How to do data analysis for information flow promotion? Avoid these 4 pitfalls!

If you work in bidding, how can you grow into a professional bidder without falling into a few pitfalls?

But since we know it is a pitfall, why don’t we avoid it?

As a bidder, you only know that malicious clicks are a big pitfall, but do you know that there are also pitfalls in " data analysis " ?

However, rather than calling it a pit, it is better to call it a misunderstanding.

Doing a good job of data analysis can make our subsequent optimization work more efficient; and avoiding these misunderstandings in data analysis can make our data analysis efficiency more efficient.

So~the main text has begun, please take your seat.

A few days ago, our customer service department and bidding department had a fight again.

Customer Service Department: The traffic you attract is too rubbish. After chatting for a long time, they just say thank you and leave without even leaving a clue. The level of intention is too low. It is a waste of time and effort.

Bidding Department: Customer service is just for chatting. The ultimate goal of bidding promotion is to enable users to have conversations and thus generate purchasing intentions. What’s more, it’s only possible to convert users if there’s no conversation at all; how can we talk about conversion?

One complains about the low traffic intention, while the other thinks it’s good enough to have a conversation and is still picky.

Dear readers, what do you think: who is right and who is wrong between customer service and bidding?

Then we need to ask ourselves a question first: When we analyze data and optimize accounts, is the ultimate goal to increase conversations? Or conversion?

Obviously it’s the effect, so let’s not optimize just for dialogue.

Bidders and customer service complement each other. Optimizing an account is not just about increasing conversations, it’s more about achieving conversions; unintended traffic is just a waste of customer service time and has no benefit for conversions.

Therefore, in bidding promotion, we must be able to clearly define:

• Which words do we want to acquire?

For example, my goal is to develop and screen potential customers through large traffic and numerous conversations with the first-stage keywords, so the customer service staff should not find it troublesome. After all, although the users of the first-stage keywords have a low level of intention and have many questions about the product, the traffic is still huge.

Suppose a certain word in an account suddenly attracts three traffics, and all of them generate conversations.

So, how would you treat this word?

Large volume? Increase price? Rabbit can only say that there is no problem with this operation, but the better solution is: conservatively increase the volume !

 Remember: do not use the data from a certain day as the basis for data analysis .

The data of a certain day has a certain degree of randomness.

For example, if I stand on the third floor and toss a coin, and it gets heads every time I toss it 10 times, can we say: the probability of getting heads when tossing a coin from the third floor is 100%?

Obviously not, because the test base is extremely low and not of reference value. But when we throw it 10,000 times, or even 100 million times, the result is different.

Therefore, data analysis must be based on a large traffic base, rather than making judgments based on data from a single day.

When we saw the second misunderstanding, we said that data analysis must have a certain data base, otherwise it is meaningless. So,

“How long would you choose to conduct data analysis?”

I think most people would choose a month, or 2 weeks. In fact, there is nothing wrong with using a long enough period of time for analysis, but the account is constantly adjusted, and we need to modify its development direction in a timely manner based on the data.

For example, the current bid for a certain keyword is 10 yuan, but its average price is 15, so some people think that there is something wrong with the system.

How can the click price be higher than the bid? But when we search the history operation, we find that the bid for this keyword was 20 a week ago;

Therefore, a cycle of one month is a bit long, two weeks is barely acceptable, and the best is a cycle of one week, which will also make it easier to calculate the effect after the last adjustment.

For bidding, the page's bounce rate and dwell time are important criteria for measuring whether the page is popular with users.

However, a long stay time of users does not mean that the content of this page is more attractive to users. It is possible that users simply forget to close the page.

Therefore, it is not advisable to blindly believe in the page dwelling time.

The ultimate goal of bidding promotion is dialogue and conversion, so the real measure of the perfection of a page is the user's conversation rate.

Therefore, we should not believe in data completely, nor rely on data completely. Data is only a means to help us optimize our accounts.

Usually, in data analysis, we may pay more attention to data analysis processes and techniques and ignore these small aspects.

Author: Little Bunny , authorized to be published by Qinggua Media .

Source: Houchang College (ID: semlianmeng )