I would like to talk about the topic of testing plans. Many friends ask “Why is the account conversion rate low, how can I avoid spending money?” It is difficult to come up with a quick and effective solution to solve the problem, such as it is difficult to explain “Why is there no explosive growth with a CPM plan of more than 3,000?”

For example, the following two plans: The CPM of the above one is about 3300, and the consumption is only about 2400. The plan below has a CPM of 23,000 and a consumption of only 1,564. Then they didn't run.

Two plans with extremely high CPM but low cost

Many times it’s just a matter of testing a plan. If it is measured, it can be put into operation; if it cannot be measured, the cost will be high and the conversion rate will be low. Therefore, “how to test a plan” should be the simplest and most practical methodology. After all, any marketing idea will ultimately have to be implemented in the plan setting to be effective.

Today we are mainly going to talk about a concept, our understanding of “planning”.

If we look at it simply, an account is made up of plans one by one. As long as we test one plan well, we can set up an account, and there will be a typical head effect in the information flow - if you can test one or two plans well, the account will run quite well.

So let’s talk about how to test a plan.

Contents of this article:

01 A story about a plan to save an account

02 How can we test a plan?

03 How to analyze and optimize accounts and plans?

04 Q&A

A story about a plan to save an account

Let’s start with the first point and talk about why it is so important to test a plan.

People always ask me why my account cannot be invested. Some industries that I have never heard of before (I used to mainly work with KA clients and promote online applications) have also started to invest, such as niche industries such as financing training.

A friend introduced him to me and he said, "What's wrong with my account? I have failed to test many plans with my account." I would tell him very calmly that it is the same for everyone and that everyone fails to test many plans. As long as you can figure out a plan during the investment, the account can be invested.

Why is it so important to test out a plan?

We can understand the backend as a carrier that connects advertisers and users. Advertisers tell the media "what kind of demands I have, what kind of users I am looking for", and the way to tell them is that the media labels their users in advance, and then aggregates these labels into buttons one by one in the backend, such as gender, age, behavior, and interest targeting. You can set the budget, bid, and targeting directly in the backend based on what kind of users you want and how many users you want.

Once you have these things set up, it becomes a plan.

So it is equivalent to an order you place with the media. After you place the order with the media, it is entrusted to the media, and the media will help you find people and make placements for you. If you put it that way, the plan doesn't seem that important; it's just an order.

But what is more important is that a plan is viable. After you launch it and click the start button, the plan will be able to generate some data, and you will find that it is like a living person, with life.

When I first had this feeling about the plan, I felt that I had a better understanding of information flow delivery . We can see how a plan will be delivered today and tomorrow. Sometimes the consumption will decrease, and sometimes it will increase. You can see that the trend of delivery and consumption will change day by day. The perception of this plan will help us test a plan.

When it comes to planning, people talk a lot about the concept of model, but because this concept is more technical, even now I am old I still don’t really understand what this model means. To put it simply, it is a wave of similar users. For example, you first find people with similar interests and hobbies. Next, he will look for similar users based on these converted people, which will form a model.

Once this model is correct, even if the conversion target you launch is not your final assessment target, for example, your conversion target is payment, but your final assessment target is ROI, then this plan may continue to be good. This is the significance of the model of a plan, I think so.

There are so many and complicated marketing methodologies, but if you focus on how to measure the plan, it is very simple. No matter who is talking about it, it is all about the options and buttons in the background. Of course, creativity is endless, and everything else besides creativity is very limited.

I am more familiar with Toutiao, so I always use the word "plan" of Toutiao. If you are working for Tencent, the corresponding level is "advertising", and if you are working for Baidu, the corresponding level is "unit".

Next, I will tell you a few stories about testing plans and my experiences in actual deployment.

Once I went on a business trip to Shanghai with a colleague. We were in a terrible situation. Our costs were always very high and our clients immediately stopped investing. I told my leader at the time that the client was about to stop investing and asked if I should report it. My leader said that the company would pay for the business trip and I should die with a clear understanding and figure out why we didn't invest. This account has been invested for almost a month, but there has been no investment. Unexpectedly, on our way to report, a plan suddenly came into being and the account was invested.

There will also be such things. Many advertisers will threaten the agents, "If you can't achieve a certain volume by such and such date, don't invest anymore." There were many times when the volume suddenly increased on the evening of the deadline set by the customer. So later, when my friends were asked to do this by clients, I would tell them that they must persevere until the last moment. Clients said that if it is a certain time of the day, they should not give up within an hour before that time, because the account may start to grow in the first hour.

Another time, we were competing with a client. This client wanted to increase the sales volume at that time, and the competition was quite intense. A new agent made 1 million in one day. We were all stunned and wondered how this agent could be so powerful. Later I learned that he had only one plan running. This was the only plan in his entire account, and this plan could generate 1 million in sales per day.

Have any classmates said that this is metaphysics? If you say it is metaphysics, then there is no way to derive a solution to this problem. We think in the direction of deriving solutions. Some plans can be run in large quantities. You haven't tested such a plan yet, but if you continue testing, you may be able to test such a plan. Think in the direction of solving the problem.

From these stories, we can see some characteristics of information flow advertising. Most plans do not produce large volumes, and this is a normal phenomenon. It is the same when you invest and when I invest.

Once a plan is tested, some plans may take a long time to run, while others may take a short time to run. However, we can assume with a high probability that if one plan can be implemented, you can breathe a sigh of relief, as it can support the account.

So if the account does not grow in volume, I will be more calm now, knowing that it is normal. Just keep testing until you come up with a plan that works.

We just said that most plans are not scalable, so is there any specific data to support this?

I have investigated some accounts before, at least hundreds of accounts, and found that this is a common phenomenon. The head effect is more obvious. In most cases, only a few plans can be run in the account at the same time, or even just one plan can account for most of the account's consumption. Later, I communicated with other agency friends and media friends, and everyone generally recognized this rule.

Of course, there will be some accounts whose consumption of various plans is relatively even, and there may be more than a dozen plans running at the same time, but from the perspective of most plans, the 2 and 8 laws still exist.

Next, let’s talk about how to test a plan.

How can we figure out a plan?

Let's first look at it from a logical perspective. We can split a plan into several parts. One part is the use of media functions, including targeting, bidding, new functions released by the media, hosting, and the like. The other part is traffic collision. There are some things we can decide during the delivery, such as the use of media functions and the creativity of the latter part, but there are also some things we cannot decide. For example, if a material is included in several plans, and one plan has a particularly good ROI, while the others are not. We have no way to explain this and can only attribute it to different traffic.

So what do we do about things we can't decide? Use a method that does not require much technical content, such as opening multiple accounts and creating multiple plans, to increase your chances of encountering good traffic. The last part is creativity. We can understand it as the use of media functions plus traffic flow to find users. Whether the user can be converted in the end is still determined by creativity, so creativity is a crucial factor in delivery.

What to do after splitting the plan into three parts?

Of course, it is to optimize each link of advertising delivery to increase the probability of successful testing. First of all, you need to be familiar with the functions of the media, especially the new functions. I need to emphasize here that I have actually encountered some cases where the investment was made because a new function was used well. Therefore, it is very necessary to understand the functions of the media. I also suggest that everyone treat the new functions of the media with a more positive attitude.

Next, let’s talk about this. In addition to working hard to increase the probability of a plan being measured, once a plan is able to run, we must also be careful not to hold the plan back. Once the plan is running, just let it run. Apart from giving him a budget, don't do anything else. What I particularly want to say here is that you should not adjust it. This is a general consensus in the information flow. It is easy to adjust the conversion bid plan to death.

So how do we allocate the budget?

First set it to a small budget and see if the effect is acceptable. If it is acceptable, you can increase the budget. If it is not acceptable, you can wait for a while.

(Don’t turn it off immediately, maybe the data is not cached. If it is a deep conversion, users still need some conversion time. For example, if you want to pay for a game, users have to play for a while before they can pay, right?)

If the data is not too bad, it is generally recommended to look at the data of the next day before deciding whether to shut it down. Do not shut down a plan too early. If you close a plan too easily, you will fall into a vicious circle of building and closing plans, building and closing plans, and in the end you will find that you have spent a lot of money, but it is still difficult to test the plan.

What are the characteristics of the running volume plan?

Many times, it is a plan with explosive volume. You will find that it will quickly hit the budget and spend money very quickly. Such a plan is more likely to become a pillar plan in your account.

In the first part, we talked about how important it is to test a plan. In the second part, we talked about how to test a plan. Both emphasized the concept of "plan". Next, let's look at how to use this concept to analyze accounts.

How to analyze and optimize accounts and plans

This is my own invention. The plans in the account can be divided into 1, 2, and 3 categories. The first category is the Genius Plan, which is often the pillar plan in the account. It is the kind of plan that we just talked about that performs very well in the beginning. It is not necessarily because you did something right. It may be that you used a certain media function well, or you hit good traffic, or it may be because of creativity. No matter what the reason is, it performs relatively well anyway.

There will be few such programs, but their effects will be great. Our operation is very simple, and we must have put some budget in it, right?

More plans are of the second and third categories. Generally speaking, all plans can be considered as second-category plans. They either spend money or have relatively low costs (high ROI). If one of consumption and cost (ROI) is better and the other has some problems, then we encounter and solve the problem. If it consumes slowly, then you should promote consumption, right? The way to promote consumption is either to increase the bid, or to release targeted budget, or to wait. It may not spend money now, but it will still spend money in a day or two.

If the cost is high, we don’t have any particularly good solutions. We usually say that you can narrow the targeting. From my experience, if I narrow my targeting, the cost generally doesn't go down. A more direct way is to either lower your bid or wait two days for the system to help you optimize your costs.

For the second category plans, what we can do is relatively limited. If after doing these operations, the effect is still not good, then you can classify the plan into three categories, and you can give up this type of plan.

There is another benefit to dividing the plans into 123 categories, which makes it easier for you to accept why some plans do not achieve sales. No one has all Category 1 plans in his account. Everyone has Category 1, Category 2 and Category 3. If you haven't measured it out, then you are now encountering Category 2 or Category 3 plans. You have not yet encountered a Category 1 plan. What will you do? Keep testing until you come across a Category 1 plan.

Let’s look at a case. This is a question from a reader. He said that as soon as his account increases in volume, he will suffer serious losses. It is used to assess ROI, but if there is no increase in volume, the ROI is quite good. He also found that the effect at night is better than that during the day. His single conversion cost is quite high, about 500, which is an account with high conversion cost.

My analysis is this. First of all, let’s not ask the reason, “Why is the ROI poor when a plan is scaled up” - what reasons can you see? What you can see is the data analysis of the account, and you can see data on dimensions such as gender, age, and region, but do these things really reveal anything? Do you know why the effect is poor?

It is difficult to draw conclusions from this kind of general user attribute data. In most cases, the data fluctuates, and it is difficult to come to an overwhelming conclusion. So we don’t ask the reason, let’s look at the simple thing and look at the account settings. You say that if you release the volume you will suffer losses, but if you don’t release the volume it’s okay, so you just don’t release the volume. Look at each plan one by one. For example, when a plan costs 1,000 yuan, the effect is okay. But if you raise its budget to 3,000 and it gets worse, then keep its budget at 1,000. Then we will create a new plan, and we will definitely be able to measure and find a plan with good ROI in terms of volume, but these plans are not that yet.

Why do I use this plan as a case study? Because I was a little surprised when I read this case. I thought everyone analyzed plans item by item in this way, but this friend said that he used to look at the data of the entire account, and not one plan at a time. I realized that my idea was wrong and it was necessary to talk about this thing.

"Analyzing the plan item by item" is the basis for us to manage our account and the first lesson in investment. An account is composed of plans, and plans are independent of each other. Although we also say that plans will influence each other because they are in the same account and will be affected by the model, we also say that plans will compete with each other, but whether they will help each other or compete with each other is something that is difficult to explain, and you don’t know whether there is more help or competition.

So what can be explained clearly? Data can definitely explain it clearly, so don’t worry about those things that can’t be explained, just look at the data. If the cost (ROI) is good, then increase the volume. If the cost (ROI) is not good, then control the budget or even shut it down. It will be easier to think about the placement this way.

I have summarized a mnemonic before: "from big to small, from many to less, from one day to several days."

What does "from big to small" mean?

What this means is that when an account is not performing well, you will not be able to see anything if you just stare at the account. You must analyze it from the big units to the small units. The small unit in the account is the "plan". When you want to see how an account is performing, you have to look at the plan in the account.

What do you think of so many plans?

From the perspective of "from most to least", from most to least refers to consumption. You should sort the consumption in descending order and start from the plan with the highest consumption. Why start with the plan that consumes the most, why not start with the plan that consumes the best and the least? Because plans with high consumption have a greater impact on account performance.

Now that we have found the plan with the highest consumption, how should we analyze it next?

Next, we need to look at the data of this plan in recent days, that is, “from one day to several days”. Click on this plan to see the data from the beginning of its launch to the present. You will have an idea of ​​whether its consumption today has decreased or increased compared to yesterday. If it has decreased, you need to be alert and see if you need to do anything to stimulate its consumption. For example, check if there is room for targeted investment, or see if the budget has been reached and whether there is room for bid increase.

If you want more details, you can also check the operation log. If the consumption of a plan suddenly drops, you should first check whether you have performed any operations at this point in time and check the operation log first. If it is caused by a certain operation, then you can adjust it back. For example, if you lower the bid and then find that the consumption has decreased, then you can see if there is room for cost and whether you can raise the bid again. If there is no operation and consumption still decreases, there is nothing we can do but to use those ideas to promote consumption.

Okay, to summarize, I have talked about three parts today. The central idea is that it is very important to test a plan. The effect of a plan can make a big difference.

So how can we test out a plan? You can focus on these three aspects: learning media functions, building multiple plans to attract traffic, and optimizing creativity. Another point is more passive - once a plan is able to run in volume, we must be careful not to hold it back, and other than setting aside the budget, try not to change anything else.

Finally, when analyzing the account, you need to analyze it plan by plan. You cannot look at the entire account together. You need to look at it plan by plan to find problems.

Q&A

Q1. How should a pitcher test a new account? How many plans should be built at the beginning? Should we release the directional running and run the headlines of food category?

This question is rather general, so I will give you a more common idea directly. Generally, for one product, you can open five accounts first, invest in two accounts first, and leave the remaining three empty. For each of these two accounts, create 10 plans a day. You can open these 10 plans in three time periods, with three plans in each time period of morning, noon, and evening.

Because the targeting of Toutiao is now developing in a more and more intelligent direction, from my personal experience, the measurable space for targeting is not particularly large. The main areas where we can make efforts are behavioral interest targeting and crowd coverage. Generally speaking, there is nothing wrong with directly launching targeted investments. If you study it in more detail, you can look at the groups of people related to food, such as the food account on Douyin.

Q2. Is it the same to put 100 plans in one group or to put them in two groups with 50 plans each? Or would it be better to have more groups?

Putting 100 running plans in one group like you is probably quite extreme. Fundamentally speaking, I don't quite understand the difference between one group and different groups. It is said that the plans in a group will compete with each other. I also feel that it is not easy for a group to have two running plans at the same time. Generally, there is only one running plan in a group.

My daily operation is to have about three or five plans in a group and create multiple groups. There is nothing wrong with the way I operate it. If you don’t quite understand it, you can follow my method. If there is a more specific explanation, you can change it.

Q3. Are there any tips for advertising?

If you look at the articles I wrote before, each article contains a little trick for delivery, such as whether there will be delays in conversion data and what is joint debugging? What is the principle behind offering a high price? When is the best time to shut down the plan? Are these considered small tricks for advertising?

Q4. If I am doing a project to improve my academic qualifications, which aspect would be better to test when I first open an account?

Now the main focus should be on testing the creative direction. You must first list all the creative directions you can think of, and also see what types of creative products your competitors are investing in. Determine the creative direction before making the materials. The creative direction is more important.

Q5. Does the test plan involve using the same material at different time periods, or using different materials at the same time period?

Frankly speaking, the rules we draw on when it comes to time periods are generally quite simple. For example, I find that the effects are better in the evening, or at noon, or in the early morning (most of the products I’ve come into contact with say that the effects are better in the evening). This is the general conclusion you can draw, which can lead to action, that is, you need to make more plans at night. Or if the data difference is more obvious, you can skip investing during the day and only invest at night.

If you want to test the effect of the same material at different time periods, I personally don’t think it is particularly necessary because it is difficult to draw a conclusion.

Author: AiKe.com

Source: AiKe.com