In the era of voice search, how to use smart speakers to sell goods？

For brands, search engine optimization SEO (Search Engine Optimization) is the most basic marketing method in the Internet age. By improving the engine weight through optimization, the higher the ranking in the search page, the greater the chance of obtaining customers. In the PC Internet era before the rise of mobile devices, the search engine company "absolute monarch" not only earned hundreds of billions of dollars in revenue, but also cultivated an industry specializing in SEO.

The tide rises and falls. With the rise of the mobile Internet, although the old search engines still occupy the terminal market (Google bought the iPhone default search engine for $9 billion), changes have already appeared. Apple’s Siri started the intelligent voice assistant, and Google Assistant came from behind; Amazon used Alexa to trigger the smart speaker war of technology giants. With the help of intelligent voice assistants and smart speakers, voice search has gradually become a major way for users to search because of its convenience.

Experts from eConsultancy predict that voice search will account for 13% of Google’s global search volume and 20% of mobile search. According to the statistics of SEO Tribunal, the search volume of Google in 2018 was 2 trillion, and the "hits" on behalf of voice search were at least 250 billion times.

13% seems like a small number, and 250 billion is definitely a big number. More importantly, voice search’s results are very different from traditional web search. How to do a good job in SEO for the increasingly important intelligent voice search has become an important issue for market participants.

Recently, Voicebot.ai, a foreign website, and Voice Insider jointly launched a report. The two teams used 4,000 questions to test the mainstream intelligent voice assistants and voice devices in the market, such as Amazon Alexa, Apple HomePod, Google Assistant, Google Home and Samsung Bixby, and analyzed the tendency of the mainstream voice search on the purchase of brands and products, and found some very interesting things.

Conclusion first:

1 Google performed best in the test.

2 The stratification of content sources in voice search is obvious.

3 Most companies have not paid attention to voice search optimization.

Third-party content such as Wikipedia and Yelp needs to be taken seriously.

5 Several major voice search systems have not performed well in pouring traffic into independent voice apps.

Please read down for the detailed analysis and results of the test. At the same time, GeekPark (ID: GeekPark) did a small test on domestic smart speakers in the way of Voicebot.ai, and also found some interesting facts. Curiosity can be pulled to the egg link.

Because of Google’s dominance, search engine optimization mainly refers to Google’s search optimization. At the same time, Google Assistant and Apple’s Siri Intelligent Voice Assistant both refer to Google Knowledge Graph, so people think that voice search’s optimization is no different from the previous Google search engine optimization.

In fact, search engine optimization has been changing, and the search results of intelligent voice assistants are often very different from those of text search.

One of the most easily overlooked factors of engine optimization change is the fragmentation of search. According to eMarketer data, in 2018, 46.7% of searches for goods in the United States came from Amazon, higher than Google’s 34.6%; Google’s Youtube is the world’s first video search engine and the second largest search engine in the world. Facebook has announced that in 2017, global users conducted 200 million searches on Facebook every day.

In addition, according to different devices, the search results are quite different. Generally speaking, the screen size determines the number of search results: computer-based web search usually displays 10-35 results; 1-15 search results can be obtained by using the mobile intelligent voice assistant; 1-10 smart speakers with screens; However, most smart speakers without a screen can "display" only one result.

At present, the search through voice assistant usually falls on the following major websites: Google /Bing Knowledge Map, Wikipedia, Yelp and Yext. These search results look similar to web-based search, but they actually hide some "small thoughts" of technology giants.

If you ask Alexa of Amazon how to treat the flu, the latter will provide the contents of the Mayo Clinic. Sometimes Alexa will provide answers uploaded by other users. If you ask for product-related information, Alexa will directly use the content of Amazon website. Similarly, if you ask Siri which is the best tablet in the world, there is no doubt that the answer is iPad. Even Google Assistant will "intercept" a part of its knowledge map in response to the company’s related products.

When using voice assistant to search, the latter will give priority to accessing multiple data sources instead of going straight to the knowledge map like web search. These data sources include some "reserved items" of giants, internal and third-party information sources. Compared with knowledge maps, these informers have obviously higher data priority. With the increase of optimized content for voice assistant search, the number of times that the knowledge map of web search is cited by voice search will be less and less.

Due to the gradual differentiation between voice search and web search, the influence of traditional web-based search engine optimization on voice search is gradually weakening, so the marketing personnel who conduct SEO need to take "special care" of voice search, which is gaining momentum.

Voicebot.AI conducted 4000 questions on voice assistants and devices such as Google Voice Assistant, Google Home, Apple HomePod, Amazon Alexa and Samsung Bixby, involving 21 product categories and 200 product brands. As a result, the Google Assistant who changed the relevant algorithm directed more traffic to the third-party Google Action, while other voice assistants still provided the original knowledge map and third-party information sources.

Among them, the product category covers: what is smart TV, where is the nearest gas station, where can I buy auto insurance, etc. Brand problems include: what is a certain brand, how to contact it and how to buy it. In order to make a better comparison, many of the questions come from Google’s search recommendation on the brand, which can better reflect the universality of user problems.

In the survey, 26.5% of the 200 brands have independent voice applications in Alexa, and this figure is 21.5% on the Google voice assistant platform. The brand share of independent voice apps on both platforms is 15%. It should be pointed out that the focus of these voice apps is not to answer the common questions of users about brands and products. This means that 75% of brand content comes from third-party resources.

This is not good news for the brand, because most of voice search’s answers come from Wikipedia or Yelp, and the content provided by the latter is not all praise for the brand.

According to the test results, the correct response data given by Google Voice Assistant reached 92%, followed by Google Smart Speaker Google Home, with a correct rate of 81%. Amazon Alexa, Samsung Bixby and Apple HomePod rank in the bottom three, which is obviously different from Google.

It is worth noting that the failure rate of other voice assistants is not entirely due to the ability of natural language processing (NLP), on the contrary, it is just unable to answer the test questions. Apple’s HomePod was the coldest in the test, and as many as 63% of the questions were rejected as "unanswerable". In contrast, Samsung’s Bixby is enthusiastic and has the highest error rate in providing answers.

In the test, Alexa of Amazon is the voice assistant who likes to quote his company’s knowledge map best. Especially, when the question contains the word "the best", 100% Alexa will provide the Amazon『s Choice "rating, selling price, buyer’s high score evaluation and sales volume of related products. So, for Alexa, this is indeed the "best", but the background platform is Amazon shopping website. Because Amazon itself has a deep accumulation of product information, Alexa is able to answer product-related questions with ease.

Another feature of Amazon voice search is that the answers to some specific questions are not from the knowledge map, but from Amazon users. At the end of 2018, Amazon launched the Amazon Alexa Answers program. According to the data, Amazon users provided answers to 100,000 questions, and these answers were queried by Amazon users millions of times. However, in the test, Alexa only provided answers uploaded by three users for more than 800 questions, so the data of Amazon’s "Everyone Ask Everyone" project is still not big enough.

Wikisego

Similar to web search, Wikipedia is the first source of all voice search’s brand content. When faced with the question "What is a certain brand?", the correct rate of all voice search links to Wikipedia is above 90%. However, what the brand needs to ensure is that the first or second sentence of Wikipedia entry should accurately summarize the brand content. It’s easy to understand why voice search relies so much on the content of Wikipedia on brand-related issues, because the latter can guarantee the accuracy of the content.

About contact information

In addition to understanding the brand content, consumers sometimes want to contact brand companies for more services. In the test, when faced with the question of "how to contact a certain brand", with the search advantage, Google voice search once again crushed its opponent with absolute advantage. The correct answer rate was as high as 96%, which was nine times that of the second Amazon. Apple HomePod gave 99% of the "I don’t know" answer, and Samsung Bixby gave 100% of the answer error rate.

However, the reason for Samsung’s 100% error may be that the contact information is confused with the user’s own contact list, so that there is a blank page, and this problem is likely to be corrected later. Amazon’s questions come from American Yellow Pages, Yelp, reference.com and wikiHow, so there are 10% correct answers. The information of Google voice search not only comes from its own knowledge map, but also draws lessons from GetHuman.com and wikiHow, so the brand should also pay attention to these third-party content platforms.

Local shops and services

Of course, knowing the brands and commodities, what users may need next is where to buy them. In this category, Google crushed its opponent again, and Google Voice Assistant got 94% correct rate, while Google Smart Speaker got 71.5%. In contrast, 7% of Apple Homepod’s correct answers came from Yelp website, and Samsung Bixby rejected all questions about location.

As an e-commerce giant, Amazon Alexa’s performance in this round is not satisfactory. Although many answers have been drained to its own e-commerce platform, when the goods are not items on Amazon’s website, the answers cited by Alexa are not wrong or can’t be answered.

Google’s eye-catching performance in this round is that it not only gives the location, but also gives the inventory and price of the goods in the store. Because in Google search, the store’s inventory and price are one of the contents that can be filled in, which is convenient for users to compare prices, which surpasses the opponent by one position.

Wikipedia is an important source of brand content

If you don’t develop Alexa skills or connect to your own third-party content, it’s best to ensure that the first or second sentence about the brand on Wikipedia can describe the brand ideally.

2 Location search depends on Yelp.

In 18% of the correct answers about the location, Alexa used the content from Yelp. At the same time, if you can’t find the location, Alexa will provide the main content about the product.

3 Amazon selection is very important

Many questions about the category of goods will be linked to Amazon’s selection Amazon『s Choice, which means that this part of the content is an important part of Alexa. If we can change the content of Amazon’s selection, it will also affect the product answers in Alexa voice search.

1 The inventory of local goods is very important.

Since the end of 2017, Google has linked the inventory of local stores with Google Assistant, which currently includes a variety of commodity categories from beer, cosmetics to digital products. If there is enough information, these items will often appear on the first page or in the optional list. When Google Assistant asks users if they need more information, the retailer information will appear in voice search’s answer.

From this point of view, large supermarkets such as Wal-Mart, Target and Best Buy are doing very well. Strangely, stores like Total Wine and More don’t have as many liquor dealers as Wal-Mart, and they often become reference answers when they encounter searches about alcoholic beverages. This means that you don’t need as many stores as Wal-Mart, and you can still have an advantage in voice search.

2 information enrichment has a better chance to rank among the top three.

When users search for "XXX around me", Google Assistant usually lists the top three, and this information is usually quoted from "Google My Business". The trick to getting on the list is to list enough details in the introduction of your store, such as providing children’s menus and personal customization. Experiments show that the shops listed in the top three of the list are not necessarily the closest to users, but often those with more detailed information.

3 Google Assistant loves lists and doesn’t love FAQs.

Google Assistant likes to use lists as answers rather than FAQ pages, because the layout of FAQ pages is difficult for voice assistants to understand at present.

Google Home, a Google smart speaker, usually refers to the blog content directly containing the brand name list, which means that the relevant content does not need to be made in the form of "adaptive voice".

Good case

When asked "How to find the best skin care method for you", Google Smart Speaker gave the answer that Dermstore, a cosmetics retailer, wrote a blog "In what order should I use cosmetics", and then listed seven steps step by step.

Bad case

When asked "How can I contact Adidas", Google Smart Speaker did connect to official website of Adidas. Unfortunately, this is a standard FAQ page, so the user heard: "The following is m.Adidas.com’s summary, product 1, basketball preparation; 2, cologne and body fragrance … "

What about Siri and Bixby

Siri, Apple’s intelligent assistant, failed to provide too many clues to the marketing personnel in charge of SEO. The test results show that Siri relies heavily on Wikipedia for common sense issues about brands. All local business issues are left to Yelp. At present, there is really nothing you can do except optimize these two content sources, because the answers to other questions are: Sorry, I can’t answer this question on HomePod, or "I can’t find a similar place". Therefore, at present, the best SEO advice for Siri is to optimize the content of Wikipedia and Yelp in addition to standard SEO.

Compared with the other three smart assistants, Samsung’s Bixby is still a younger brother. At present, it only iterates to version 2.0, and its performance is not as good as that of its opponents.

1 Optimize the content of Wikipedia and Yelp files

Reason: Wiki and Yelp are the content that all searches will use.

2 Apply for Google Express or find a way to get into Amazon Select.

Reason: Therefore, it can be an optional answer when users search for purchase options.

3 Use Google Local Commodity Information Flow

Reason: When users ask about local commodity purchase information, complete information details will be a powerful choice answer.

4 build Google My Business file

Reason: When users search for local services, they are more likely to appear in the answers.

5 release voice App, or more.

Reason: Google and Amazon’s diversion for independent voice apps is not very strong at present, but the big platform will continue to iterate their related algorithms, making independent voice apps the first choice. In addition, the voice App is best only for some keywords about products, or only for specific topics.

6 add CFIR and Implicit Invocations to voice App.

causeCFIR and Implicit Invocations are related projects of Amazon and Google aimed at voice search respectively. At present, the effect is average, but they will be useful gradually, so we should seize the opportunity.

7 using Alexa Conversations to build Alexa skills

Reason: Give Alexa more signals and give your voice skills a greater chance.

8 Optimize the information page of aggregation websites

Reason: Aggregation websites like WikiHow and GetHuman can help voice search connect directly to your webpage.

9 using Schema to build data tags

Reason: The microdata constructed by Schema.org can better highlight the important information in web pages, such as contact information.

10 provide a more concise answer

Reason: The playing time of 29 words from text to speech is 10 seconds to 12 seconds, which is a relatively standard time for users. Try to answer 5 "W’s" in voice apps and web pages.

All the above are foreign voice search results. You must be curious about what answers voice search and intelligent assistants in China will give if they test.

As there is only one Baidu smart speaker at hand, I conducted a simple test on Baidu voice search with the methodology of Voicebot.ai, and found the following:

1 brand issues, Baidu Encyclopedia is the main source of content

As for product problems, Baidu’s own short video platform "Second Understanding Encyclopedia" has sprung up, but the matching degree between answers and questions is not high. Ask, "What is the iPhone? As a result, electronic music was played for one minute (it should be a prelude to a short video).

3 Questions about "where to buy" are divided into Baidu know categories, and the high-priced answers provided by netizens are more naive.

4 "The best XX product", the only problem that is divorced from Baidu’s knowledge map, refers to the recommended information of this product from Pacific Computer Network and another unknown website respectively.

Editor-in-Chief: Sleeping Worm

Image sources: Voicebot.ai, 9to5Mac and Apple, Google and Amazon official website.