Email marketing: what it is and how to implement it

Understanding of Real-time Bidding System

 Performance marketing is a part of the digital marketing environment, and it is possible to properly approach it only if you have an understanding of the bidding system. Without this understanding, it is impossible to delve deeper into conversion optimization. Why? This is because the bidding system is a bidding system that determines which advertisements are exposed to whom and at what price with some logic, and if you don't understand this, you can't expand your thinking map except to keep creating creatives or changing your targeting options. 

Why was the Realtime Bidding System (RTB system) created?

Step1. Advertisement products using traffic from web pages

To make it easier to understand the process of building the online advertising ecosystem, the story is as follows. First, traffic began to flow into an online space. A lot of people started flocking to a specific page to get some information. Naturally, if people flock to you, you can place advertisements. The owner of the online space leaves one side of the web page blank and sells banner advertisements on a daily or hourly basis to advertisers who want to place advertisements. This type of advertisement is the origin of CPT (Cost Per Time) and CPD (Cost Per Day) advertisements.

Understanding of Real-time Bidding System

Step2. Increase in advertising price due to increase in traffic

As time passes, the size of the traffic increases. As web pages develop, more and more people flock to see more information. As a result, profitability compared to advertising costs began to improve, and more and more advertisers wanted to place advertisements on the webpage. The owner of the webpage made a few more spaces for advertisements and gradually increased the number of ad slots.

 Web page owners are also becoming more profitable, and advertisers are also benefiting from being able to place more ads on non-best-placed, but profitable, web pages. Web page visitors can also obtain additional information from advertisements in addition to content. In this process, the advertisement price P increases significantly, and the quantity Q of advertisements also increases, but it is insignificant compared to the increase in P.

Understanding of Real-time Bidding System

Step3. The emergence of personalized advertising to address the growing number of unsold advertising accounts

But a problem arises. Although the number of ad slots has increased and the influx of advertisers has also increased, all slots are not sold unless it is a popular time slot of a popular slot. In addition to the non-sale of ad slots, the queue of advertisers for popular times of popular slots is lengthened, resulting in a loss of time and money for both the owner of the web page and the advertiser.

 To solve this problem, we need data about people entering the web page. If people can come in and know what products and services they are interested in, advertising can be tailored to each individual, allowing advertisers to monetize even when not in the popular slots of popular slots. Carpet bombing on the battlefield can of course annihilate all enemies, but hitting a strategic base with a pin-point guided missile strike can remove obstacles much more effectively. 

Understanding of Real-time Bidding System

Step4. Increased profitability due to the advent of personalized advertising

Now, even if you access the same web page at the same time, you will see different advertisements according to the individual characteristics of the visitors. Ad slots that were sold in chunks can now be sold individually. As the owner of the web page, the individual unit price P became very low as the advertisement slot, which was sold at a very high price per hour, was split into individual units, but the number of advertisements increased exponentially with the number of inflows Q, resulting in the final revenue size P *Q is an opportunity to dramatically increase.

Step5. Increased complexity of advertising management due to the growing number of market participants

As time goes on, the number of web pages on which ads can appear increases. Advertisers have more options to choose which web pages to place their ads on, but the complexity of management has also increased. In addition, the trouble of having to publish advertisements by verifying the right targets for each individual web page has also arisen. The same goes for the owners of web pages.

 As the number of advertisers increases, it becomes difficult to distribute ad slots, and the number of web pages that can display advertisements becomes so large that the competition intensifies the advertisements and makes their web pages spam advertising websites. had to avoid The owners of web pages faced the difficulty of finding the right line for themselves to sell ads to advertisers as much as possible, while not making visitors feel tired.

Understanding of Real-time Bidding System

Step6. The emergence of the RealTime Bidding System

To solve these problems, DSP (Demand Side Platform), SSP (Supply Side Platform), Ad Network, and Ad Exchange are introduced. DSP is a platform that negotiates advertisements with web page owners on behalf of advertisers. The advertiser only presents the cost to pay for an advertisement to DSP, and the DSP finds a web page that can display advertisements at that price. 

SSP is a platform that negotiates advertising costs with advertisers on behalf of the web page owners (hereinafter, publishers) who have advertising accounts. DSP and SSP move so that advertisement purchases can be made at the point that maximizes the profits of advertisers and publishers through the Ad Network. do. Broadly speaking, the advertisement purchase decision system composed of DSP, SSP, Ad Network, and Ad Exchange is called Realtime Bidding System (RTB system), and the advertisement purchase process takes place within 0.1 seconds in the RTB system.

Understanding of Real-time Bidding System

Schematic of RTB system

Let's take a closer look at the RTB system. Suppose a user browses the web. While user A was browsing the web to find information related to real estate, he arrived at a Tistory blog, and while browsing the contents of the blog, he clicked on the official sale advertisement of Company B located between the contents and landed on the sale homepage. Let's take a look at this process according to the diagram of the RTB system below.

 

Understanding of Real-time Bidding System

First, company B requests the DSP to place an advertisement. The DSP receiving the request makes an advertisement request to the ad network again. At this time, in addition to Company B, many advertisers such as B1, B2, and B3 who want to sell various real estate such as officials, apartments, and studios will want advertisements.

 The SSP responds to the Ad Network as to which publisher's account is suitable for displaying the ad, based on the publisher's ad accounts that are eligible to serve the ad. How much an advertisement is exposed is determined by the principle of supply and demand. If the number of advertisement accounts that can expose advertisements is small and the number of advertisers who want to expose advertisements is large, the advertisement exposure cost will be set at a high level. 

 

Understanding of Real-time Bidding System

If you draw only the schematic where the advertisement is exposed to the user, it is as above. When a publisher requests advertisement delivery to Ad Network (here, it is described as Ad Exchange, but performs the same function as Ad Network) (1), Ad Network sends information about advertisement space and users to DSP and advertises on the advertisement space. Requests to proceed with bidding for advertisers who want to expose (2).

 DSP receives detailed data of users accessing the page from DMP, analyzes which advertiser's advertisement is suitable (3), decides which advertisement to expose on the advertisement page and at what price, and returns it to Ad Network (4). Ad Network conducts a Second Price Auction with a large number of advertisements and advertisement prices received in this way, and finally decides which advertisement to be exposed at what price (5). And it transmits this price to the DSP. After that, the advertisement is posted on the advertisement page (7), and how the user reacted to the advertisement is tracked and fed back to the DSP (8).

DSP and media that can maximize conversion through the RTB system

The above process is an algorithm that determines the appropriate advertisement exposure price. However, in general, the purpose of placing ads online is to maximize conversions. However, maximizing the number of conversions cannot be achieved only by exposing advertisements with a low CPM. This is because maximizing the number of conversions is possible only by exposing advertisements to “ users who can cause conversions, ” but it is difficult only by exposing advertisements to more users at a low price.

 Since there are relatively few “users who can convert,” DSPs will have to purchase advertisement accounts with a higher CPM to expose advertisements to them. If DSP and media are performing these roles well, CVR should be formed at a satisfactory level even if CPM is high when an advertisement is set with the goal of maximizing conversion.

DSPs and media with RTB systems that are difficult to maximize conversion

Media operated based on machine learning undergo advertisement exposure and subsequent inflow and conversion in the same way as above, but media that are not operated based on machine learning are slightly different. Naver PowerLink does not individually target users who can convert. Instead, it operates based on the assumption that CVR will be higher if it is exposed to the top in search results. 

Similarly, it is difficult to find users who can convert among DA media, so there are cases where we focus on influx first and think about conversion later. In this case, CTR should be formed at a satisfactory level even if the CPM is high when an advertisement is set suitable for inflow or high exposure in search results. will be greatly affected

Conditions that may affect the RTB system

Who will you expose to?

The RTB system will be affected depending on the user segment to which the advertisement will be exposed. In addition, the performance of the system may vary depending on how well the user's behavior data is structured and organized and whether the RTB algorithm is well designed. The deeper the user activation and behavior within the medium, the more room there will be to structure the data. 

what to expose

The RTB system will be affected by the nature of the advertising material. This is because, according to the expected response rate of advertisements (CTR, CVR, etc.), the system will determine to whom, how much, and how quickly advertisements will be exposed, and receive data feedback to advance advertisement delivery.

where to expose

The RTB system may be affected depending on the various types of advertising space the system has. It may be affected by the form of feed advertisement, network advertisement, search advertisement, etc. Even if the same feed advertisement is an image feed and a video feed, even if it is the same network advertisement, the advertisement result data that is fed back to the system is Because it will be different.

How to adjust budget spending within the period

Adjustment of budget expenditure is sometimes manually adjusted by the advertising manager, but in systems that are being advanced, a pacing algorithm is used. Pacing is an algorithm that adjusts user inflow by distributing a given budget in such a way that the system can maximize goal achievement.

The above factors can ultimately be thought of as factors necessary for estimating CTR and CVR and adjusting the RTB system to allocate budget in a way that maximizes performance. We need to organize what changes occur when these conditions are changed and design a test set that will maximize the expected effect when those changes are combined. This part is connected with the understanding of machine learning in advertising media. After all, the RTB system and machine learning of advertising media are inseparable.

When advertisements are posted through the RTB system, the following are largely determined:

• 1. Campaign Goals: Which of the following goals will be maximized: impressions/clicks/conversions/other goals?

• 2. User Segment: The user segment with the highest expected advertisement effect 

• 3. Advertising CTR & CVR: Estimated CTR and CVR when advertising is exposed under constraints 

• 4. Performance data: Actual advertising operation data (CPM, CTR, CVR, CPC, CPA, etc.) 

• 5. Advertisement space: Advertisement space where the actual advertisement is displayed 

These decisions can be fine-tuned by changing the setting options of the campaigns and groups in each medium, mainly. This means that by adjusting the media's campaign and group options, the direction of movement of the RTB system that the media is leaning on can be adjusted.

RTB system and media conversion optimization connection

Aside from understanding the machine learning of advertising media, based on the discussion so far, it can be explained as follows by connecting the RTB system and the media conversion optimization. Fewer people are willing to buy, and a higher CPM is required to expose ads to them. This is because it assumes that there is a competitive state among advertisers. When competition occurs, a higher CPM is determined based on the secondary price auction. Regardless of the size of the target, there are relatively few people who are willing to buy, but if there are many advertisers who want to sell to these people, the CPM will go up.

Going a little further, if I sell carbonated water, my competitor is not just carbonated water companies. Companies that sell carbonated water-like colas, sodas, water, and perhaps even succulent fruits such as cucumbers and tangerines can compete together. Therefore, in the online environment, even if my products and services are niches, I cannot monopolize an advertising account for that niche target, and there is always competition. In media that advertise based on meta, Google Ads, or other high-performance machine learning optimization engines, they try to reach high-value users by executing conversion-optimized ads, and as time goes by, the overall CPM of the media increases.

To optimize conversion effectively in such a situation, the RTB system algorithm and conversion optimization machine learning algorithm of the medium itself have been advanced, so that the cost is reduced by properly exposing effective advertisements to a valid target, or strong competition is accepted as a constraint and the function of the medium You need to run a variety of operational optimization tests that can fully utilize the company, or develop powerful creatives that can make conversions more effective than your competitors.