marketing strategy for brand advertising, accurately outlining the customer's process on products or services. The funnel model is mainly used to decompose and quantify each link in the process, help effectively find problems and optimize them, so as to improve the overall operational efficiency.
This strategy proposed by Lewis was later called AIDA model, namely consciousness interest desire action. In the next 100 years, with the promotion of funnel model, in order to adapt to the new media platform and the change of user behavior path, it has been modified and expanded for many times, producing various derivative versions, such as the well-known AIDMA, AISAS, AARRR and other models. Since these models can still guide our work in normal work, we will introduce them respectively as follows:
1. AIDMA model
The AIDMA model is based on the AIDA model (Attention, Interest, Desire, Action), which adds Memory and forms a model of attention → interest → desire → memory → action (purchase). From attracting the attention of consumers to arousing the interest that users can turn to desire, and being able to remember enough time for users to take action (purchase at the next time).
The AIDMA model is mainly applicable to brand marketing. Of course, many Internet products have begun to build their own brands, such as Pinduoduo, Tiaoyin titled variety show, popular H5 screen painting, Netease Cloud Music's subway screen painting advertisement, etc., all from arousing users' interest, strengthening brand memory, and attracting potential users. However, the user process of AIDMA is not transformed immediately, and there is a lack of user feedback after purchase.
2. AISAS model
Because the AIDMA model lacks the link of user feedback, and with the completion of Internet user education, consumer behavior patterns have changed, and then the AISAS model (Attention, Interest，Search，Action，Share)， That is, attention - interest - search - action - share. From receiving the product's promotional and marketing information (hard or soft), to arousing interest, users began to search for information (Baidu, Zhihu, Weibo, Taobao), to online download or payment, and subsequent evaluation sharing links (within the product, WeChat Weibo).
The AISAS model is more in line with the characteristics of the Internet and has strong timeliness, but like the AIDMA model, it still lacks quantitative standards, and the effects of each link cannot be fed back through data.
3. AARRR model
AARRR model is a business growth model proposed by Dave McClure (founder of 500 Startups) in 2007. It includes five stages: acquisition, activation, retention, revenue, and referral. It is regarded as the five most important indicators that the company focuses on, because these indicators effectively measure the growth of products, and are simple and operable at the same time. In previous articles《 Super detailed analysis of APP data index system | Recommended Collection 》If you are interested in improving the AARRR model, you can check it.
3、 How can e-commerce websites use the funnel model?
The funnel model is most widely used in e-commerce websites. As shown in the figure, most users of e-commerce products need to go through several steps from the home page to the final payment: goods/browsing categories - viewing product details - adding shopping cart - generating orders - starting payment - completing payment. For e-commerce products, the main purpose is to place an order for payment. Therefore, the transaction conversion rate is a global indicator to measure the whole process. For a single link, it is generally UV, CTR (click through rate), page dwell time, conversion rate, bounce rate, etc.
We need to monitor the user's behavior path at all levels of the process to find the optimization points at each level; Draw the transformation path for users who do not follow the process, and find out what can be improved User experience To shorten the space of the path. The following will describe the flow on the home page - search list page - details page - add shopping cart - submit order - re purchase stages:
1. Homepage traffic
We usually channel traffic from various external channels to the home page through various means. At this time, the quality of traffic is very important. The indicators commonly used to measure the quality of page traffic are page UV click rate, page dwell time, and bounce rate. Users interested in the home page will generate necessary click behavior, and click behavior will generate two data: page UV click rate and page bounce rate.
Page UV click rate=page click times ÷ page UV number
Jump out rate=the number of times to leave after entering through an entrance ÷ the total number of times to visit through the entrance.
The higher the click rate, the more effective the content presented on the page can attract users' attention; The higher the bounce rate, it means that the page presentation content and copy do not meet the user's expectations. Therefore, the optimization of the home page is to improve the click rate of the page, reduce the skip rate of the page, and try to let users enter the next page.
In addition, this data can also be used to judge whether the quality of traffic sources is up to standard. Generally speaking, when excluding page problems, low page traffic quality often has three characteristics: low click rate, high skip rate, and short page dwell time. These low mass flows are mainly caused by:
① The copy content presented on the channel drainage is inconsistent with the landing page;
② Other reasons, including but not limited to technical jump errors, such as errors in the follow-up page;
③ The channels that do not match the target users are launched, which means that the channels are not accurate (as shown in the figure, the quality comparison of the external channels).
2. Search list page
The search list page plays an irreplaceable role in large-scale e-commerce websites, and it is also the main source of traffic in the site. It undertakes the task of searching goods and category layout in the site.
The search page is an overall search based on the keywords entered by the user and presented to the user's commodity display page. The list page is directly related to the background of the website's commodity category, presenting the most comprehensive category of pages. Both functions are to give users a better and faster positioning of the commodities they want to view. Therefore, the data indicators in this level of pages include:
(1) Search click through rate =Click times/search times, this indicator measures the presentation quality of the search page;
(2) UV to detail page conversion rate =Details page UV/search or list page UV, which is also applicable in search and list. It is used to balance the possibility of cheating in click rate and is one of the indicators reflecting the quality of details page;
(3) Search No Results : It is used to reflect the missing or unrelated indicators of brand categories involved in keywords. Of course, the lower the number of searches without results, the better. For categories with empty search term results, it is necessary to decide whether to open investment invitation and introduce product lines for relevant categories after comprehensive evaluation; For categories that are not associated, the page needs to be optimized and re associated;
(4) Hit rate of the first screen of the search results page =The number of clicks on the first screen of search/number of searches is used to measure the ranking quality and presentation quality of the first screen of search results. The quality of the data index can indirectly reflect whether the page ranking presented by the search term is reasonable and meets the needs of users;
(5) Number of searches and people : The number of searches generated by a search term or the number of people searching for it. The higher the number of searches and the number of people searching for a search term, the higher the demand for the category involved in the term, and vice versa;
(6) Advanced Filter Item Hits : In the search list page, the advanced filter items at the top of the page are set up to provide quick positioning, and the click times and usage rate of the advanced filter items can also provide operators with a reference to the popularity of goods.
After sorting out these indicators, how to analyze these indicator data?
① Focus on the keywords with high search volume ranking high according to the number of times and numbers of search terms. High search volume represents high attention and strong acceptance;
② Popular search terms focus on their click rate and arrival rate of detail pages. The click through rate is too high, and the arrival rate of UV to detail page is average, which may lead to click cheating (used by merchants to refresh rankings); High click rate and high arrival rate. It indicates that the search result page of the keyword works well, and it also indicates that the ranking of the keyword page is reasonable, otherwise it needs to be optimized;
③ If the number of clicks on the parameters of advanced screening items is too low, and the utilization rate is low, it is necessary to reset the display items of advanced screening items to improve the utilization rate;
④ For keywords without results, in-depth analysis is needed to determine whether they are system problems or involve products that have not been introduced, and feedback is given to the investment promotion and procurement department as the basis for procurement.
In general, the data analysis of the search list page can be summarized as follows: Focus on optimizing high search terms to improve their click conversion; No result word analysis feedback; Click on the page to focus on high filtering applicability, which is convenient for users to quickly locate. The ultimate goal is to let users sink to the details page.
3. Details page
As a key page for traffic conversion, the details page is the most basic unit for carrying commodity information and the most important part for users to decide to place an order. Therefore, when analyzing the detail page, the data indicators are more about the quality and conversion rate of the detail page. From the perspective of data quantification, the quality of detail pages is the average page dwell time and the number of shopping carts added.
(1) Average page dwell time=total page dwell time/number of UVs visited This indicator is obviously related to the presentation layout of the page, including product parameter introduction, detailed picture description, customer service online situation, favorable rating, etc.
(2) Number of shopping carts added : It is a key step to reflect the number of buyers of the product. The number of shopping carts added is basically determined by several factors, such as the comprehensive quality of the detail page (pictures, layout, display, parameter description, after-sales information), the comprehensive service index of online customer service (response time, online duration, response satisfaction), and the evaluation information (positive rating, negative response content, and presentation information).
4. Shopping Cart
For e-commerce websites of FMCG and standard products, shopping carts are set up to save the payment time for users to select multiple goods, and to improve the customer price. Shopping carts are bound to get twice the result with half the effort in combination with full discount coupons and other promotional means.
If there is a large backlog of goods purchased by customers in the shopping cart, but the user has never placed an order to pay, then you need to use SMS, email, push and other means to promote user transformation.
5. Order
The order page is the last link of vertical conversion. The main purpose of this interface is to make users pay as soon as possible to achieve the final conversion.
Effective order conversion rate=number of clinched orders/number of effective orders. It is relatively simple to facilitate conversion at this stage. If the effective order conversion rate is low, it is necessary to analyze whether there are problems in the payment page, whether the system submission process is wrong, etc. After the system problems are eliminated, you can also use SMS or push to urge payment.
Finally, as the overall user conversion indicator, UV transaction conversion rate=number of transaction orders/number of UVs on the page;
As an indicator to assess the overall user value: average UV value=transaction amount/page UV number.
6. Re purchase
Replenishment rate=the number of customers who repeatedly purchase in a period of time/the number of customers who purchase in a period of time. This indicator requires us to analyze data from the horizontal time dimension.
A mature shopping website whose regular users contribute about 60% - 70% of the total sales. Therefore, when we see the traffic funnel transformation model, we should also deepen the hierarchical management of members, and maintain old users with good services and creative activities. If the repurchase rate is low, the following measures can be taken:
① The old members can be sufficiently awakened and activated through SMS push, offline advertising or activities;
② If more resources have been invested in the promotion of new products recently, which has led to the increase of new customers and reduced the recovery rate, it is necessary to verify the data of the promotion activities;
③ If it is super low price or super preferential activities, it will also lead to the introduction of a large number of new users, and will also have an impact on the recovery rate.
The above is the whole process of e-commerce ordering process. Of course, many modules are not mentioned, such as intelligent cross recommendation. We only need to understand the logic of data analysis.
4、 How to guide the landing page of funnel model
The analysis of landing pages is often subjective, for example, landing pages should have a sense of picture, prominent advantages, vivid and eye-catching, etc. These words often appear in landing page analysis. A slightly better team will have a landing page for special data analysis. The landing page generally has three purposes: developing users, facilitating transactions, and collecting clues.
The existing landing page analysis is generally a funnel model, and each level of funnel is determined by the page, such as landing page → purchase page → order page → purchase. However, such process analysis often leads us to deviate. For example, the conversion rate from landing page to purchase page is low, so many entries will be added to the landing page to induce users to enter the purchase page. The final result of such a revision is often that the conversion rate of this step has been improved, but then the conversion rate has declined, and the overall conversion rate has not improved significantly, even bothering users and causing the overall conversion rate to decline.
The purpose of our analysis of funnel transformation is to improve the final transformation, not the transformation at all levels. If the user does not have a genuine purchase intention, no matter how high the previous conversion rate is, the final payment phase still needs to rely on the user's actual purchase intention to reach a transaction. Therefore, the purpose of our revision is actually to stimulate the purchase intention of users.
Regardless of the form of our landing page, consumers themselves need to go through these steps. In fact, the "user perspective" version is the real logic behind the funnel transformation. Let's first sort out the whole purchase process of users on the general landing page.
(1) Put a link at the traffic entrance, which can be in various forms, such as banner, open screen, text, etc. The purpose is to attract users to click to enter the landing page. In this step, the user will generally go through two stages: "attention" and "interest". Attention is to let users pay attention to our advertisements at the entrance. The purpose of arousing interest is to make the users who see it have the intention to click on the advertisement.
The data of this step generally includes: advertisement exposure, click through, click through rate. It can be seen that the step of "arousing interest" is difficult to be reflected through these data, and the click rate data can only show the comprehensive effect of "attracting attention" and "arousing interest".
(2) When the user enters the landing page, the user will first see the header information. We often place the core highlights of activities and product selling points on the front page. If the header is not attractive enough, and the user does not get the value provided by the activity/product, then ordinary users will not decline. In this step, users are still "interested". Therefore, the trigger ratio of the slide down operation on the landing header screen can be regarded as a measure of "interest".
(3) The user starts to read the details provided in the landing page. For users, this stage is about "collecting information". In this stage, there are several data that can be reflected, such as the trigger buried point data of the second screen and the third screen, the proportion of reaching the bottom of the page, the length of page reading, and so on.
However, the higher the data, the better. If the proportion or duration is too high, it may be that the information you want to convey is too much, too miscellaneous, and not enough focused. If the proportion or duration is too low, it means that the information collected by the user is not in line with the previously motivated interest (such as the promotion of "buy one get one free", which turns out to be a 100 yuan product and a 10 yuan gift), or the information collected cannot stimulate the purchase intention. Therefore, the data in this stage is the most complex, and it needs to have the past case The control group was used for reference.
(4) After the previous stage, users begin to consider whether it is worth buying. At this time, users will care about the price. Many landing pages tend to place prices directly on the landing page for users to view. But then we can't find out how many people care about the price from the data.
If the price is hidden and displayed on the purchase page, we can know how many users have reached the stage of "evaluation scheme". Some people may have questions. Isn't this going to increase user click costs? Such changes will increase the proportion of people entering the purchase page and decrease the final success rate, but the overall conversion rate will have little impact.
In addition, if the price is directly placed on the landing page, users will first pay attention to the price, and then see whether the product/service information is worth the price. Even some of the more expensive goods, users are directly scared away. Therefore, if the landing page is designed to display prices directly, it can be modified to separate them. To improve our understanding of user purchase decisions.
(5) Finally, I decided to buy. The user learns the price of the product/service on the purchase page, and will evaluate the transaction. If the price is reasonable, the user will buy. Therefore, the final order conversion rate can be seen as the data performance of this stage.
5、 Tips
1. The most important thing for data analysis is not to tell the business side what happened, but why it happened. Through the analysis of the purchase decision process from the perspective of users, the data of each step can show the problems that users actually encounter, and these problems often determine the final result. In addition, when doing funnel analysis, we should also do some competitive product analysis. We should be aware of the transformation of similar data in the same industry and try to reduce the loss of users.
2. Some funnel analysis involves many links and has a long time period. At this time, the number of funnel links should not exceed 5, and the percentage value of each link in the funnel should not exceed 100 times. Because there are more than five links, there are often multiple key links, so analyzing important issues in a funnel model is prone to confusion. The difference of numerical magnitude is too large, and the fluctuation relationship between numerical values is difficult to be detected, and information is easy to be missed. We can consider whether the length of the funnel can be shortened, whether the order of process nodes can be adjusted, and how to avoid the disconnection of the funnel process.
3. When conducting funnel analysis, it can also be analyzed in combination with the attribution model. According to the actual needs of the product, the credit before achieving the goal (forming the transformation) can be allocated to each transformation node according to the set weight. The significance of attribution model is to find channels that are really beneficial to the development of products at this stage and expand the advantages.
4. The funnel model can also be used in reverse to infer some basic data required for the normal operation of the product. For example, a video website that plays the main role of bullet screen needs 20 people to send bullet screens online at the same time. According to the three-layer funnel model, it can be roughly calculated that the PV of the website's home page must exceed 20000, which can guide us how to find traffic.
For more dry goods, please follow the public account: DataHunter
