How to increase the volume of information flow bidding advertising?

For all advertising operations, there are always two huge mountains on their shoulders: cost and volume . But at the end of the day, what is the most important goal? It means increasing volume. After all, you can only make money if you produce in large quantities. Even if the cost is low and there is no consumption, it is just hooliganism.

In 2020, when it comes to the delivery of information flow advertising, operators seem to feel increasingly powerless: Why are the costs exceeded? Why don't you run more? I was running yesterday, why am I not running today? …

Someone once asked me: As products and systems become more and more intelligent, will optimizers become unemployed?

What I want to say is: if you ask this question, you may be unemployed.

My understanding is that with the current level of development of artificial intelligence, it is far from being able to replace human work. The current product tools are more of operational efficiency tools. There are no operations that are eliminated by tools, only operations that are eliminated by the company because they do not know how to use tools.

This article is written based on the Bytedance Engine delivery platform. The text length is 5,600 words and it is expected to take 10 minutes to read. I hope I don’t waste your reading time.

1. Delivery Logic

First, let us understand how each exposed advertisement is delivered in the information flow bidding advertisement . To deeply understand the mechanism of advertising delivery, we need to look at the advertising delivery mechanism from two perspectives.

1. User perspective

As a user of Toutiao, when you open the app and scroll down to browse information in the recommended channel, the multiple pieces of information presented to you will include advertising information. When an ad position is about to be displayed (or filled with an ad) during the sliding process, this display opportunity is an ad request, and for each ad request, there will be N ad creatives competing for the opportunity to show you.

But in the end, only one advertising idea is successfully displayed, and what determines who can be displayed is the logic behind the delivery. We briefly outline this with the following diagram:

We still take the advertising opportunities you browse in the recommended channel of Toutiao APP as an example to briefly explain:

Advertising warehouse: all advertising plans that are "being delivered";

Matching ads: Match ads that match the ad sizes in the recommended channels of the “Toutiao” app;

Frequency control: There is a limit on the number of times each user can see the same type of advertisement in a day;

Dislike: If you click "dislike/not interested" on a certain type of ad multiple times, that type of ad will not be shown to you for a certain period of time;

Estimate: For each ad that meets the above steps, the system estimates the probability of you clicking on the ad (CTR) and the probability of conversion (CVR) after it is displayed to you. A formula is used to calculate a value (ECPM) for each ad, and then the values ​​are sorted from high to low.

Article Mixing: The highest ranked ad will be considered as a normal article by the system. If the content quality is too low, it will not be displayed and a normal article will be displayed in its place.

Placement: If it is not "killed" by "mixed article placement", then this advertisement has defeated multiple competitors and successfully come to you.

2. Operational perspective

As an excellent bidding advertising operator, you need to understand that we serve users rather than customers. Therefore, it is very necessary for operations to have a thorough understanding of how users see an advertisement. Next, let’s look at how your ads are delivered from the perspective of operations and advertising operators.

We use the following figure to briefly summarize:

Take an advertising plan in the "delivery" section of your account as an example: first, the system will filter out ad requests that are not within your targeting range. Then your plan needs to have enough budget to filter out ad requests from users who do not like the ad or have seen it often enough. Only then will your plan begin to enter the stage where the system model estimates the ECPM value. And this is done not just once, but multiple times. After multiple estimates and sorting, the final winner will enter the article mixing stage.

As you can see, you have to go through layers of screening and competition to grab the opportunity to be displayed. In a day, your advertising plan will go through the above process countless times. The number of times it is displayed means the number of successes, which is the display volume of your plan for that day.

In this process, as operations managers, what steps can we take to improve the competitive success rate of your plan?

Targeting and budgeting are the basic account settings you can make. It is almost impossible to influence whether the user clicks "dislike" or has seen enough of the same type of ads. Of course, if your ad creative is too annoying, more users will definitely “dislike” it.

Coming to the estimation and ranking stage, this is the key step for our operations colleagues to truly improve through operational means. Here we need to introduce the key formula:

ECPM = estimated CTR * estimated CVR * bid

Only if the ECPM is high enough will you rank high in the competition and ultimately win. It is not difficult to find out from the formula that the three key indicators that affect the ECPM value are the estimated click-through rate (CTR), the estimated conversion rate (CVR), and the OCPM bid you set.

It should be noted that, combined with the explanation from the user's perspective in the previous article, the model makes an estimate for your advertising plan and for each ad request (that is, the user of the request). The model's estimated values ​​are different for different requests (that is, different users). In other words, if we want to have a large advertising campaign, what we need to do is to make the advertising campaign have a higher estimated value for more people .

2. Volume expansion strategy

Now that we understand the logic of bidding advertising, let’s officially talk about strategies that can increase the amount of advertising. (Note: If you have no experience in operating the Bytedance platform, you may not understand some of the following.)

I don’t know if you have noticed that, up to this point in the reading, the word “optimizer” has only appeared once in the previous text (and it was said by someone else), and all the other words used are “operation” and “operation”. Because I think that a good "optimizer" is not just an account optimizer, but should also have overall control capabilities. Account optimization is only an important part of the job.

Therefore, we need to have the ability to formulate delivery strategies .

So what is strategy? I believe that on the premise of fully understanding the products being launched, fully understanding the needs of customers, and being fully familiar with the systems and users of the delivery platform, knowing what kind of delivery methods and choices can maximize customer needs , this method and choice is the strategy.

Back to our title, I have summarized the following strategies that are conducive to increasing the volume of delivery. Moreover, if many strategies can be combined, the effect will be twice as good with half the effort.

1. Multiple accounts

Don’t limit yourself to one or two accounts. It is recommended to place your ads in more than one account.

The core reason is that the traffic of the giant platform is large enough (tens of billions of requests per day), and it is impossible for each account to cover (have the opportunity to deliver to) all the traffic; and currently, almost all performance bidding is OCPM bidding, that is, bidding directly on the conversion target. The model's ECPM estimate of the advertisement is greatly affected by the account's accumulated data. Therefore, using multiple accounts to deliver can increase the possibility of your product covering more target audiences, and it will not cause one account to have poor performance, thereby limiting the model's understanding of potential target audiences.

The second reason is that multiple accounts can have more delivery methods, such as testing different conversion goals for each account (whether it is "effective customer acquisition" or "activation and payment", or "key behavior"), for example, each account delivers different resource ends. The advantage of this is that the model of each account is more accurate in the same type of event or environment.

2. Multiple plans

The core reason is basically the same as the multi-account strategy. "Plan" is the smallest unit participating in bidding ranking. Therefore, the quality of each plan, or whether it can cultivate a "explosive" plan, is the key.

Here, I want to emphasize a wrong view about multiple plans: planning without thinking. The current platform has very efficient planning tools (diversity replication, MKT API docking, one-stop Zeus, etc.), but the accumulation of large numbers of identical plans is completely unnecessary and even has a negative effect.

The core of multiple plans lies in multiple creativity, as well as different targeting, bidding, etc. If you are really confident in an idea and want to reach more people with it, then it is recommended to combine a multi-account strategy and go for multiple plans.

3. Multi-resource testing

The biddable resources on the massive platform are mainly divided into four ends within the site (Toutiao, TikTok, TikTok Volcano Edition, and Xigua Video), as well as Pangolin Alliance resources. If the client does not have clear restrictions, or the massive platform does not have restrictions, then it is recommended that all resource ends can be tested. Each resource end will basically have a target audience that meets the requirements for the product launch, the only difference is the amount.

As for the method of multi-resource testing, you can directly " optimize ad positions ", or you can build plans for different resource ends , or even test different resource ends for different accounts .

4. Improve the cold start pass rate

Increasing the cold start pass rate of plans within the account as much as possible will help the account model establish an understanding of the target population as early as possible, thereby improving the ECPM estimate of the plan for the target audience.

Speaking of cold start, here is a general definition of a successful cold start: during the cold start phase, if the accumulated number of effective conversions exceeds 20, the plan passes the cold start. The core is the number of conversions, not the cost of conversions, so the key is to launch new plans as quickly as possible and accumulate conversions. As for how to improve the pass rate, we will discuss it below.

5. Make good use of tools

Do you remember what I said at the beginning? Products are becoming more and more intelligent, and there are more and more optimization tools. Where is the value of operations colleagues reflected? That is to make good use of tools and let the products and tools help you achieve the goal of increasing sales. You must not let the intelligent product tools prevent you from finding the way and think that whether you can increase sales is metaphysics and depends entirely on the "volume-increasing symbol". This is not in line with the core socialist values ​​at all.

3. Tool Usage Guide

In the next part, the author summarizes the massive platform product tools that can currently help with volume expansion, as well as the core points of each tool.

1. Delivery Manager

The first one I would like to introduce is the Delivery Manager, which is currently a tool at the forefront of product intelligence. The predecessor of Delivery Manager was called "Hosting Lab". It used to be a tool that required an account to be cleared before it could be used. In December 2019, the product was upgraded to become a full-scale product, and the entrance was integrated into the AD background for easy use.

The core highlight of the Delivery Manager is that you only need to follow a few simple steps to create a project (equivalent to the promotion group in the AD background). The system will automatically create a plan within the project group and dynamically adjust it according to the target cost (i.e. bid) set by the project group. The system will automatically decide whether to shut down each plan and when to create a new plan. Combined with programmatic creative functions, it supports the optimization of up to 50 videos/pictures * 30 titles in a single group, giving full play to the power of machine learning.

Some people would say that this product tool mainly improves the efficiency of planning, but how to improve the ability to increase volume?

If you really think so, then you have not fully understood and used this tool. Through "operations" such as automatically creating and starting and stopping plans, the Delivery Manager is actually constantly trying to find target audiences that meet cost and volume requirements . This is a very efficient and effective way to increase the estimated value of advertising plans for more people. Let's take a closer look.

With this understanding, the key points of using the delivery manager will become clear:

1) Be sure to include more creative ideas (videos, pictures) and titles to give full play to the power of programmatic creativity.

2) Try not to manually intervene in the start and stop of the plan, even though such a function is provided. Try to view the "project" as a whole to observe the data. The start and stop of the plans within the project are just the system's methods to complete the overall requirements of the project.

Finally, pay more attention to the new function upgrades of the delivery manager, such as the recently launched segmented industry scenarios and pangolin scenarios. This product tool will definitely be an important delivery tool in the future and every operations colleague cannot ignore it.

2. Automatic bidding

Automatic bidding is also a product tool that has been developed from applying for whitelisting to full launch. This product is easy to use, and it can be said that it is just a matter of switching on and off , but I still put it in the second recommended position, which shows its ability to increase sales.

As the name suggests, automatic bidding is a product that the system automatically bids. Although operations colleagues are allowed to choose automatic bidding and then manually set an expected cost, there is still an essential difference between this and direct manual bidding.

Let’s take a look at the product positioning of automatic bidding: a bidding tool that helps customers grab volume, comprehensively considers “cost” and “budget utilization rate”, and controls costs through budgeting. Specifically, there are different bidding strategies according to different stages of the plan.

1) Cold start phase: aggressive exploration to improve the cold start pass rate.

2) Mature stage: intelligently explore and find the optimal cost.

To translate, during the cold start phase of a new plan, by increasing bids, the plan can quickly accumulate enough effective conversions and start from a cold start. After entering the mature stage, bids are adjusted intelligently to find the optimal cost.

Note that this is the optimal cost, not the lowest cost. What is the optimal cost? It is the minimum cost that can spend your set budget. The ability to increase volume is evident. However, don’t worry about whether the cost will be high. The system will make a cost estimate based on the plan settings (industry, conversion tracking type, budget settings, etc.), and it will not bid randomly. So for plans using automatic bidding, the cost trend is generally as follows:

Let me summarize two points of experience:

1) For new plans with automatic bidding, do not set the daily budget too high in the initial stage, because the system will judge your running volume needs based on the budget, which will affect the bid range. It can also be observed during background settings that the higher the budget, the higher the estimated cost of the system. Therefore, it is recommended to adopt a budget strategy of "small steps and fast progress", that is, to give a smaller budget in the initial stage, and when the budget is spent to complete the cost, continue to increase the budget, and increase it again after the cost is completed.

2) In the initial stage, you can set certain audience targeting appropriately, circle the core target group, set a higher budget, and use automatic bidding to quickly accumulate conversions without causing the cost to be too high. After a cold start, you can relax the targeting appropriately (using the smart volume-enhancing tool described below) and let the system find the target population.

3. Intelligent volume expansion

Smart volume expansion is a product tool that starts with applying for whitelisting and ends with full volume launch. Its core function is that the system gradually relaxes the targeted settings based on the accumulated characteristics of the converted population while ensuring costs.

Please note that it means relax, not release . The significance of relaxation is that the system is no longer limited to the scope of the population you define, but looks for people with the same characteristics and conversion intentions instead of directly targeting everyone. After understanding this meaning, you will understand why smart volume expansion is recommended as a volume expansion tool.

As for the method of use, it can be said to be very simple. After enabling it, just check the direction that can be relaxed by the system.

4. One-click volume

One-click volume, currently it is still a tool that requires application for whitelisting to use. Its product positioning is: a function specially provided for the planned cold start phase, which increases prices through short-term optimization, quickly gains volume, and improves the cold start pass rate.

One-click scaling is only available when the plan is in the cold start phase. By setting a "startup budget", the system will spend the budget within 6 hours and provide data display during the startup period. A plan only has one chance to be used in a lifetime.

Some students may ask: Is there any difference between this and manually bidding a high price?

Good question. If you haven’t thought about this question, you need to reflect on yourself. My explanation is:

① Everyone knows the compensation policy for cold start. You cannot adjust more than 2 times in one day. However, if you use "one-click start", the number of adjustments will not be counted (but the consumption of one-click start will not be compensated, and the conversion will be counted into the compensation threshold);

② You can understand that one-click volume increase is an opportunity for the system to boost your potential plan, and this boost effect is more powerful than manually bidding a high price. Based on past experience, it is recommended that after a new plan has accumulated about 5 conversions and you find that you are having difficulty increasing the volume, you should use one-click to increase the volume, which will have a better overall effect.

5. Programmatic creative/landing page

I believe everyone is familiar with programmatic creativity . In one plan, multiple videos and pictures are uploaded, combined with multiple title copywriting, to form an exponential number of creative combinations; then the system intelligently allocates display opportunities based on the performance data of each combination to achieve the best effect.

The principle of programmatic landing pages is similar to that of programmatic creativity. Through the use of multiple components/multiple landing pages, multiple different landing pages can be delivered within the same plan. The system then intelligently allocates display opportunities based on the performance data of each landing page to achieve the best results.

PS: Currently, the programmatic landing page group is supported to add the "own landing page" function.

After briefly introducing their core functions, are there any students who would like to ask: Isn’t this a tool to improve efficiency and effectiveness? How can the volume be increased? If you are really in doubt, read the article again from the beginning.

If you have the patience to read this conclusion, it means that this content is useful, plain and not boring to you. There is too much that is not covered in this article regarding the question of “How to increase the volume of information flow bidding ads?” I believe that experienced operations colleagues also have their own set of effective methods. I hope that this article can provide some inspiration and help to those students who have just entered the workplace and hope to quickly learn experience, have a more thorough understanding of bidding advertising, and quickly prove themselves.

I also hope that students who are currently operating massive platforms, if you are confused, can recognize your core value and the direction of your future efforts in an environment where product tools are becoming more and more intelligent and creative factors are having an increasingly greater impact.

You want to ask me, what is the direction? I would say: Improving cognitive abilities, learning to make good use of tools, and improving creative and innovative abilities are the abilities that outstanding operations students in the future (or now) should possess. The platform will continue to work hard to improve the core capabilities of all students in the industry.

Author: Tiyangyang

Source: OceanEngine University (ID: oceanengine-univ)