"Using data analysis, accurately positioning users, accurately discovering user needs, and accurately recommending products" is a story that is being blown in many places.
However, just like the children's story "Gudong is Coming", everyone has heard of this Gudong (precision), but what exactly is the accuracy and how the data is accurate, no one has ever explained it clearly. Many students are also very depressed: the leader makes me precise every day, but how can I be precise? ! Today we will explain the system.
1. Completely imprecise scene
There is a classic cross talk called "Universal Cigarettes", which depicts the daily life of salesmen in the 1980s and 1990s: I am a cosmic cigarette. The girl smokes beautiful, the young man smokes handsome, the old man smokes longevity, the child smokes Entrance to university-this is the typical complete imprecision. It seems that everyone is suitable, but in fact, ghosts know that girls and children can't smoke well.
Although many products are essential to people's lives, when it comes to a brand, a product, a certain period of time, and a certain price, no one may buy them. The longer the conversion chain, the greater the possibility of user loss (as shown below). The widespread use of the Internet has exacerbated this problem. In the past, there were only a few commercial streets in a city with a lot of people, and it was easy to rely on flagship stores to monopolize traffic. Nowadays, there are various information channels on the Internet, and the information forms are rich and colorful, resulting in lower conversion efficiency. Fortunately, Internet platforms can record data, which is the prerequisite for accurate analysis through data.
Therefore, precision is not an absolute value, but a relative concept. That is, it is more accurate to find users than shouting "the bastard boss and his sister-in-law ran away with his sister-in-law, the whole audience will be 9 yuan", which is called "precision". When referring to precision, it is often accompanied by a discussion about how much the current situation is and how much needs to be optimized. This point will be mentioned later.
Second, improve the accuracy of delivery
The first step in using data to improve accuracy is the distribution channel. Information has to reach the user before it can be converted. There are two common methods in channels: one is to subtract, cut down the channels with a low conversion rate, and save costs; the other is to add, find new channels with high target users and guaranteed conversion rates. Improve efficiency. These two methods are often used in combination in actual work, because multi-channel delivery is the norm. To improve efficiency, you have to look at them one by one and make permutations and combinations (as shown below).
3. Improve content accuracy
The second step is to improve the accuracy of the content, because the information reaches the user. If the content is not good-looking, the user may not click, and then the process will be directly interrupted. At this time, you can do ABtest directly, or tag the content to observe the actual effect of different content, to reverse inference: users actually prefer the XX type. Similarly, design problems can be tested/grouped to observe the effect, which is better than direct data prediction.
Fourth, improve product accuracy
The third step is to improve the accuracy of the product. The basic idea is not to simply push a product, but to make more choices to increase the chance of user conversion. Note, however, that there is a limit to doing more, because users have limited energy. If you push too many models at a time, it will easily cause difficulty in choosing. Therefore, try to find users to buy categories with high potential and focus on pushing a few models.
Five, improve price accuracy
After clarifying the channels, content, and products, the final step is price. The so-called price accuracy means that big data is familiar, and the more loyal users are, the happier they are, so as to maximize profits and sales. This may be what the Internet is best at. Because traditional channels are so blatantly engaging in price discrimination, not to mention that users will rebel, and the Industrial and Commercial Bureau will call the door. However, in the Internet channel, the information between users is not so smooth, the prices of many products are not high, and they are consumed in real time, so they are killed.
6. Steps to achieve precise analysis
It is precisely because there are many links that affect accuracy, so when the actual project is carried out, you must first clear and precise goals, sort out the business process, understand the current data status, and understand what the business can do in the business process. Only in this way can we find the point of force, from simple to difficult sweeping problems, and improve the effect (as shown below).
The basic idea of the work can be summarized as follows: first catch the white rhino, then the yellow toad, and give up the black swan. which is
1. Obvious shortcomings in the conversion process, first improve
2. After solving the big problem, then work out the details from the smallest
3. Sudden, random, one-time problems, just give up
Because from the perspective of results, there are too many influencing factors for one promotion, and it is impossible to expect one analysis to cover the world. Accuracy of data analysis requires process, time, and experience. Deal with the obvious problems first, otherwise there are too many details that can make people addicted, and in the end, the effect will not be improved.
It’s easy for newcomers to do data here to make a problem, that is, don’t look at the scene and just use the model. For example, when many students heard "precision", they were eager to make recommendation algorithms (and still use association rules, because collaborative filtering or calculating distance functions requires a lot of tags, which is too much trouble and the association rules are so simple). No matter the specific business scenario, no matter the price, no matter the content, it is difficult to produce good results. Many students are addicted to: Taobao is XXX. Please, the user volume, data volume, and user stickiness of such a super application like Taobao are not comparable to other platforms, and the Taobao platform does not need to be responsible for a certain SKU of a certain merchant, so it cannot be simply copied. It is important to be down-to-earth and understand your own process, current situation, and business goals.
But in business, there are more problems with precise understanding, such as...
7. Common misunderstandings of precise analysis
It is estimated that many planning, marketing, and operations will say: This feels inaccurate! It's Di, people who don't understand data, the accuracy in their minds is probably like this (as shown below).
In essence, there is no such thing as a brand + a product + a certain period of time + a certain price and everyone will consume it. Even if it is as powerful as WeChat, it only has 1.1 billion daily activities, which only covers less than 80% of the Chinese people. So don't expect your small brands, non-explosive models, and prices that can't do anything to do with Pinduoduo's products.
Therefore, the business side should not attempt to be the shopkeeper in such jobs as precision marketing. It is a better way to improve by sharing marketing plans, investment costs, combat intentions, content creation, and co-creation with data analysis. If you are interested, follow the WeChat public account [Grounding Qi School]. In the next article, we will share a case of precision operation, so stay tuned.
Author: Chen grounded gas, micro-channel public number: down to earth school. A data analyst with ten years of experience and CRM experience in multiple industries.