Only by placing accurate advertisements in this way can customers who really pay the bills see them.

In order to obtain the greatest benefits with the least advertising cost, many advertisers choose "precision delivery" when placing advertisements, that is, "precisely" placing advertisements in front of the people who are most likely to buy. However, is the so-called “precision delivery” really accurate? I'm afraid many advertisers will be disappointed.

Every advertiser has this hope: to get the greatest benefit with the least advertising cost!

How to get the maximum benefit with the minimum investment? Advertisers are keeping a close eye on “precision delivery”.

As the name suggests, precision delivery means “precisely” placing advertisements in front of the people who are most likely to buy.

It would be best if users could click on the ad and place an order directly after seeing it!

With this expectation, once someone tells him: "We can place advertisements more accurately, and we can place advertisements to anyone you want", it is difficult for advertisers not to be tempted.

So, how is this kind of advertising where you can vote for whoever you want to vote for achieved?

A common method is that big data suppliers capture users’ search behaviors, click behaviors, etc. on a daily basis and create labels for users. Advertisers then select the customer data they want and deliver targeted ads.

For example, a friend who placed an advertisement for a vacuum cleaner on Toutiao believed that "people with carpets at home will need a vacuum cleaner to take care of them." Naturally, his users should have the "carpet" label, so he found users in the database who had purchased carpets on JD.com in the past six months and targeted them with the advertisement.

With such “precise” targeting, there should be “precise” users coming to you soon, right?

However, the results were not as ideal as expected.

He couldn't help but ask me: "The backend data shows that the number of clicks is pitifully small, let alone the number of impressions. What went wrong?"

There are many more cases with similar problems than the one above.

Advertisers’ pursuit of accurate users, coupled with the exaggeration of certain organizations, has led to higher and higher expectations for the “precision” of big data, and people’s increasing reliance on it.

It seems that as long as we have big data, advertising will be more accurate and effective.

Unfortunately, the actual data may disappoint them.

According to survey data in my country in recent years, around 2010, the advertising industry achieved big data technology-guided advertising. In the following two years, the proportion of advertising expenditure in GDP increased wildly. After that, people obviously realized the limitations of big data and the growth rate declined, but unfortunately the proportion remains high.

Simply put, with the help of big data precision delivery, the cost-benefit ratio of advertising has been reduced.

You might be wondering, why is this so? Is it that big data is not accurate enough?

Of course not.

You know, the reason why advertisements made by first-class advertisers have a high return on investment is because they grasp the motivations of users.

As for big data, it does capture and record user behaviors accurately and classify them in detail. However, how to understand and utilize this big data still requires the control of professional advertisers.

In other words, big data can allow advertisers who understand human nature to make more accurate judgments, but to ordinary people, it is just a pair of numbers.

So, as an ordinary advertiser, how should we deal with these big data without professional training?

Today we will talk about how we can make good use of big data to make our delivery more precise.

1. “Effective tags” and “Associated tags”

Like my friend’s vacuum cleaner launch case just now, the reason he chose the “carpet” label is: one of the usage scenarios of the vacuum cleaner is carpet.

In his opinion, if you buy a carpet, you must take care of it; if you need to take care of it, you will need tools ; a vacuum cleaner is a tool that can help users save trouble, so users will need it, and he even specially created an advertising plan targeting users who have a history of carpet purchases.

Do you think it’s reasonable? If the data cannot be put into use, does that mean the big data is not good?

Actually, it is not.

"People who buy carpets need to buy a vacuum cleaner to clean them," this is the perception of merchants. In fact, carpets are not easy to clean, and it is best to use a vacuum cleaner.

Unfortunately, this is not what users think.

Let’s look at the merchants’ perception: “Carpets need to be used with a vacuum cleaner to be easy to clean.” A person who has a carpet may want to buy a vacuum cleaner, but his motivation for buying a vacuum cleaner is not because he bought a carpet, but because it is not easy to clean.

Therefore, the correct solution is to look for users who are truly tired of cleaning the house.

Let’s guess who will be the users of the carpets?

Young people who have just rented a house, newlyweds who have moved into a new home, or even those who are sitting at home and browsing shopping apps may have a sudden inspiration...and a carpet is less than two square meters in area. If it is difficult to clean, are these impulsive users more likely to choose to roll it up and not use it, or to buy a vacuum cleaner worth more than 2,000 yuan?

Most advertisers are accustomed to using a single, intuitively visible attribute to mechanically group certain similar users in the market together, misinterpreting the correlation between keywords and user behavior as a causal relationship.

Just like when ice cream sales increase, the number of drownings increases year-on-year, it does not mean that the hot sales of ice cream cause drownings. Rather, it is the hot weather that has led to hot sales of ice cream and an increase in the number of people participating in water sports. As the base number increases, the number of drownings will inevitably increase proportionally.

So in the matter of “the carpet needs to be vacuumed to be clean”, “carpet” and “vacuum cleaner” are in a correlation relationship, while “vacuum cleaner” and “more convenient and cleaner cleaning work” are in a causal relationship.

Looking back, if we observe those among our friends who buy high-priced small appliances and are happy about it, they are often new mothers and housewives.

Because they need cleaning tools that can really help to reduce their workload.

Therefore, keywords should be captured for people who "want cleaning to become easier" rather than people who "want to buy carpets".

It can be seen that clarifying the relationship between keywords and products is a task we must do.

2. Prioritize users who send consumption signals

Detective Edmond Lockard believed that when people perform a certain behavior, they always come into contact with and exchange with various substances.

Correspondingly, in the era of big data, every action of a user will leave data in the places he touches.

These data are consumption signals. Analyzing and organizing these signals can help us detect user status and predict user behavior.

Let’s take the vacuum cleaner example again. When you realize the relationship between carpets and vacuum cleaners, you don’t need to immediately launch the keyword “carpet”, but explore the relationship between carpets and cleaning actions.

Let’s first talk about why the “carpet” keyword is ineffective:

Whether in a physical store or an e-commerce platform, you will find:

1. Most users buy carpets because they look good and they buy them to decorate their homes. At this time, they don’t think too much about cleaning. Even if they have some concerns, the sales staff will dismiss them with a few words, giving users a feeling that it is very easy and simple to take care of.
2. If you fully realize at the time of purchase that it will be troublesome to take care of in the future, you probably won't buy it;
3. After you buy it and use it for a while, you find that the carpet is easy to get dirty and difficult to clean, but you have to clean it. This is often caused by some factors.

These factors may trigger the behavior of "buying a vacuum cleaner" or the behavior of "throwing away the carpet".

We don’t have data to prove that “when users find carpets difficult to clean, they will choose to buy vacuum cleaners”, so the keyword “carpet” is obviously not “accurate”.

Let's talk about the correct state of capturing and analyzing signals:

Where are the users who bought the vacuum cleaner? The easiest place to find them is probably on the review page for vacuum cleaner products on e-commerce platforms. If you carefully read the after-sales reviews of some popular vacuum cleaners on e-commerce platforms, you will find that the following keywords appear very frequently:

1. Human hair;
2. Snack crumbs;
3. Cat/dog hair.

Combining this information, we can conclude that what motivates users to buy is not the carpet, but the small, difficult-to-clean garbage on the floor of the home - when the optimizer captures this signal, he or she can make a corresponding delivery strategy.

For example, some of the users that vacuum cleaners are looking for are young families who always buy snacks, families with children, families with cats and dogs, and even families whose home decoration style is mainly light-colored (it is easier to find fallen garbage on light-colored floors than on dark-colored floors).

Correspondingly, ads can be targeted to users who have just started buying pet food on JD.com, or users who have a history of buying light-colored decorative building materials in a short period of time, or users who frequently buy children's toys or snacks... and so on.

3. Try to stay on the same wavelength as motivated users

After understanding the user's consumption motivation, does it mean that as long as we push product information to him, we will be able to sell the product?

Not necessarily.

Of course, if you place an ad at this time, the effect will be much better than using related words such as "carpet" on the spur of the moment. But this is not enough, you also need to further consider whether your product level is at the same level as the user's consumption level and habits.

For example, if we sell a vacuum cleaner that costs RMB2000+, then the corresponding ones are:

1. What consumption level of users will become our target users ?
2. What are their requirements for this type of product and for life?
3. For the same vacuum cleaner, should we recommend the latest model to users? Or should I recommend the classic model that is currently on sale?
4. …

This requires data collection and analysis.

In most cases, brands understand data collection as collecting customer static data and transaction data:

For example, static data is a static data file of a customer, such as background files such as name, address, contact number, and annual income;

Transaction data is fluid, such as the attributes and specifications of the transaction products, activity participation data, customer service records, and product characteristic codes, which are used to record information related to the transaction products themselves, such as price, origin, functional description, word-of-mouth data, etc.

These data are of course very important and are an important part of our user portrait .

However, it is easier to find precise keywords in the products that customers are trading. It is easier for us to collect information that is more valuable for analysis on products consumed by similar users during the same period.

This friend who was promoting the vacuum cleaner later found out when he retrieved the platform data: many of the users who bought vacuum cleaners on the same platform also bought consumer upgrade appliances such as haze removers, air purifiers, and dishwashers. By analyzing these related products, we can predict the purchasing power, model selection and other requirements of the target users of vacuum cleaners.

When we find that many users who buy vacuum cleaners also buy a certain brand of Type A purifier, we can judge that Type A purifier is the associated brand of this vacuum cleaner.

Assume that the price of Type A purifier is 3,000 yuan, while the average price of other brands of purifiers sold on the same platform is 1,500 yuan, accounting for only 28% of the sales volume of similar products on the platform. In other words, Type A purifiers are high-priced products on this platform.

It is a related product to the vacuum cleaner we want to promote , which means that our vacuum cleaner user group is the same as that of Type A purifier. The target users are those with high spending power and high pursuit of quality of life.

Therefore, when we launch advertising, we should focus on users with high consumption power and high living standards. The creativity and landing page style need to match the characteristics of such users.

The author of this article @婷婷 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 Games