Douyin live streaming sales - the secret of data

All apps have their own algorithms, including Zhihu's like rate, comment rate, and collections, Meituan's praise rate, collections, Taobao's praise rate, collections, and comments, and the same is true for Douyin. So what secrets are hidden in this data?

Let’s take a look at the meaning of our data indicators and how to convert between indicators. You can see the main data viewing points of our data compass and live broadcast screen.

1. PV: defined as the number of visits. The number of times a user comes to your site is counted, and this is cumulative.

2. UV: defined as unique visits. No matter how many times a user comes to your home, it is counted as one visit (within 24 hours);

3. Conversion rate, which generally refers to the ratio of the number of people entering the live broadcast room to the number of people placing orders, is an indicator to measure the “people, goods and venue”;

4. Exposure-conversion rate, the number of direct transactions/number of product exposures, measures the quality of short video materials and live broadcast rooms;

5. Click-to-conversion rate: number of people who directly make transactions/number of clicks, which measures the quality of product arrangement and product combination, and the overall strategy of the live broadcast room;

6. UV value: GMV of transaction amount/PV of live broadcast, mainly to see the average order value of the live broadcast room. The higher this value is, the better the live broadcast is, and the more the platform algorithm likes it, which determines the level of traffic coming in;

7. ROI: The amount invested/the amount sold is used to calculate your profit value;

8. GPM: UV value*1000=field views/GMV*1000, GPM=1000*GMV of a single product/PV of the live broadcast room of the product;

9. Click-through rate: generally refers to the click-through rate of a product, the number of people entering the live broadcast room/the number of people clicking on the product;

10. Stay/average viewing time: the average stay time of the total number of viewers in the broadcast room (the longer the stay time, the more attractive the content and rhythm)

11. Live broadcast natural traffic: the number of people/total number of people who enter through live broadcast recommendation or live broadcast square;

12. Other traffic: the number of people/total number of people who entered the live broadcast room through sharing, guess you like, etc.;

The important data and formulas are all out, so where do we look at the data? Currently, there are several key points for looking at data scientifically: the data compass in the backend of Doudian, a conversion funnel, a large data screen, and third-party data platforms, such as Chanmama and Douchacha (which is mostly used to disassemble and diagnose other people’s live broadcast rooms).

The data consultant in the background updates the real-time traffic situation of each game in real time. It can not only be used to split the traffic sources after the broadcast, but also to check the traffic during the live broadcast.

Main points:

The overall dashboard is divided into traffic data, user data, interaction data, and transaction data. The traffic indicators, interaction indicators, and transaction indicators we have been talking about come from here. Our site views, click-through rates, UV values, conversion rates, etc. are all calculated using formulas based on the values ​​in the data advisor.

For example, you can intuitively see the number of people in the entire event - the number of views, the number of orders sold - the order volume, the total sales amount - the GMV, etc.

Product click-through rate - more used to explore the potential and decline of our products, mainly used for iterative updates and testing of potential hits;

Conversion rate - helps us see the popularity of products, the status of anchors, and the quality of strategic rhetoric;

UV value - my understanding is that it affects the overall situation of the account. The following formula is derived: UV value = GMV/PV of the live broadcast room, GMV = average customer unit price * order volume * conversion rate. UV determines your traffic level (I think it is because the platform earns more);

The reasoning is: cater to the behavioral logic of consumers - cater to the platform's algorithm, don't think about a quick hit, and focus on healthy and sustainable growth.

For example, if we need to adjust our average order value, we will then adjust and test the sales pitch, anchors, scenes, and supply chain. This is based on the user's psychology and behavior. The length of time the user stays and the interaction will determine the corresponding algorithm, which means continuous push and then improved conversions. The more sales we make, the more money the platform earns, and we earn more accordingly. This is the link for the sustainable development of a live broadcast room.

However, live streaming is not static. Mature accounts basically rely on refined operations, starting with improving the ability to read data, studying the data of a live broadcast, then moving on to data for a week, a month, and then to data at the minute-by-second level. You can look at the following curve, combined with the screen recording of your own live broadcast room, and then see what happened during the traffic trough, what product was being promoted, and what the sales pitch was. Or you can look at the traffic peaks and the transaction peaks at the minute-by-second level to understand what happened in the live broadcast room, what product was promoted at the time, and whether it can be replicated. Therefore, testing is also needed. Perhaps the explosion of orders is hidden in these tiny things.

The third is the traffic source, product ranking and user portrait, investment flow, natural traffic, and traffic from other ports are all here. The main thing is to look at the second picture, the number of clicks and click-through rate of each product. This is the trace left by potential hot products, especially those that do not have much explanation and have a high click-through rate. You can return to the market the next day to continue testing;

Other data can also provide feedback on potential problems or good points in the live broadcast room, such as the number of people waiting to pay, which can be urged on the spot, orders can be kicked off, whether the interactive indicators are very good or very bad in that short period of time, what was done during that period, and the processing and review of details;

The next thing to look at is the peak traffic level. Do we place orders during the peak traffic level? Does our strategy reach the optimal solution during the peak traffic level? Because placing orders during the peak traffic level can instantly generate a large number of orders, which can instantly increase our GPM. You can understand this by looking at the formula above.

The fourth is the conversion funnel

The conversion funnel composed of 5 data models, from the number of people entering the live broadcast room to the number of transactions, reflects the path of a normal user from entering the live broadcast room to placing an order. The main analysis is the loss of fans

The first and second funnel models focus on optimizing the accuracy of the fans in the live broadcast room and the acceptance of the content in the live broadcast room for fans;

The second and third funnel models focus on optimizing the sales pitch and guiding customers to the shopping cart.

The third and fourth funnel models focus on the anchor and team's sales promotion skills, as well as the grasp of the rhythm of the whole scene;

The last funnel model is the last step of holding orders, the smoothness of the previous process, the execution of the strategy, and most importantly, the release of the live broadcast room to that critical point;

Here are the highlights of the data main screen:

One is the details of the traffic source. The main focus is on natural recommendations - feed and live broadcast square, the proportion of recommendations on the video side (the majority after short videos become popular), and then the proportion of paid recommendations.

In the second half of this year, you will slowly discover that the weight of short videos in the entire ecosystem has increased, because from the algorithm point of view, interests and content are throughout the entire platform. This is the positioning of the platform, and short video content is in line with the ecosystem. First, once the content racing results are good and it explodes, the traffic will increase. Second, the traffic of short videos is the most accurate, and the link from fans to the live broadcast room to place orders is particularly short. Third, the traffic of short videos is long-tail, and it will not be OK on the first day and gone on the second day, so please pay attention to short videos starting from the second half of the year.

The first few games with natural traffic are basically very small, because the algorithm is still learning the recommendation model. From 10%-50%, hundreds to thousands of views, it is normal. It is fine as long as it can increase step by step. If there is not much change after 10 days or half a month, and there is no natural recommendation, try another account.

Let's talk about the user portraits of viewers and transaction users. The user portrait of viewers is strategy-oriented. Generally speaking, it will get closer and closer to the target audience of the product. Otherwise, the label is inaccurate. It is normal that the first few shows are inaccurate. You will need to study your fans from here later. The transaction user portrait corresponds more to your own transaction audience label. It can only be said to be relatively accurate data, because sometimes people who do not look like your user audience will also place an order. For example, when you sell lipstick, some men buy it as a gift for their girlfriends, and when you sell men's shoes, some female users buy them as a gift for their husbands and family members. These can be tested when doing Qianchuan's crowd targeting and Leica. It can be regarded as providing some ideas.

Regarding the 5-minute traffic trend chart in the upper right corner when the broadcast starts, there will be a number of people leaving and entering. You can roughly estimate the traffic situation of your live broadcast room in half an hour or an hour. For example, 100 people enter and 80 people leave (5 minutes), then the traffic in one hour is 10,000 people and 8,000 people leave. Then compare it with the audience of your previous few games, which is mostly used to consider whether to stop the broadcast or not.

In fact, after analyzing some of the data in this issue, we finally found that most of them are suitable for post-broadcast review, exploring the traffic codes from the data, and then refining operations, testing, and optimization.

Okay. Thank you for watching. Finally, I’ll attach a data review table for everyone. Remember to record the situation of each game.

Author: Deer holding flower

Source: Deer holding flower