Three serious misunderstandings about information flow advertising targeting. Whoever steps on them will suffer. Let’s take a look…

After reading this article, I hope you can improve in the following aspects:

1. User attribute research;

2. Effective insights into user attributes;

3. Small data sampling, big data profiling;

4. Analysis of consumer purchasing power.

I heard a friend in the advertising industry discussing an information flow advertising case:

The owner of a mattress distributor of a well-known brand (hereinafter referred to as Boss A) targeted his information flow ads to users in the city who use the IOS system. In other words, the advertising excluded all Android users. Everyone thinks that this boss's approach is very interesting...

In fact, it sounds a bit confusing at first:

Generally, advertisements that require downloading software will be directed to mobile phone systems. When will buying a mattress also require different systems? Is it a smart mattress? Need to match the system to download the APP? What does the boss think?

After much discussion, we found Boss A himself, who explained that the single-choice placement for Apple users was the result of "research."

"This mattress is more expensive than its peers, and its sales are average. Not long ago, the clerk accidentally discovered that almost all the people who came to the store to consult in the past month used Apple phones, and the six owners who placed orders recently also paid with Apple phones. So I wonder if high-priced mattresses and high-priced phone users are more compatible?"

“Is the delivery effective?”

"After 4 days of advertising, the click-through rate is a bit bad," Boss A said to himself. "It's strange. I spent my advertising money very quickly before, but I didn't spend much this time. I originally had a daily budget of 1,000 yuan, but now I can't even spend 200 yuan. The optimizer said that choosing only Apple users is too narrow. I want to try it for a few more days."

Listening to Boss A's case reminded me of a time when a local scenic spot commissioned a data company to conduct customer research. Survey data shows that users who are interested in this scenic spot are also very interested in Audi cars .

If we follow Boss A's idea, should the scenic spot target its promotional advertisements only to users who are interested in Audi?

No.

The scenic spot made another decision: referring to the user characteristics given by the survey results, the scenic spot’s sense of quality was particularly highlighted in the information flow advertisements promoting scenic spot activities. From parking lot management to internal transportation in the scenic spot, from scenic spot sketches to restaurants, all were very sophisticatedly displayed in the advertising pictures and landing pages .

★ Limited phenomena VS effective insights

As front-line marketers , we all know the importance of marketing research. It is common sense to gain effective insights from research to guide our marketing planning work.

Taking Boss A and the scenic spot as examples, although they are both faced with the same survey data, their levels of insight are different and the advertising decisions they make are also different, resulting in vastly different advertising effects.

After discovering that all consumers had a common feature, Mr. A came to the conclusion without much thought: just target Apple users with advertising. However, judging from the current delivery of information flow ads, it seems that Apple users have not shown any special interest in mattresses as Mr. A expected.

After learning about the connection between tourists and Audi, the scenic spot was not blinded by appearances, but went a step further to explore the intrinsic relationship between tourists in the scenic spot and those who follow Audi cars.

After further analyzing users' demand for high-quality life, a new wave of scenic spot information flow advertising content focuses on style rather than promotion, and successfully attracts the attention of precise users. Periodic passenger flow statistics show that with a slight increase in total passenger flow, the number of various high-end cars visiting the park has increased by more than 4 times month-on-month.

Let’s look at the case of Boss A according to the thinking logic of the scenic spot (assuming that there is no problem with the survey data itself, that is, all the customers who come to the store for consultation are using Apple phones). After discovering that all customers use Apple phones, the insight has just begun. We should consider in depth:

The value of any single phenomenon is limited. What we need to do is to look deeper, look farther, and see through the essence behind the relevant phenomena, gain insights that match user needs, and make effective marketing decisions.

★ Small data sampling VS big data portrait

The premise of our analysis in the previous paragraph is: assuming that there is no problem with Boss A’s survey data itself. What if there is a problem with the survey, that is, not all of the sampled users who come to the store for consultation are Apple users, how should we judge and analyze it?

In fact, like you, I heard of such a coincidence for the first time - almost all the people who came to see the mattresses used Apple phones - how credible is such an observation? Is there something wrong with Boss A’s statistical method, or is there something wrong with consumers’ purchasing power?

If you are careful, you must have noticed that the survey emphasized by Boss A is just a rough statistics of store customers. The sales staff did not ask every consumer what mobile phone they used, but only counted the consumers who held the mobile phone in their hands. So, what kind of phones do consumers who don’t hold their phones in their hands use? There is no way to know.

An important part of research planning is to design a sample that represents and explains the population. Only by selecting the right sample can researchers make accurate estimates of consumers' thoughts and behaviors.

For example, suppose we want to measure the average height of Chinese men. It would not work to select only Yao Ming or only Pan Changjiang, because neither is representative.

Similarly, suppose there are 1,000 consumers entering Boss A's store, half of them using Android phones and half using Apple phones. Boss A randomly selects 10 people, and it happens that 9 of these 10 people use Apple phones. He will draw the one-sided conclusion that all customers use Apple phones, while the actual situation may very well be the opposite. If the samples are selected incorrectly, there is no point in discussing the research results and making marketing decisions.

Let me ask you a question first: On a hot afternoon, the glass between the iron bars at the entrance of a coffee shop exploded. Do you think that the glass exploded because it couldn't withstand the heat? Or did the iron fence expand due to heat and squeeze the glass, causing it to shatter?

Needless to say, you probably haven't even considered the second possibility.

Why do you take it for granted that the first possibility is the one without even listening to the question in full, even though the probability of the second possibility is very high? Because you are disturbed by your self-consciousness.

Due to the limitations of memory or knowledge, we will unconsciously use the information we are familiar with or can imagine to analyze and judge when making predictions and decisions, just like when we see reports of a plane crash, we will subconsciously remind ourselves not to take a plane, even though in reality the probability of a plane crash is the smallest among all means of transportation .

Behavioral economists Kahneman and Tversky call this cognitive phenomenon the availability bias.

We do not rule out the possibility that Boss A may have the same cognitive error.

For example, Boss A himself and the sales staff in his store all use Apple phones. When he sees a consumer coming into the store to inquire, he will be particularly sensitive and will unconsciously implant it into his memory. When he recalls it again after a period of time, he will feel that all the customers who came here seem to be using Apple phones. The high price of mattresses is information that has long been implanted in their brains, so he naturally connects Apple users with the purchasing power of high-priced mattresses.

In fact, the information we remember can only be used as part of an evaluation decision.

When making predictions and decisions, we must consider a large amount of relevant information, which also has an important impact on correct evaluation and decision-making. The practices of scenic spots are a good interpretation of this.

For some advertisers with smaller businesses, it is almost impossible to find a marketing research company specifically. Many of them can only use some simple research methods to collect information like Boss A, and the reliability of the information cannot be evaluated.

Instead of doing research based on our own feelings, we might as well use the big data platform.

Taking the advertising backend of Toutiao as an example , the targeting method it displays is itself a research framework.

For some local advertisements, except for the delivery area, it is not recommended to do targeting in other aspects at the beginning. After a period of delivery, you can analyze the background data to determine:

What advertising mainly does is to obtain big data analysis and combine it with reality to determine what kind of decision-making power these people who are interested in you have, and then make corresponding marketing copy according to the corresponding power to attract users to take action.

The most important task of a marketer is to discover and meet needs from various phenomena, rather than creating needs out of nothing based on feelings. If you have encountered interesting "insights" like Boss A in the marketing process, please leave a message to discuss with us.