Product strategy for search ads

Bidding advertising products: product strategies for search ads , and also summarizes the search ads designed by myself: bidding top.

01 Search Advertising Overview

The essence of the bidding transaction model is to remove quantity constraints from the transaction process and only adopt a simple decision-making plan of "the highest bidder wins" to place each advertisement. Bidding conforms to the trend of targeted advertising towards refinement , and also finds possible monetization channels for a large amount of surplus traffic that cannot be sold by contract, greatly enhancing the possibility and enthusiasm of a large number of small and medium-sized advertisers to participate in online advertising .

Search advertising has the status of a flagship product in bidding advertising and even the entire online advertising. In addition to its advantages in monetization capabilities and market size, more importantly, some of the core product strategies and technical solutions in computational advertising come from search advertising. Therefore, a deep understanding of search advertising is critical to understanding the entire bidding advertising market.

Search advertising is a typical bidding advertising product. Its characteristic is that advertisers compete in an auction-style for the advertising display opportunity of a certain target (here is a keyword), and occupy several positions for the advertising display in turn according to the results of the competition.

This is completely different from the display volume contract: first, the guarantee of quantity no longer exists, and advertisers need to adjust the balance between effect and quantity on their own; second, the price agreement is also removed, and each advertiser can adjust the bid for each keyword at any time.

02 Search advertising product form

Search advertising is an advertising product that targets audiences based on contextual query terms and is sold through bidding and settled on a CPC basis.

Usually, search ads are displayed on the search results page. The display area of ​​search ad creatives is generally divided into three parts: north, east, and south. All positions in the north and east regions constitute the position set of the same keyword auction. The positions are sorted as {north1, north2,···, east1, east2,···} during bidding, which is basically arranged according to the click-through rate of each position. At the same time, it is not necessary to place ads in all these locations, which is significantly different from banner ads. For advertisements in the South Region, different search engines have different product processing methods. Some directly copy advertisements in the North Region, while others directly copy the first few advertisements in the East Region.

The most basic form of search advertising is a text link that is consistent with the natural search results. It usually has a background color and words such as "Promotion", "Promotional Link", "Ads", etc. in the corners to distinguish it from the natural results. The purpose of this is to minimize accidental clicks by those who are not interested in the ads, thereby reducing advertisers' ineffective consumption and improving user experience.

The subject of search advertising bidding is the bidding keyword. The query entered by the user is matched with the keyword to determine whether the advertisement can be triggered. The matching method can be simple exact matching or more extended methods. Query expansion is also a key product strategy for search advertising.

The search ads discussed here are mainly based on general search engines. Many vertical searches, especially e-commerce searches, also have strong advertising monetization capabilities, but the product form may be different. In the entire product spectrum of Internet advertising, search advertising has a particularly important position and has the following distinctive product and technical characteristics.

(1) The monetization capability of search ads, i.e. eCPM, is much higher than that of general display ads. Therefore, research on some unique issues and algorithms related to search advertising has received great attention. The most critical product reason for the high monetization ability of search ads is that the queries actively entered by users directly reflect the user's intentions .

(2) Audience targeting tags for search ads, which are contextual search queries. Since search terms strongly represent user intent, search ads can be targeted very accurately. Compared with such contextual information, the importance of interest tags obtained based on user historical behavior is greatly reduced. On the one hand, this is because their signals are far less strong than search terms, and on the other hand, this is because such a task with clear user intentions must not be interrupted. Therefore, the eCPM in search advertising degenerates from r(a, u, c) in the general case to

r(a,c).

(3) The display format of search ads is very similar to that of natural results, with only subtle hints in the background color and text links. This kind of product design makes it feel like native advertising and further improves the advertising effect. On the other hand, the relevance requirements of such advertising results are much higher than those of display ads, so very sophisticated strategies and techniques are required when matching ads based on queries.

(4) The bidding transaction model developed from search advertising has gradually become the most mainstream transaction model for Internet advertising. This model fundamentally changes the way advertising is operated and releases huge productivity to rapidly improve its effectiveness.

03 New forms of search advertising products

From the perspective of product design, search advertising has the following three exploration trends.

(1) Enrich the creative display form of text links to make them more expressive and increase click-through rate.

(2) Take advantage of the fact that the eastern region has a slightly lower requirement for relevance to design some extended advertising products.

(3) Optimize the relationship between advertising and natural results to increase revenue while ensuring relevance and advertising effectiveness.

1. Creativity beyond text links

Clearly, the blue hyperlink is not the end of the road for search ads. As search results themselves develop towards richer displays, more direct content acquisition, and increasing industry specificity, search ads are also exploring some forms of display that can deliver more value, in order to simultaneously improve user experience and monetization efficiency. There are two key directions for exploration in this area.

(1) Add more expressive information points to the universal advertising link.

In addition to the standard content of text link advertising creatives such as title and summary, you can also add the advertiser's logo, main content link, contact number and other content. Practice has proved that these will improve the direct effect and brand value of advertising. More importantly, this development direction has little to do with the industry and can be replicated on a large scale, so it has been widely adopted in search advertising.

(2) Directly display a structured summary of the advertising content, and even provide some directly accessible functions, which can reduce the cost of user jumps and improve promotion efficiency.

As shown in Figure 2 above, the advertising product directly displays some main content links of the travel website and provides a quick entry for booking air tickets and hotels directly in the results. Such commercial results often come directly from direct search products, such as Baidu's Aladdin, 360's OneBox, etc., but because the results are paid, they should also be classified as advertising or commercial content. It should be pointed out that in order to ensure the consistency of user experience, this type of content often cannot be operated entirely using the bidding method of search advertising. The general operating method is to conduct periodic offline bidding in a certain industry.

The above development direction provides new space for improving the effectiveness of search advertising, but it has not yet become the mainstream of search advertising. The reason is that this model requires product and operation plan design by industry, and its scalability is limited. In addition, if a single product is directly connected, complex engineering problems such as product library connection, which is the same as personalized redirection, will arise.

2. Weakly relevant advertising formats

The product principles on the right side of the search engine are slightly different from those on the left side, that is, within a reasonable and explainable range, some generalized content with lower relevance requirements can be added, thus providing new space for the design of this part of the advertising product. We use two examples to introduce our exploration in this area.

(1) Part 3 of the above picture is a common form of advertising with a certain brand connotation in search advertising. It is a set of navigation search keywords corresponding to the advertiser, which displays the advertiser's brand promotion creativity when users search for these words. Such products can increase advertisers’ users’ awareness of the brand and subsequent stickiness.

(2) Part 4 of the above figure shows similar recommendation functions provided by some search engines. In this case, they are related travel websites. Obviously, the list and sorting of this part can be operated according to the logic of bidding advertising. This advertising product can provide search engines with advertisements that are a little further away from decision-making and aim to reach potential users. Of course, such products should never be displayed in the results on the left side of the search, because that would cause users to question the relevance of the results.

3. Native Exploration

Search ads have had the characteristics of native ads from the very beginning: their commercial results are the same as natural results, triggered by the user's active intent, and their display format is not much different from natural results. In fact, a considerable proportion of users are completely unable to distinguish between natural results and advertising results. Therefore, exploring the natural combination of search ads and content and placing ads in an increasingly "native" way is also an important product direction.

Unlike other search engines, Baidu's search ads have been mixed with natural results using a paid ranking method from the very beginning. The mixed format has actually always existed in Baidu's advertising products and still plays an important role in revenue. Although this mixed ranking has received mixed reviews and actually reduced the quality of results for some queries, it can be said to be an exploration of the native nature of search advertising.

Another native trend is to mix commercial results into some information searches directly targeting products and sort them in a unified manner. Compared with only providing organic results, this can enhance monetization capabilities; compared with only providing advertising results, this can avoid the decline in result quality caused by insufficient demand. This type of advertising can also be called "direct product advertising."

Search advertising is actually a more important origin of native advertising than information flow advertising, and some important features of search advertising can be borrowed into the products of native advertising platforms.

04 Search Advertising Product Strategy

In addition to innovation in product form, search advertising delivery and optimization strategies are also an important part of the product. To understand the key points in product strategy, let’s first look at the basic process of search advertising decision-making.

The entire decision-making process of search advertising can be divided into several stages, including query expansion, retrieval, ranking, placement, and pricing. Query expansion is a strategy unique to search advertising, which aims to automatically expand relevant query terms for advertisers and increase purchasing traffic. Ad retrieval and ranking candidate ads according to eCPM are relatively common core processes of advertising systems. Pricing is a very core strategy for bidding ads.

It should be noted that in search advertising, the basis for ranking, namely eCPM, can be simply expressed as r(a,c)=µ(a,c)·bidCPC(a). However, in actual products, the click-through rate µ will be replaced by the quality score. In addition to considering click-through rate, quality also needs to take other factors such as landing page quality into consideration to get a more comprehensive evaluation of advertising quality. Its purpose is to avoid malicious behavior of advertisers and promote long-term market development.

1. Query expansion

Query expansion is a key strategy in search advertising. For advertisers, it is not easy to find a combination that meets their needs from the vast sea of ​​keywords. Therefore, search engines provide some services to match keywords in advertisements to more relevant queries. Several common matching methods are as follows.

(1) Exact matching, that is, no expansion of the keywords provided by the advertiser is made in any form, ensuring that the keywords are accurately executed in accordance with the advertiser's intention. The execution method of exact matching is to first segment the query entered by the user. For example, the query "English training" can be divided into the word set {English, training}. When this word set is exactly the same as the keyword set set by the advertiser, exact matching is triggered. Taking the keyword "English training" as an example, under the exact matching method, the two queries that may trigger the advertisement are "English training, training English".

(2) Phrase matching. When the user's query completely contains the advertiser's keywords and the inserted or inverted form of the keywords (including synonyms of the keywords), it is considered a successful match and the corresponding advertising candidates can be triggered. Taking the keyword "English training" as an example, under the phrase matching method, the searches that may trigger the advertisement include "English training, English training summer classes, which English training institution is good, English training, English related training, English training" etc. But words like "English songs, computer training" will not be triggered. This method uses more detailed concepts to match broader concepts, so it is generally more accurate.

(3) Broad match. When the user's query words are highly relevant to the advertiser's keywords, they may be matched even if the advertiser did not submit these query words. Taking the keyword "English training" as an example, under the broad matching method, it may trigger queries such as "foreign language learning classes, Level 4 certificate". There is no clear definition of the logic of broad matching. Generally speaking, it is automatically obtained using data mining algorithms. Therefore, although turning on broad matching can obtain more traffic, the accuracy of the traffic will often be discounted.

(4) Negative matching. Since phrase matching and broad matching are completed automatically by the system, some inaccurate matching results are inevitable. Therefore, it is necessary to provide advertisers with the function of negative matching, that is, to clearly indicate which words cannot be matched, so that some inefficient traffic can be flexibly shut down.

Phrase match and broad match are both typical methods of keyword expansion. By combining them with negative match, advertisers can more freely find a balance between traffic and quality. However, due to the inconsistency between the interests of the media and advertisers, the keyword expansion method provided by search engines may not be the best for advertisers.

2. Ad placement

After the ad candidates are ranked, the number of ads for the north and east regions needs to be determined respectively. This step is called ad placement. Since these two areas constitute a unified bidding queue, it is actually necessary to set the conditions for entering the North District and the East District separately, among which the most critical is the condition for entering the North District.

The North District is a prime ad placement and is crucial to search ad revenue. Generally speaking, the majority of advertising revenue for general search engines comes from the North Region, so the average number of advertisements in the North Region is directly related to revenue. But at the same time, since the North District advertisements directly lower the position of natural results, it will inevitably have a certain impact on the user experience. Therefore, in order to achieve a better balance between commercial interests and user experience, we should pay special attention to the quantity and quality of advertisements in the North District while paying attention to revenue indicators. When examining the impact of a new algorithm on RPM, we should also pay attention to the average number of advertisements in the northern region of the entire search engine, NFP. When NFP changes significantly, the corresponding RPM changes often have little reference value.

There are two key factors to consider when determining whether an advertisement can enter the North Zone: first, whether the advertisement is relevant enough ; second, whether the advertisement’s RPM is sufficient . The former is to ensure user experience, and the latter is to make efficient use of placements. In addition, generally speaking, the North District will set an upper limit on the number of advertisements. Based on the overall NFP constraints and revenue targets, we can easily determine the optimal thresholds for correlation and RPM through data simulation methods.

As mentioned before, search advertising decisions generally do not take the influence of user u into consideration, but there is an exception when it comes to determining the number of advertisements in the North District, which is personalized advertising placement.

Due to different personal preferences and understanding of ads, different users have significantly different tolerance and click-through rates for ads: some users always skip ads and start browsing directly from natural results; but some users either don’t distinguish ads very well or are highly receptive to them and treat ads the same as natural results, resulting in a large number of clicks.

Obviously, for the former type of users, the number of advertisements in the North Zone should be reduced, so as to obtain a better user experience without significantly reducing revenue; while for the latter type of users, the location in the North Zone should be used as much as possible when there are eligible advertisements.

05 Search Advertising Product Case

1. Google AdWords

AdWords is a keyword bidding advertising product launched by Google, which generates tens of billions of dollars in revenue for Google every year. In the initial stage of the product, AdWords was targeted based on search terms and sold using the then-popular CPM method, but unfortunately the sales results were not ideal. Later, it adopted Overture's CPC sales method and creatively introduced the concept of click-through rate into the CPC sales method, expressing the relevance of advertisements through the click-through rate of advertisements, and achieved very high revenue.

When Overture invented the CPC selling method, the way of advertising was very simple. It was just sorted by advertisers' bids, and the ad with the highest bid won the opportunity to be exposed. Google discovered the flaw in this model: the ads with the highest bids won a lot of impressions, but if no one clicked on them, the platform still had no revenue. Therefore, Google changed its strategy to estimate the click-through rate of each ad during the delivery process, and then sort the ads by the product of the click-through rate and the bid. This has formed the logic of making decisions based on eCPM that is commonly used in bidding ads today.

After introducing click-through rate to express relevance, an ecosystem that is beneficial to advertisers, users, and platforms has been formed under the CPC model. In this ecosystem, advertisers can target audiences by selecting keywords and reduce bids by improving ad creatives to increase click-through rates. Google increases its revenue by delivering high-click and high-paying ads to users. Because each time users are presented with a "personalized" ad with a high click-through rate, they see more relevant paid information when using search engines, reducing the interference of ads on users.

As relevance increases, advertisers will win more clicks with lower bids, so advertisers will have the motivation to increase the click-through rate of their ads through more precise keyword conditions and better advertising creativity. The revenue of the entire system also increased accordingly.

It should be noted that in order to simplify the concept, the AdWords introduced here refers to its past product category. Currently, Google has unified the demand-side entrances for its display advertising network and mobile advertising network under the AdWords product, with the aim of providing advertisers with a unified market to facilitate unified effect tracking and optimization. This is actually an important trend in the demand-side interface of advertising products.

2. Taobao Express

Compared with AdWords, a general search engine advertising product, Taobao Express is an advertising product of Taobao, a C2C e-commerce company, that specifically serves sellers. Taobao Through Train ads mainly appear on the search results page of Taobao. However, since the display of product search results is very different from that of general search engines, the corresponding advertising display is also very different. The ads only appear on the east and south sides of the page.

The predecessor of Taobao Express was Yahoo Express, which allowed sellers to place advertisements on Yahoo Search and Yahoo websites. It was renamed Taobao Express in 2008. The initial ad ranking rules are very simple, that is, advertisers with higher bids win the opportunity to be placed. Taobao later improved its advertising strategy and began to adopt the same eCPM ranking method as AdWords. Compared with the previous advertising ranking rules, the conversion rate of sellers is improved because the click-through rate of the ads is relatively high, and Taobao also obtains more revenue due to the increase in the overall click-through rate of the products, forming a win-win situation.

Unlike general search engine advertisers who come from all walks of life, Taobao Through Train advertisers come from the seller group on the Taobao platform. Taobao and its advertisers basically have a symbiotic relationship, which gives Taobao greater influence on its advertisers and significant advantages in the following aspects.

(1) Advertisers can promote individual products and directly reuse product images, descriptions, and other information without having to create creatives for specific advertising channels. This makes the additional cost of using the Through Train very low for a large number of small and medium-sized sellers.

(2) Since on-site search is strongly related to user purchasing intent, the advertising effect is better.

(3) Since Taobao has a comprehensive understanding of the entire conversion process of its advertisers, Taobao Through Train can use subsequent data to optimize the advertising system, such as conversion estimation and synchronization of product listings and delistings, to a level that is difficult to achieve with general search advertising.

Therefore, although Taobao Express is just an advertising system for a vertical search engine, the ecosystem supports most of the revenue and profits of Alibaba, China's largest Internet company, and is also one of Taobao's main ways to monetize. This example tells us that in some vertical search engines with high commercial value (such as e-commerce, real estate, automobiles, and app downloads), using the product system of search advertising to monetize is the most prioritized method of traffic monetization. In the vertical recruitment industry where I work, the bulk of my revenue currently comes from refresh and top products that are monetized using the search advertising product system.

06 Position Auction and Mechanism Design

Bidding advertising, represented by search advertising, actually sells advertising display opportunities like an auction. In other words, the system decides who gets the ad space for a certain display based on the advertiser's bid and the eCPM calculated from it. In the initial stage of competitive bidding advertising, advertisers adjust bids in stages; in the real-time advertising transaction stage, advertisers can adjust bids in real time for each display. However, from the macro perspective of the auction market, there is no essential difference between the two types of bidding.

Let’s first take a look at how to describe the bidding advertising problem and understand some important conclusions from a macro market perspective. As shown in the figure below, suppose there is a set of ad slots that can be occupied. These ad slots are ranked according to their empirical value and denoted as s=1,2,···, S (for banner ads, S here is generally 1). In a certain advertising request, there is a group of ads a=1,2,···, A bids to participate in the auction. The bid of each ad is denoted as ba. The system puts the first S ads with the highest bids in the first S sorted ad positions in turn. This problem is called a position auction.

According to the previous discussion, when an ad a is placed at position s, its expected revenue, i.e., eCPM, is ras=µsνa. We have made some assumptions here, for example, the click-through rate µ is only related to the position s, and the click value ν is only related to the ad a.

How to design some important mechanisms in such position auction problems often has a huge impact on the revenue, stability, fairness, etc. of the entire bidding market. This type of problem is called mechanism design problem. Common mechanism design issues in advertising include pricing, market reserve price, price squeeze, etc.

07 Pricing Issues

The most important mechanism design around position auctions is the so-called pricing problem, which explores how to charge appropriate fees to advertisers who finally obtain a certain position in a position auction given the bids of each participant and their expected returns.

In an auction environment where participants such as advertising can continuously adjust their bids for the same object, it is entirely possible to create higher returns and better market stability for the entire market through smart pricing strategies.

From the perspective of the entire market, we need to focus on studying the returns and other characteristics of the market when it is in a stable state. The so-called stability means that the entire bidding system is in a Nash equilibrium state, that is, each advertiser has obtained the position that best suits his or her own interests through bidding. For a certain position bidding, its symmetric Nash equilibrium state can be expressed as follows:

Note that the meaning of the subscript here has been adjusted. Here νs refers to the click value of the advertisement in position s, not the click value brought by the s position, and qs refers to the fee charged by the market to the advertisement in position s, that is, the pricing, which is the advertiser's single investment. The significance of this equilibrium state is easy to understand: for each advertisement in the final position ranking auction result, its revenue is higher than that in other positions. Obviously, under such conditions, each advertiser has reached his or her optimal state, and the entire system has stabilized.

In the above formula, the only strategy that the market can adjust is the way to determine qs, that is, the pricing strategy. Depending on the pricing strategy, the macro-return situation and the degree of stability when the market reaches a stable state will vary. Therefore, the main purpose of research on the macro properties of bidding markets is to find better pricing strategies to optimize overall revenue.

The most common pricing strategy in the online advertising bidding market is the GSP scheme; there is also a VCG pricing strategy, which is more reasonable than GSP in theory, but because the principle is more complicated and difficult to explain to advertisers, it is not widely used in actual systems. Below we introduce these two pricing strategies respectively.

1. Generalized second highest price (GSP)

Let’s first take a look at what the second highest price is. The so-called second highest price refers to the bid of the next advertiser to win the position in an auction with only one position. Such an auction is also called a Vickrey auction. In the auction process of search advertising with multiple positions, it is easy to intuitively generalize the second-highest price strategy into the following strategy: for advertisers who win each position, they are charged according to the bid of the next advertising position. This is the generalized second-highest price.

The intuitive rationale for the second highest price and the broad second highest price has been explained above. But in fact, the second highest price is the optimal pricing strategy for single-position auctions, but the generalized second highest price is not the optimal pricing strategy for multi-position auctions (the optimal strategy is VCG pricing to be introduced below). Although it is not theoretically optimal, the generalized second highest price has many advantages in operation, such as being simple to implement and easy to explain to advertisers. Therefore, it is the most mainstream pricing strategy in the actual bidding advertising system.

When I was doing search advertising and bidding for top products, I also used the broad second-highest price pricing strategy.

If settlement is based on CPM, then the broad second highest price can be applied very directly. However, in the CPC-settled bidding advertising system, the advertiser's bid is for clicks, while the bid is for eCPM. Therefore, the two must be converted to achieve the generalized second highest price under CPC. The pricing formula is as follows:

If we multiply both sides of the equation by µs, we can see that the generalized second-highest price is actually still the second-highest price in terms of eCPM. Regardless of whether it is CPM or CPC settlement, under the broad second-highest price scenario, the pricing for an advertiser will definitely not be greater than its bid. In fact, this pricing strategy is also applicable to the bidding market for CPS settlement, and you only need to replace µ in the formula with µν.

The ∆ at the end of the formula is generally the smallest unit of settlement currency in the advertising system, such as 1 cent. This is a historical practice and, to some extent, makes advertisers feel more fair psychologically.

2. VCG

VCG pricing is a theoretically superior pricing strategy obtained by Vickrey, Clarke and Groves when studying the equilibrium state of the bidding system. The basic idea is that for an advertiser who wins a certain position, the cost he pays should be equal to the value loss caused to other market participants by occupying this position. Under this principle, VCG's pricing strategy can be expressed by the following formula.

The intuitive rationale for this pricing strategy is easy to understand. Theoretical analysis shows that the superiority of VCG pricing strategy is reflected in the following aspects:

First, in the stable state of this pricing strategy, the entire market is truth-telling. The so-called truth-telling can be understood as every advertiser finding his or her own optimal state.

Secondly, compared to other pricing strategies, this pricing charges the least to advertisers. In the case of a single ad space auction, the VCG pricing strategy degenerates into the second highest price strategy.

Despite the many advantages mentioned above, VCG pricing is not a mainstream method in bidding advertising.

This is mainly because the logic of this pricing method is too complicated and difficult to explain to advertisers. In addition, when there is a game relationship between advertisers and the media, it is difficult to verify whether the media has correctly calculated the "value damage caused to other market participants." However, this pricing method also has its market space. Some advertising companies, such as Facebook, have adopted this pricing mechanism in their actual bidding advertising systems.

08 Market Reserve Price

In order to control the quality of advertisements and maintain a certain selling price, the bidding advertising market often sets a minimum price for winning an auction position. We call this price the market reserve price MRP, commonly known as the "starting price" or "reserve price."

Advertisers can only get a bidding opportunity if their bids are higher than the market reserve price. At the same time, after winning a certain auction position, if the payment calculated according to the pricing strategy is lower than the market reserve price (taking the broad second highest price as an example, it is easy to verify that this situation is possible), it also needs to be adjusted to the level of the market reserve price.

Determining MRP is an important product strategy for bidding advertising. Setting MRP too low or too high is not conducive to maximizing the profits of the entire market. Generally speaking, when competition is sufficient and advertisers have sufficient depth, MRP can be set higher ; otherwise it should be appropriately lowered. There are two ways to set a market reserve price. One is to use the same reserve price for the entire bidding market, and the other is to set different reserve prices based on the characteristics of different subject matters (such as keywords in search ads). If you follow the latter method, you should obviously set a higher MRP for keywords with more intense competition.

There is another point in the practice of search advertising: although the advertisements in the North and East zones share a bidding queue, in order to ensure that the prime location in the North zone is better monetized, we often set a higher MRP for the North zone separately.

There are also some theoretical research and practical methods on how to calculate the optimal market reserve price. Simply put, the optimal starting price for a specific keyword is only related to the distribution of bids and quality scores. By fitting its distribution truncated by the reserve price as the core, the optimal starting price can be solved through a theoretical calculation formula. Since the paid search business model was formed, search engines have been constantly trying to adjust the starting price of keyword auctions to improve their own profit levels while being acceptable to the market.

In 2001, Yahoo! In 2008, Yahoo! set its starting price uniformly at 5 cents. Combined with the theoretical derivation results, some Yahoo search keywords are dynamically priced, and keywords of different values ​​are adjusted to different starting prices. After this price adjustment, based on continuous observation, the impact of the starting price increase on revenue in the third quarter of 2008 was +13%. Google has not disclosed its starting price calculation method and process, but it is generally believed that its starting price setting strategy is relatively advanced in the industry. As early as Yahoo! Before implementing dynamic starting prices, researchers discovered that Google's starting prices were also dynamic.

Whether it is adjusting the overall MRP of the market or dynamically setting the MRP based on different traffic segments, the basic principle is to find a CPM floor price that does not significantly reduce the fill rate based on the eCPM distribution of bidding advertisers, and then calculate the CPC floor price based on the quality. In fact, due to the consideration of quality, the dynamic reserve price can be related to the three parameters (a, u, c) at the same time, making it completely dynamic.

09 Price Pressure

In CPC-settled advertising products, eCPM can be expressed as the product of click-through rate and bid, that is, r=µ·ν=µ·bidCPC. However, in the design of bidding mechanisms, this formula is sometimes fine-tuned to become the following form:

Here, κ is a real number greater than 0. We can consider two extreme cases to understand the role of κ:

When κ→∞, it is equivalent to sorting only by click-through rate without considering the effect of bidding;

On the contrary, when κ→0, it is equivalent to sorting only by bids.

Therefore, as κ increases, it is equivalent to our role in squeezing bids in the entire bidding system, so we call this factor the price squeeze factor.

The main function of the price squeeze factor is to more proactively influence the bidding system to develop in the desired direction according to market conditions. For example, if it is found that there are a large number of advertisers in the market who bid high but have low quality, the impact of quality and user feedback can be emphasized by raising κ ; if it is found that the bidding in the market is not fierce enough, competition can be encouraged by lowering κ . If there is short-term financial pressure, this can increase overall revenue in the short term; if the purpose is to encourage advertisers to improve the quality and relevance of their advertisements, the impact of bids can be reduced by increasing κ.

10 Conclusion

The author works on a recruitment website and designed a search bidding ad: bidding at the top. Overall, there are some differences between this product and standard search bidding advertising products.

And because the current competition environment is not fierce, the sales department prefers to sell the top advertising space according to the fixed price, so there are now two types of top products.

1: Pricing is set at the top.

2: Bidding is at the top. The bidding top will only be displayed when the priced top is fully booked and the competition is fierce.

Subject matter: word bag. Package and sell the hot search terms of C-end job seekers.

Settlement method: cpt. It is different from general search advertising which is settled by CPM or CPC. More like an evolved version of ad slot ads.

Pricing: Use the broad second-highest price, that is, for each position won, the fee is charged according to the bid of the next person.

Market Reserve Price : A reserve price is set for each word package to win an auction position. An advertiser's bid will only be given a chance to bid if it is higher than the market reserve price. Currently, we have set different minimum prices for word packages based on different subject matters (word packages).

Price squeeze : Currently, the price squeeze factor is not considered in the design because the current market bidding is not fierce enough, which is equivalent to using a very low k to encourage competition.

There are many areas that can be optimized later:

1. Continue to optimize word packages. The generation rules of word packages directly affect the effect of top placement.
2. The market reserve price is set directly based on popularity and city levels, and is not calculated using historical statistical data. Based on the historical data distribution, we can find a minimum price for the word package that does not significantly reduce the fill rate.
3. In order to encourage competition, the current sorting is based directly on bids, not ecpm.
4. In actual ranking, click-through rate is not considered and is not estimated. The experience is not friendly to the C-end. In order to achieve healthier long-term profits, when designing products, we should start to consider moving towards a more refined direction in traffic monetization.
5. Find a balance between sales efficiency and product complexity. Currently in the recruitment industry, HR seems reluctant to spend energy learning these complex commercial products. People are still more willing to spend time lobbying candidates. When HR uses bidding products, the uncertainty of bidding is already a headache. Sales also find it very difficult to understand the concept of bidding. Can you explain the bidding products settled by CPC to HR? These are all issues that need to be considered
6. A pay-per-performance model seems to be the most ideal, such as pay-per-delivery. However, we have no way to achieve ideal profit by achieving CPS in a short period of time. Faced with the company's revenue pressure, CPS's attempts to develop new products came to an abrupt halt. But it is undeniable that this should be the product direction of the recruitment industry in the future. For example, I might consider finding a pay-per-delivery location in the recommended information flow and try a new product called CPS.

Author: The Prince Turns into a Frog

Source: The Prince Who Turns into a Frog