Information flow advertising: Before you fight with the data, think about your purpose

Let me first tell you a real case we received:

Xiao Q placed an investment and franchising advertisement for an educational product on a certain information flow platform. He received a lot of leads in just a few days . He was very happy and felt that the platform was very effective. But the feedback from the customer service call after following up on the clues made him feel a little frustrated because:

More than 50% of the leads were responded to over the phone with "I'm interested, but don't have time, you can add me on WeChat to chat" and then there was no further follow-up. WeChat additions were not accepted and no one answered the phone calls again.

"Are these leads all fake? Are they all fake? If so, what should I do?" He couldn't help but think of the worst. "How can I determine whether the traffic on these platforms is valid? If I find invalid traffic... what should I do?"

This colleague's feelings are not an isolated case.

Many pitchers will have similar feelings when analyzing the delivery data: there are a lot of clicks and inquiries, but very few successful conversions , and then everyone will think of the same thing: Is there traffic fraud?

Regarding everyone’s confusion, let’s first talk about what traffic cheating is.

The industry has different definitions of traffic fraud. For your convenience, we will temporarily divide traffic fraud into three types:

The first is to use machines to simulate the access behavior of real visitors . It mainly uses programs that simulate real-life algorithms to control machine cheating, using different IP addresses, generating different access and delay conditions, and even simulating real-time click behavior .

The second is to hijack real people's visits through Trojan programs or other means . For example, when you are surfing the Internet, a Trojan is implanted in your computer, and you accidentally or are induced to click on some ads. This ad is indeed clicked by you, the real person, but it is actually not the same as the traffic that the advertiser wants .

The third method is to cheat on traffic by organizing real people to increase the volume. There are often special individuals or companies that organize communities and issue order-brushing tasks with detailed operational requirements, such as which keyword to search for, which advertising entrance to click, how many pages to view, how long to stay, what to talk to customer service about, etc. - you can tell from the description that this type of traffic has the highest degree of simulation, not only are the visits real, but the interactions are also real, making it difficult to identify.

But no matter which one, the essence of cheating is to imitate the behavior of real visitors, so our ideas for identifying fake traffic can also be unified, that is:

Using the behavioral characteristics of real visitors as the standard, fake traffic can be identified and verified by comparing and discovering behavioral differences in certain traffic.

For example, on the same platform, we can determine whether there is fake traffic by segmenting the time dimension and comparing the behavioral differences between real visitors and "suspected" visitors , such as page dwell time, average visit depth, click heat map, visitor daily visit time distribution, conversion funnel and other aspects.

Data indicators from other platforms can also be used for reference. Because there are people behind the platform, although the media attributes are different, the attributes of people are consistent, the access behaviors are similar , and they can refer to each other.

For example, if an ad was originally launched on platform B and is now launched on platform A, if the same ad creative is used on both platforms, we can make auxiliary judgments based on the differences in indicators such as the duration of stay of users who click on the creative on the two platforms, the average visit depth, and the click heat map to see if there is fake traffic.

Some large companies will develop special anti-fraud identification models, incorporating all relevant dimensions into a data model to make comprehensive judgments. In essence, this is also based on the above principles, using the behavioral characteristics of real visitors as the standard .

As for how to monitor the entire behavior path of visitors, we need advertising monitoring tools to help us complete it. For example, paid tools include Ptengine (Platinum Analysis), doubleclick, etc.; free tools, such as Google's GA, are sufficient for most advertisers, but the technical requirements for users will be relatively high.

From the above analysis, we can understand that monitoring and analyzing traffic requires financial costs, and at the very least requires overcoming some technical barriers. However, for the majority of small and medium-sized advertisers, considering the advertising budget and other practical factors (for example, the platform may not necessarily open user access data to advertisers), it may not be cost-effective to purchase monitoring tools specifically. So in this case, how should advertisers deal with possible fake traffic?

Our suggestion is not to argue with the general environment .

Start from the purpose of your own project and begin with the end in mind .

Assuming that your purpose of placing information flow ads is to acquire users, then what you really need to care about is the actual customer acquisition cost . In other words, when you find that there are problems with the quality of users and customer acquisition costs brought by the advertisement, what you need to do is to make a better choice between various delivery methods and platforms, rather than spending a lot of time worrying about whether there is traffic fraud .

For example, in the absence of conditions to identify the authenticity of traffic, we can determine whether to continue the current delivery based on the settlement method and platform performance .

Take the example at the beginning of this article:

Let’s first determine whether the current settlement method is appropriate. When doing specific operations, we can first establish a reference system:

Case 1: Assuming that you are currently investing in CPA, and the cost of each lead is 90 yuan, then compare the total cost of acquiring an effective user before using CPA ( note that it is the total cost! That is, the cost of obtaining an ineffective lead is spread over each effective cost ). For example, it is 220 yuan per user. After investing in CPA ( do not consider whether there is cheating ), the total cost of acquiring an effective user is reduced to 190 yuan per user. Then continue to invest in CPA. Even if you suspect that there is a problem with some of the traffic, as long as the total cost is reduced, your investment purpose has been achieved.

The same consideration applies to situation 2: for example, the original total cost was only 150 yuan per piece, but now it is 190 yuan per piece, then you can just switch back to the previous settlement method.

(Note: When doing this analysis, it is necessary to exclude whether the customer acquisition cost itself has an increasing trend and other related factors)

So we can basically get the following formula:

• Case 1: If the total average cost of acquiring a customer is lower than the total average cost of acquiring a customer in the previous campaign, the campaign will continue;
• Case 2: If the total average cost of acquiring customers is higher than the total average cost of acquiring customers in the previous campaign, then switch back to the previous approach.

In other words: The total average cost of acquiring a single customer = the total investment in a certain delivery period ÷ the number of valid leads in that period

Then check if there are any issues with the platform . You can ask yourself based on the results of the delivery, the same information flow platform:

• 1. Is the current customer acquisition cost of this platform relatively ideal?
• 2. Is the quality of acquired customers relatively high?
• 3. Is the efficiency of acquiring customers relatively fast?
• 4. Is there any cooperation between the current platform and other platforms? If so, how much contribution does it make to customer acquisition on other platforms?

If the answers to the above questions are all yes, then continue to invest, and continue to pay attention to the changing trends of the investment data, and make timely adjustments and optimizations.

As for how much water there is in the traffic of the delivery channel itself, we can temporarily regard it as an uncontrollable factor in the delivery process. After all, the purpose of your advertising is not to fight with the platform for fake data, but to complete our own marketing tasks, namely:

Find matching users on the right platform, say words that motivate them to take action, and minimize customer acquisition costs through continuous optimization.