A survey on the effectiveness of Toutiao CPA shows that 90% of people ignore this issue!

A friend complains every time he meets her:

"I'm really upset lately! The CPA I did for Toutiao for my client was stable and the cost was controllable..."

“That’s great!”

"That's the problem. Many of the applicants have no interest in the application, or they just don't answer the phone. I've spent money, but not a single deal has been made! The client is furious. What if he beats me up and makes me disabled? Does it count as a work injury..."

His words reminded me of Mr. Song Xing’s question about Toutiao’s traffic last week and Toutiao’s official answer. Let’s briefly review the whole incident:

On July 18 , Mr. Song Xing posted an article in his official account (Internet Data Officer): "Is Toutiao's advertising campaign ineffective? It may not be your problem! ", the article pointed out the strange data phenomena that appeared in the Toutiao account of a friend of Teacher Song, such as excessive concentration of click hotspots, sudden traffic peaks in the early morning , etc.

The original text is roughly as follows:

Preface: This article only describes the problems I encountered and does not involve any value judgments.

Recently, a friend of mine in the education field discovered something "unusual" in his company.

Around June 7 of this year, the traffic of Toutiao, which had been running well before , suddenly dropped to zero. To be precise, the traffic is still there, but there is no conversion suddenly.

The specific phenomenon is as follows:

My friend’s company puts information flow ads on Toutiao. After clicking, you will be directed to your own landing page , and it is a single-page promotion . Before the problem occurred, the traffic conversion was about 30 per day. After the problem occurred, the traffic conversion was directly reduced to 3, and then became 0. Yes, you read that right, it became 0.

Everyone was shocked and thought it must have been caused by the e-commerce 6.18 promotion - even though there was no evidence to support this. So we will continue to put in ads and make some creative optimizations, hoping that after June 18th, everything will slowly get better. But disappointingly, the traffic has not improved since then.

The data segmentation revealed two strange phenomena:

1. Traffic peaks at 4 a.m.

2. The page click heat map is too concentrated;

3. Android traffic is extremely high.

As shown below:

On July 21 , in response to the questions raised by Mr. Song, Toutiao officially published an article through the official account "Toutiao Marketing Observation" to answer: "What is the effectiveness of Toutiao's advertising ? This article gives you a more comprehensive understanding. The article provides corresponding investigations and explanations for the reasons why the heat map is too concentrated and the traffic peak in the early morning.

1. We chose accelerated delivery , which resulted in concentrated consumption of the budget. Traffic began to rise at 4 a.m. on the same day (in line with the mobile traffic distribution curve) . Backend statistics showed that the click-through rate of the ad was 20% higher than the average, and the ad was highly competitive. The ad budget was quickly consumed before 8 a.m. on the 13th.

2. Ptengine heat map: By dragging the heat map slider, we found that clicks are not limited to these three areas, but also in other areas. Toutiao matches the landing page access IP with the Toutiao click IP based on the advertiser's landing page log, and finds that user visits basically come from Toutiao, and there is no abnormal traffic.

Toutiao also stated that it will continue to cooperate in the investigation and resolve the issues raised by Teacher Song.

On July 22 , Mr. Song wrote a message to thank Toutiao and asked:

1. Heatmap

In the official description of Toutiao, this heat map is as follows:

The default heat map is as follows:

Why is there such a difference? The main ones are here (see the red box):

When the button in the red box is dragged to the left, the heat map will tend to show all clicks in darker colors. In other words, the display of the heat map will reduce the difference in the actual number of clicks at each location.

For example, I drag the slider slightly to the left to the middle position. Please see the screenshot:

Is it a little darker than before?

If I keep dragging, is it a little deeper?

Only when I drag this bar to the far left will the situation described in the headline appear. Please see:

Therefore, this situation of clicks everywhere will only occur when the scroll bar is dragged to the specific position on the far left.

Regarding this issue, I also consulted the third-party tool provider Ptmind about why such a heat map display method was set up. The other party’s answer was that the purpose of doing so is mainly to allow users to know where the click occurred, but dragging it to the far left can no longer show the relative difference in the number of clicks at different positions.

This means that although there are clicks in other places, the difference in the number of clicks is completely incomparable to the number of clicks concentrated in those three places. Therefore, a large number of clicks are concentrated on three points, which is still an anomaly worthy of attention.

2. Early morning traffic peak

According to the general night traffic, around 3 or 4 in the morning is when people sleep the deepest. According to the traffic trend, it should be highest at 0:00 in the morning and drop to the lowest at around 3 or 4:00. This is more in line with the "traffic distribution curve".

In this case, the opposite happened. Traffic started to surge at 3 or 4 o'clock. This is indeed a bit confusing.

However, I would also like to point out that the way Toutiao’s advertising works is not an ordinary way of allocating advertising resources. In theory, it is a traffic allocation based on population and interests with a complex algorithm behind it. Therefore, it is possible that the background algorithm believes that it is very worthwhile to advertise during this time period. So, it feels abnormal, but it can still be very reasonable, it just needs a reasonable explanation. But it seems that "conforming to the mobile traffic distribution curve" is not a very clear explanation.

At the same time, Mr. Song is also very grateful for Toutiao’s hard work and dedication, as well as its honest attitude in solving problems, and will continue to follow up together.

The above is the whole process of the whole thing.

Let’s go back to the question of the friend mentioned at the beginning of the article. Without considering the heat map and the early morning traffic peak, the editor wants to start with the effective data obtained from CPA .

Why do those people sign up but then say they have no intention or simply refuse to answer the phone?

Of course, this is already a level outside the platform, and it seems that it has little to do with the promotion stage. The factors that affect subsequent conversions are nothing more than the following reasons:

1. Customer service follow-up time and frequency;

2. Customer service communication skills;

3. Activity content, including price, cycle, quality, service, etc.;

4. Users make their own choices;

…

The above are all factors that affect follow-up , but one point that is often overlooked is: the source of the data.

So we set out to study the available data.

After investigation, the following two problems were found:

1. Most of the people who have no intention or do not answer the phone call leave their name as Ms. X/Mr. X;

2. Ms. X/Mr. X account for about 10% of the total valid data;

Why do they keep such a name? Out of personal habit?

Is this data ratio normal?

With this question in mind, we continued to investigate the effective CPA data of several other friends and found the same problem:

1. There will be some X ladies/X men more or less, accounting for an average of 10%. Among them, the proportion of accounts with better results is slightly lower, and the proportion of accounts with poorer results is relatively higher;

2. The location of the valid registration phone number is consistent with the background targeting area after inquiry, and there is no abnormality;

3. In addition to those who call themselves Ms. X/Mr. X , there are also a few who call themselves Ms. X/Mr. X ;

4. The common characteristics of these people are that they have no intention, want to think again, or even refuse to answer the phone.

This article does not contain any emotional color, bias or smearing intention. I just want to ask everyone if they have encountered similar experiences , so that we can discuss and research solutions together.