How to do data analysis for information flow promotion?

It seems that many entrepreneurs like to talk about conceptual things. For example, the Internet + in the past two years, big data later, and the recent blockchain...

However, among the people I see who talk a lot about Internet+, big data, and blockchain, few really understand what these things are, how they work, and how to achieve the "magical" effects that the world sees.

When people don't understand something, it's easy to deify or demonize it .

For example, this advertiser’s complaint:

"I sell tea and use this data to advertise, and I get pretty good results. The cost of a button is less than 2 yuan, and the cost of adding fans is as low as 30 yuan. I'm selling honey recently... Why is the effect not so good? One fan costs more than 200 yuan!"

And this one’s voice:

"Look at this beauty product landing page , it's pretty good. It impressed me. You should also help me make it like this and put it on the market! It will definitely be effective! "

It's like last time you burned incense to the Goddess of Land, your grandson's fever went away. This time you are sick and burn incense to the Goddess of Land again, but why do you get even sicker?

When you don't know enough about big data, all your actions are based on superstition , which is equivalent to burning incense and worshiping Buddha in a temple, but it does not play any substantial role .

So, how does big data "magically" solve problems? Where you can't see, many professionals are operating in ways you don't understand. To them, big data is just a tool .

So, if you are not a professional data analyst , how should you look at big data scientifically?

Idea 1: “Understanding business” is more important than “understanding data”

We all know that data analysis is inseparable from the tracking of key delivery indicators. For example, CPM, CPS, GMV, DAU, etc. These indicators are often closely related to your business (such as customer acquisition cost, average customer unit price, number of customers, etc.).

But many people may not know that the tracking of these indicators also has priorities . The specific situation is closely related to the company itself.

The same indicators mentioned above may be feasible for mature companies with mature business models , because mature companies focus on execution and the business model verification stage has been completed. Companies only need to copy the existing model, follow the steps, and expand the scale to operate smoothly and achieve their business goals. Correspondingly, indicators such as GMV can measure operational performance, and tracking these key indicators can also effectively measure our delivery and operational activities.

But for new projects that are still in the exploratory stage of their start-up, it may be a "vanity" indicator .

Why do I say so? Because in the start-up phase, most companies may not even be able to fully determine their business model. They are still constantly modifying their promotional activities and looking for the right products or target customers. With all these uncertainties, how can they determine indicators?

Even if advertising is necessary at this stage, the purpose is to test and verify the degree of match between the product and the market. In other words, before the funds are exhausted, advertising tests are used to understand whether the current product and business model can allow the company to survive. The important indicators that need to be evaluated at this time are not GMV and DAU, but how many users will continue to use their products after the advertising is stopped , what is the repurchase rate, and how many referrals are there ? And other data closely related to the company's survival.

This is what marketing analysts often say: if you don’t understand the business, you won’t be able to correctly interpret the data. For business, what is important is not the data itself, but the effective business insights discovered from the data from a business perspective.

For example, if you have just opened a Taobao store , because there is no natural traffic , you try to use information flow to attract traffic to your store. In this case, the number of customers and turnover of your store every day are not the indicators you should care about most. What you should focus on is : how many customers who placed orders during this period repurchased within a certain period of time, how many customers started to enter the store by searching for your store name or your product brand name, etc. If these data are not clear, you will not know whether it is the traffic that helps you or the competitiveness of your own products that attract customers. The subsequent store activities and marketing strategies cannot be determined, and the subsequent business scale may be limited as a result.

Idea 2: “Understanding human nature” is more important than “understanding data”

The purpose of advertising is to influence consumer decisions. So how do consumers make decisions? We often hear that big data can predict human behavior. So, can we use big data that can predict human behavior to influence consumer decisions and make advertising more effective?

I regret to tell you that the big data platform can accurately collect the behavioral data of each individual, but not make accurate predictions . This is also why Taobao recommends to you the products you just bought, and Toutiao always pushes to you the movies you just watched.

Prediction, on the other hand, is based on behavioral data and uses human intervention or machine learning to build a predictive model to infer their next possible action and intervene accordingly.

In fact, before the emergence of big data technology, businessmen still did business.

A good businessman collects data through observation, interprets the data through thinking, and then uses the data by adjusting sales behavior. Compared with today's big data, it is just not as convenient and does not have such a large base. I won’t list them all here, the well-known brands you can think of at the moment are all good examples in this regard.

Today's big data simply saves us the process of collecting and processing each data one by one. To gain insight into this matter, human brain is still needed so far , and big data technology can only assist rather than lead.

When we do advertising, whether it is information flow or other aspects, we are actually studying people’s purchasing decision-making process, that is, studying the entire process from users “seeing” to “becoming interested”, to “generating a desire to buy”, and finally completing the purchasing behavior. Of course, these studies are based on the assumption that "people are rational." If you have done market research, you will find that real needs are often hidden . For marketers , the most troublesome thing is that decisions are often made due to hidden motives. If you want to uncover hidden motives, you must learn to understand human nature .

We always give an interesting example : a person originally wanted to buy Brand A car after seeing an advertisement, but later bought Brand B car because the model of Brand B car was prettier. Do you think this person's decision makes sense? From a human perspective, yes.

There are two decision-making paths for people. One is the central path , that is, considering the purchase issue is based on sufficient research and thinking, and tends to be "rational", while this temporary decision-making behavior points to the marginal path , that is, the purchase decision comes from certain clues outside the product.

For example, a decision maker's mate-choosing self is activated by car models. Under this secondary self behavior pattern, consumers' decisions tend to be risk-averse and are prone to impulsive consumption - this is one of the more reasonable explanations at the human nature level.

The same is true when we analyze information flow advertising data. The data can only give us signs. We need to look at the reasons behind these signs from multiple perspectives, even from the deepest human perspective, and think about why the audience reacts in this way? What is the underlying logic behind their conversion or non-conversion? Only by repeating this process over and over again can you find the direction of optimization and find customers one step ahead of your competitors.

Idea 3: "Understanding the profession" is more important than "understanding the data"

Data analysis can be divided into two methods: qualitative analysis and quantitative analysis. Among them, qualitative analysis refers to: making judgments on the nature of things, that is, answering "what is it". Another type of analysis is called quantitative analysis, which specifically refers to: making statistics on the quantity of things. What we usually talk about is click-through rate , conversion rate , etc., which refers to quantitative analysis.

Because data such as click-through rate and conversion rate are relatively intuitive, optimizers and bosses will pay more attention to this type of data. The most discussed topics in various information flow exchange groups are quantitative data such as "What is your click-through rate today?" and "What is your conversion rate today?" There is nothing wrong with this kind of communication, but when everyone uses these quantitative indicators to measure the effectiveness of the delivery without distinguishing between industries and delivery stages, you need to pay attention :

(1) For an unfamiliar project, there is no past reference data to refer to. When starting from scratch, what you need is qualitative analysis skills based on marketing expertise rather than quantitative analysis skills .

For example, you can analyze the demand scenario on which the product was developed and why the company developed this product. These qualitative analyses will help find possible effective directions for appeal.

(2) AB testing with quantitative analysis is certainly necessary, but over-reliance on it can cause fatal damage to the delivery .

For example, if you want to optimize a landing page, there are dozens or even hundreds of variables in AB testing (layout, selling point copy , emotional arousal, color matching, product combination, pricing, gifts, etc.). If each one needs to be AB tested, then when can it be put online? (The boss is waiting for you to make money!)

At this time, we need to enable professional marketing analysis first, rely on professional judgment to select 3-5 versions for testing, and then combine the delivery data analysis to choose which one is better for large-scale delivery.

(3) When there is a discrepancy between the conclusions of data analysis and one’s own professional judgment, in most cases, one should listen to the professional judgment rather than the conclusions of data analysis.

For example, an advertiser came to consult about a BMW car advertisement he had placed on WeChat Moments . However, users who received the advertisement left comments saying things like “I can’t afford it.” “Isn’t big data said to be very accurate? How could it have found the wrong person?”

First, the data itself is incomplete.

What advertisers consider to be precision is that the platform displays its ads to users who are interested in BMW or cars of the same level as BMW. Of course, theoretically, if the platform captures all user behavior data, it can achieve 100% accuracy, but in reality advertising platforms cannot do so because their data is incomplete.

The reasons for incompleteness are complex. For example, it may be for the protection of user privacy; for example, the data between various platforms are not fully connected, and data islands are bound to exist for a considerable period of time. Like the user comment on the ad above saying that he can't afford it, the most likely reason from a data-level analysis is that the platform did not accurately grasp the user's financial situation. The platform that has the most accurate data on the audience's financial situation may be various banks, or national tax departments, etc. However, whether for privacy protection or data security considerations, these data cannot be opened to various advertising platforms in a short period of time. Therefore, incomplete data is a reality that we marketers must face.

Second, human nature itself is unpredictable.

Even if data silos are connected and the data itself is complete, there is no guarantee that major information flow platforms can accurately identify and predict the potential audience's subjective response to advertising, because: human nature itself is unpredictable.

Behavioral economics believes that people are irrational in many situations. It can even be said that humans are inherently irrational. It's just like when you fall in love with someone. When you love him/her, you don't even know why you love him/her. All you can remember are illogical fragments one after another. But is there any pattern behind this? Of course there is. The logic behind this is the subject of evolutionary research. In a sense, all of our human behaviors are for the survival and reproduction of the species, and human behavioral responses also come from this.

Let’s go back to the user who left a message under the BMW ad saying that he couldn’t afford it: Can they really not afford it? Not necessarily, it's just that they don't think buying a BMW immediately is the top priority for them at the moment. It's not urgent enough for them, so it's not worth spending the money of a BMW to complete. If you give these users a reason , such as:

"You spent 19 yuan to buy a cup of unbranded coffee for your goddess.

He used his BMW, which cost 19 yuan a day, to send his goddess home. ”

Then "buying a BMW with the money for a cup of coffee every day and winning the heart of a beautiful woman" will become an urgent task that needs to be accomplished .

(Historical data picture)

Users will not do the math, nor will they say they can’t afford it .

You should know that when you don’t give them a reasonable reason to buy, they will tell you that they can’t afford it, and they themselves will also think that they don’t buy it because they “can’t afford it”. We can simply summarize this situation as "saying one thing and thinking another."

The situation of "saying one thing and meaning another" often occurs in our marketing research scenarios. Of course, you must first be able to recognize that the consumer is "saying one thing and thinking another" before you can find a way to make him "tell the truth." Offline, you can guide and confirm through methods such as unique behavioral observation. When placing ads online, you need to create copy that can move their hearts based on their true "intentions."

The author of this article @杜江 is compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting!

Product promotion services: APP promotion services Advertising platform Longyou Century