Starting from SEM automated management tools

Document directory

• A. Rule Optimization
• B. Combination Optimization

From: http://semwatch.org/2010/01/sem-autobidding-tools/

 

Author: gaoge Date: January 7, 2010 category:

Paid search for SEM-18 comments

We often say that SEM includes
Seo and narrow SEM (paid search. Although the two Ad display forms are very similar, they are achieved through advertising and rankings, but there are huge differences in operation. For seoer, it is very common for one or two cool people to beat the world, And Seo service providers are even more cool. Relatively speaking, the narrow SEM is far from a complicated system technically, and it is not what one or two smart heads can do. In this field, most of the advertising consumption is swallowed up by a batch of evolving automatic optimization tools and the large team behind them.

The concept of automatic optimization is still relatively unfamiliar in China, but in the North American market, automatic optimization has become the main means of SEM management for large and medium-sized customers. On the other hand, although automatic optimization has evolved several times in the past year 45 s, it is still far from perfect, and there are still some arguments about the basic theory of automatic optimization. In fact, from the discussion of the brand term, we can see that
SEM optimization has a broad scope. Because there are too many variables, fuzzy judgment is often introduced in the optimization process. In general, I don't think there will be any automatic optimization tool that can fully replace humans. In particular, the current Chinese search marketing market is still immature, from the perspective of platforms, agents, and advertisers. Automatic optimization is a luxury for many advertisers. However, to do well, you must first sharpen your tools. Understanding the advanced tools in the North American market should be of reference to Chinese search and marketing administrators.

First, we need to clarify the concept of automatic optimization. Automatic Optimization products on the market mainly come from two directions: one is that traditional monitoring data service providers or web Analysis Service Providers (click to manage) start to enter the Search Marketing Management (click to manage) market; the other direction is the independent search engine bidding service provider, that is, the CPC Bidding Management Service Provider. The main products in this field are automatic Bidding Management tools.

SEM advertising includes four parts: 1, website; 2, keyword; 3, advertisement; 4, bidding.Why have automatic bidding products become the mainstream in the current market?Let's consider this issue from the perspective of AD rank. Ad Rank = QS * max CPC. QS is the quality score, which is a black box. There are many optimization factors and there is no simple rule to follow. In fact, for a well-prepared advertising series, once the ads are put on the market, it is difficult for QS to be established and further improved. Most advertisers have a long period of improving their QS, with a small margin. The adjustment of bidding price becomes the main means of routine management of SEM.

Currently, technical service providers in the North American market often emphasize that automatic optimization is a system solution, but so far, there have been advantages and disadvantages of the products from the above two directions, there is no truly mature automatic system optimization product. Of course, the progress in technology and business strategies is amazing in the cash-rich market of search and marketing. In the past year, the so-called end-to-end Automatic Optimization products have become a strong trend in the market. Such cross-platform system optimization products are still not fully automated in the sense, but are already enough to form a huge challenge. Automatic system optimization is the trend of the times. There is no overlord in this highly competitive market. Even the most advanced service provider, once it has not followed suit, it may fall into the dust for a year or two.

Automatic CPC spot price optimization is only an integral part of the optimization process, but we have discussed it before, which is the most direct part of the impact on account performance. To put it bluntly, the CPC Bidding Management can determine the profit and loss of a search marketing plan. At the same time, CPC management often occupies the main time and energy for manual optimization. Especially for large and medium-sized customers, the price management of thousands of keywords may be enough to keep search marketing administrators away. It is reasonable to use CPC Bidding Management as the starting point for automatic optimization.

Existing CPC spot price Automatic Optimization products on the market generally use APIs to capture click data from ad publishers and adjust the CPC Based on the conversion data provided by monitoring. There are two main adjustment methods:

A. Rule Optimization

If you use the "catch up, fight, and fall" principle to adjust the spot price for keywords, you only need to input the original data and adjustment Rules (rule) into the model, and the computer can immediately give the adjustment value, this is the so-called rule-based optimization (rule based approach ). In the initial phase of automatic optimization, this optimization is the main choice. Currently, it is still a very active and popular automatic optimization method. Most Automatic Optimization products in the North American market are based on rules. This is especially true for the automatic bidding products provided by data monitoring service providers. The famous ones include Atlas search, omniture searchcenter, Web Trends, and dart search. independent service providers such
Did-it, acquisio, and clickable are also optimized based on rules. Google adwords conversion optimizer is probably the most well-known one for Chinese users, because it is the only truly free automatic optimization tool on the market (with unsatisfactory results ).

Rules-based automatic optimization technology is relatively low, because the judgment (that is, Rule Making) is manually selected, the advantages and disadvantages of each product are mainly reflected in the rule-making principles and practical applications. Essentially, this is more like a large decision tree system. The cost of spot price products with optimized rules can be relatively low, and can be shared in the SEM market. Even a relatively simple product can free up the time and energy of the search and marketing administrator from the heavy work and put it into a direction that requires fuzzy judgment, such as the development of new ad content, new Keyword mining and so on. On the other hand, such optimization still relies heavily on the manual judgment of the search marketing administrator.

B. Combination Optimization

Because large and medium-sized customers often have a large number of keywords, the performance of different keywords varies greatly. Some keywords have a high income, but they are highly competitive, and the cost is also high. Some keywords may not be searched, and their income is very low, but there is almost no competition. Once converted, the return rate is very high. This situation is a bit like the product performance in the financial market, so some people have introduced the investment portfolio model originally used for Wall Street into the search marketing management, through a series of product portfolios to reduce risks and increase high returns. This method is called portfolio-based optimization (portfolio based approach ).

Let's look at this example: we have two keywords in an account at the same time. If keyword 1 is ranked first, 15 registrars can be obtained. The unit registration cost is 2 yuan. If it is ranked second, 10 registrars can be obtained. The unit registration cost is 1 yuan. If keyword 2 is ranked first, you can get 20 registrations, and the unit registration cost is 2.5 yuan. If it is ranked second, you can get 10 registrations, and the unit registration cost is 2 yuan. Assuming that each registration value is 2 RMB, what should we do?

If you use rule-based optimization. It is very simple. If keyword 1 is pushed to the first place, 15 registrations are obtained. Keyword 2 can only be kept in the second place, and 10 registrations are obtained. At this time, the average unit registration cost is 2 yuan, meeting the requirements, the total registration is 25.

If combination-based optimization is used. Keyword 1 is placed in the second place, and keyword 2 is put in the first place. At this time, the registration cost of keyword 1 is 1 yuan, and the registration cost of keyword 2 is 2.5 yuan. A total of 30 registrations are obtained. The average unit registration is (2.5*20 + 1*10)/30 = 2 yuan, still meets the requirements.

Based on combination optimization, we have obtained 5 additional registrations at the same unit registration cost, which is the advantage of combination optimization. For an inaccurate analogy, this is "Tian Ji Horse Racing ".

The concept of combined optimization is intuitive in this example, but it is much more complicated in actual operations. If the customer has thousands, tens, or even hundreds of thousands of keywords, how can we select the best CPC match in the keyword combination to obtain the best overall performance in the current market environment, this is not a simple four-Rule operation, but a self-evolved prediction and decision-making system.

The source of portfolio optimization, the portfolio model, is a typical operational research application (that is, the operational research generated during World War II ). Similarly, composite optimization is also a typical operational research application. Its scope includes statistics and probability, matmatical modelling, simulation, and optimization), rule algorithms (algorithums), data mining, and data warehousing ). To put it simply, composite optimization is to establish a mathematical model by  on the performance of existing accounts, and simulate the performance of all keywords at various levels in the model, in addition, the rule algorithm is developed to filter out combinations that may produce the best results from the massive possibilities and put them into practice. Then, the simulation optimization is carried out through the  of the implementation results, so as to achieve continuous optimization with a spiral rising.

Due to the complexity of combination optimization, there are fewer Automatic Optimization products using this model. The most famous of course is efficient frontier, which is the first of these products. The world's largest independent SEM agent service provider is also the only automatic optimization product service provider that has entered the Chinese market. Searchignite/360i, followed by EF, is well known in China. In terms of competitiveness, marin software and kenshoo have sprung up in the past two years. These two products provide a complete automatic optimization solution that includes automatic bidding products, and both are self-help products, so the charges are significantly lower than EF or 360i. In my personal opinion, the majority of the Search Marketing market is small and medium customers, so
The idea of marin or kenshoo will gradually become the mainstream. EF has served eBay and Amazon, and these giants have left, and Expedia and Microsoft will leave sooner or later. When the SEM scale is large enough, there is no reason to outsource this business, and the benefits are too great. Therefore, EF and 360i must develop self-help products to adapt to small and medium customers. Google is the real competitor of all automatic optimization tools ......

Someone often asks me what automated tools I use at work, or thinks what tools are better. Looking at the glossary of the two concepts, it is obvious that the combination optimization should be more advanced.The word "algorithm" is used in the SEM field, except that the search engine is a combination optimization.In fact, from the perspective of market share, rule optimization is still the mainstream in the market. Not all advertisers need the most complex algorithms. Composite Optimization achieves self-evolution through a large amount of data accumulation, which may be short for Small and Medium customers. If the customer scale is small and the keywords are relatively simple, the combination method may not be more efficient than rule optimization. Therefore, my answer to this kind of question is: as long as the product has been sold for more than two years on the market, it is good that bad products cannot survive in such fierce competition. SEM
The marketing effect is good. In the end, it is not because the automation product is not good enough, but whether the product meets the needs of advertisers and whether the search marketing administrator can make good use of the product.

Automatic Optimization has gradually become an important method of CPC Management in Search Marketing Management in North America. It is closely related to the maturity of search marketing in North America. The search marketing investment occupies a very high proportion in the entire marketing, highly competitive market, relatively mature user search behavior, more advanced search marketing platforms and technical means. All these make the daily management of search and marketing more complex, the workload is greater, the difficulty is higher, and the popularity of automatic optimization has become an inevitable trend.Is Automatic Optimization suitable for the current Chinese market?In my opinion, after two years of rapid growth, at least some large customers in the industry have realistic and meaningful demands for automatic optimization. On the one hand, the fierce competition in domestic search marketing sometimes exceeded North America. On the other hand, the customer's understanding of search marketing is becoming more and more detailed. Practice has proved that appropriate automatic optimization products can play a huge role in the domestic market. No matter whether it is a simple automatic report system, a preliminary rule optimization system, or the most complex combination optimization system, there is sufficient application space.

Whether in North America or in China, the basic principles for measuring Automatic Optimization products are the same:

1. Is the product developed by the supplier? Only those who have mastered the product can truly understand the full potential and value of the product;
2. Is there any precise mathematical mind behind the product? Although the application of operational research requires professional mathematicians, the leading optimization products are basically behind a group of the best mathematicians. The rule organization of rule optimization is crucial. Mathematical support is also required for how different data is weighted and how different decisions are optimized. Said Hal Varian, chief economist of Google, that marketing is the next Finance and the metering trend of marketing.
3. Whether this product has full and meticulous prediction capabilities. Considering that automatic optimization is the selection of simulation results, keyword/advertisement-level simulation prediction is the basic element. This is a means to distinguish between Rule Optimization and combination optimization, because Automatic Optimization Based on rules usually does not require a lot of simulation.
4. Whether you have the learning ability. The big closed-loop learning approach is the basis for ensuring that products can adapt to continuous optimization based on the environment.
5. Whether the model's preview and postview capabilities can adjust the historical reference range as needed, so as to calmly cope with fierce market changes.

These are suitable for good automatic optimization products. WhileFor the Chinese market, it is clear that this product also needs to be able to support domestic search platforms, as well as the product supplier's understanding of domestic customers and the market environment. China's search market is still evolving, and one of its manifestations is a serious blow. Effective Combination Optimization requires not only precise mathematical models, but also a large amount of capital investment to verify the results from real advertising and continuously correct them. Because the entire market will have a huge impact on the large Closed-Loop Correction of the model.

These reviews are based on the automatic optimization of the product itself. In fact, the prerequisite for automatic product optimization is the customer's understanding and needs for marketing management. We need to discard the idea that automatic optimization can replace our search and marketing management. Even automatic optimization tools with self-learning capabilities can only play a role in an appropriate environment. The relatively low domestic search marketing market is first manifested in advertisers. If the customer does not know what search marketing requires, the difference between Manual Optimization and automatic optimization is meaningless.

In the example of Combined Optimization, we only need to slightly adjust the direction and the results are completely different. For example, we no longer pursue the maximum number of registrars, but the biggest profit. In this way, you do not need to use a combination of optimization methods, just follow the profit margin of each keyword. A penny is also earned, and a penny is also lost. The highest profit point is achieved at ordinary times in terms of marginal profit and marginal cost. In practice, a considerable number of customers are pursuing the biggest profit, but the final result is often an ROI indicator. Why? If the customer cannot answer this question effectively, it means that they do not need automatic optimization at the present stage.

To put it back, excellent automatic optimization products are only input and output on machines. What machines do well: massive data computing and implementation of algorithm rules are not human resources (or the cost is too high to be acceptable ). However, machines are dead and humans are active. Manual intervention by the search and marketing administrator cannot be replaced by any automatic optimization product: Setting effective targets, developing rule policies, and responding to special changes. Different levels of customers in different industries may have different account performance, with different environments and objectives for each customer. Therefore, there is no "best" Automatic Optimization product. The best performance can always be achieved with the efficient operation of search and marketing administrators with clear mind, whether through automatic optimization or manual optimization.