Under the AI wave, the convenience of voice input has greatly improved the user experience, and has become an area of input method products.
Editor’s note: This article is from WeChat public account “Technology to order” (ID: xiangling0815 ), the author has bellmed.
Under the AI wave, the convenience of voice input greatly enhances the user experience, and it has become an area for input method products to compete for.
However, the three recent reports on the market competition of voice input methods should have become an important reference for us to observe the industry, but they have many contradictions between them, which is puzzling.
-
The “authoritative” CCTV market research organization CTR released the “2019 mobile phone input method voice input function market research report” (hereinafter referred to as CTR report), the three input methods led by Sogou, Xunfei and Baidu Occupy nearly 95% of the market share, Among them, Sogou market share of 76.9% exclusive 3/4, daily average voice request 584 million times, more than the sum of Baidu and Xunfei two times;
>
-
The third-party mobile phone input method research report of China in the first half of 2019 issued by iiMedia Research, a third-party data mining and analysis organization of the world-famous new economic industry (hereinafter referred to as the “Ai Media Report”) It is said that in the first half of this year, Sogou input method and Baidu input method users ranked the top two, accounting for 45.5% and 41.0% respectively. Baidu input method has the fastest monthly growth rate, and the voice input user scale reached 271 million. Ranked first in the industry, the daily voice input per person is the most, and it has become the strongest voice input method in China;
-
The other “authoritative market analysis organization” Sino Market Research also released the “Analysis of the Voice Input Method of 2019” (hereinafter referred to as the Sino report), said that the flight to maintain the voice input method leader Status, voice input awareness of up to 90.35%, far ahead of other input methods such as Sogou, Baidu, etc..
In all reports, everyone recognizes the market structure of Sogou, Baidu, and Xunfei. However, in the case of market share, especially voice input, these reports each have their own “first” method.
There is no doubt that someone is lying, at least deliberately confused. The origin of each number cannot be falsified, and the reporting agencies each have their own rhetoric, but from the logic of the report itselfAnd the comparison with the data of external independent institutions, etc., too exaggerated arguments can always show the original shape.
What kind of report statistics are the most scientific?
To make a market research report, in the simplest terms, it is divided into three steps: Select a statistical sample in one way, then count the data in some way, and finally add some institutions. Evaluation.
In the three reports, it seems that the CTR report has investigated more people and the results seem to be more reliable.
However, only from a statistical point of view, when n≥30 or at least n≥3(k+1) (k is the number of explanatory variables, ie the number of indicators to be described), the sample size is sufficient Sufficiently, the minimum sample size of the three major reports has reached about 2000, and the impact of the difference in quantity between the two has been negligible.
At this point, what affects the statistical results is how the sample is selected, how it is calculated, and whether the sample matches the statistical interval.
1. How are samples sampled and calculated?
The CTR report marks the survey at the bottom of the homepage with the survey of China’s Internet information center as a sample quota setting. This is an obvious single source random extraction method, and does not specify the calculation model:
The Ai Media report on the home page refers to data from the Ai Media Information Business Intelligence Database, the Polaris Internet Product Analysis System, the Strawberry Pie Data Survey and Computing System (Strawberry Pie) and other sources, using market research on the calculation model. In-depth industry interviews, desktop research and other methods, and use the major data computing systems and related computing models of Ai Media Information:
The Sino report is called “simplification”, so there is no mention in this regard.
There is a very simple reason in statistics. When the sample size is low, increasing the sample size is to increase the accuracy; and when the sample size is increased to a certain extent, the sample selection method and the calculation model are accurate to the result. The impact is getting bigger and bigger until it becomes the deciding factor.
When the sample size exceeds the threshold, the sample supported by the scientific calculation model can maximize the data of the restored market. In this sense, the CTR report has 4,000 samples but data. The single source and the “ambiguity” of the calculation model have greatly reduced the credibility of the report.
2, how is the time interval of the sample and statistical conclusions defined?
You can’t take a sample of A time to get the conclusion of B time, which is basic common sense and does not even require statistical thinking.
Only from the title, the CTR report is somewhat “weird”. In September 2019, the report was written as “2019” for the whole year, while the Ai media report was “smart”. It only applies to the “first half” of 2019.
It’s strange to say that CTR’s report last year was written for 2018. The release time is a few days at the end of 2018. It makes sense. Why is it that this year will be in a hurry to sum up the whole year in September, and its credibility will hit again? Discount:
Correspondingly, the Sino report is only a brief analysis of 2019. It seems that the world is only talking about 2019 during 2019, and it is not expected to run ahead.
External contrast, is the number scientific enough & consistent with normal cognition?
A report can certainly give some shocking conclusions to highlight its value, but it is too different from the data of external independent authoritative third parties, or it clearly deviates from the general public’s perception, and obviously does not have credibility. degree.
The CTR report said that Sogou, Baidu, and Xunfei accounted for 95% of the market share, but in terms of specific allocation, Sogou was nearly 10 times higher than Baidu, and nearly 8 times higher than the news:
However, according to the widely-recognized professional market research company iResearch, the latest monthly installed capacity of Sogou, Baidu and XunfeiAccording to this, it is very different:
On the ranking of one or two, Sogou is indeed ahead of Xunfei Baidu, but it is obvious that the data is far less than 8 times and 10 times in the data, and Baidu’s input method is A larger proportion exceeds the Xunfei input method. Even though the data of iResearch may not be accurate, the report of CTR is too exaggerated and unconvincing.
Compared with this, the Ai media report and the iResearch data (the two have a competitive relationship, the reason is not linked to each other) is closer, the first is still Sogou, but with two or three The gap is in a smaller range:
The data is not completely compatible, but it may not be accurate, but the Ai media report on the big surface is more rational. The CTR report is over-exaggerated. Some of them are impatient to highlight the “first” status of Sogou. However, who Nor did it deny the position of Sogou’s market share.
In these three reports, the Sino report has avoided the market share of Xunfei. In general, the third position of Xunfei is also recognized, and there is a certain gap between Sogou and Baidu.
In addition to the market share data, Sogou claims to have 584 million voice requests per voice input per day, which exceeds the frequency of Baidu and Xunfei. In this respect, there is no similar external mechanism of iResearch, but Sogou input method. The leading voice of Baidu and Xunfei in voice is also obviously contrary to the general perception of the industry and the public. After the CTR released the report, some netizens questioned (screenshot source: Technology Times):
Although netizens’ comments are too extreme and not worthy of all, at least, these concentrated public perceptions show the deviation of CTR reports from popular perceptions, and at the same time Expressed the degree of recognition of Baidu and Xunfei input methods in terms of voice.
The logic is correct, have you done it?
A market research report, each data is difficult to break through, but the report itself is at least logically self-contained. If there is a contradiction between what he said before and what he said later, then the credibility will disappear.
That is, even if we leave the truth of the various data undiscussed, analyze whether a report is credible, at least from the report whether it has been consistent.
For example, in the media report, Baidu leads the industry in voice input, both in terms of user size and daily voice input:
These numbers are not as high as credibility, but at least, the histogram is not much different from the user share of Sogou, Baidu, and Xunfei in the same report, but Baidu voice input is in its users. The ratio and usage are slightly higher, multiplied by a reasonable multiple to get here first. At the very least, there is no obvious problem with the consistency of the logic before and after the report itself.
Review of the CTR report, If the Sogou market share mentioned in the first half of the report is nearly 10 times higher than that of Baidu, it is nearly 8 times higher than the news. When the second half talks about the voice, the voice input is daily. The frequency of requests, why Sogou is only one more than the flight, Baidu more than one point:
According to the logic of the CTR report, the only reasonable explanation is that the voice usage of Xunfei and Baidu is far better than that of Sogou, so it will be used in the case of market share far less than Sogou. Achieved “catch up.” However, in the same CTR report, it is mentioned that Sogou users use the voice input function for the longest time each day:
The problem is coming. The market share is far ahead, and the user’s voice usage is the best. According to the reason, the daily average request should be widened again, from 8 times, 10 times to 9 times, 11 times, etc. Waiting for it, why is the gap greatly reduced, only one double?
Let’s look at the report of Sino, and only talk about the user’s voice input ratio. The proportion of voice users in Xunfei is the highest, and the voice recognition rate of users in the second half of the report is also the highest. These two indicators are inherently consistent. If there is a cognition, there will be a proportion of use, at least report that there is no logic that is not right.
Entering to subvert yourself, the leader must first abandon the old competitive thinking
An industry research report that not only has data, but also has an insight into the future.
In terms of voice input, technology is the thing that determines the future. It is better to look at the data in front of you (not necessarily right) to explore the future technological potential.
Technical competition, seeing potential rather than immediate utilitarianism, technology does not lie, the best use can ultimately win the user.
Baidu and Xunfei are competitors on the AI track. Correspondingly, the Ai Media Report and the Sino report also emphasize the voice technology capabilities of the input giants. The media reports are as follows:
Sino Report is as follows:
While looking at the CTR report, it does not talk about the specific voice industry comparison, but is constantly emphasizing data, proportion, and market possession.
This is not wrong, but as the CTR report said, voice input has penetrated into various scenes of work and life, and the needs of users have become increasingly diversified. Input method as an “old” Internet product, under this wave is facing an excellent reshaping of cognitive opportunities, changing the past simple tool product positioning, bringing more business model space.
Ai media report, gave AIInput method application innovation brought about by input:
Among them, similar volley handwriting, smart chat assistant can be separated from the scope of mobile phones, PCs, open a broader intelligent terminal application space for the input method.
In any case, the new story and new hope of the input method is to subvert yourself to create a new product form, which requires the leaders to abandon the old and backward competitive thinking, no longer entangled in market share, but greater To the extent of innovation. The competition of input methods is likely to open up a new track instead of competing on the original track. This is also the reason why Baidu and Xunfei, who are ranked second and third in market share, are actively engaged in voice input. Where.
The CTR report over-exaggerated the Sogou input method, which was already at the top of the market share. In September, it launched a 2019 annual market report, and there are still some logical mismatches in the middle. The extent may reflect the overall anxiety of Sogou.
In early August, Sogou announced its Q2 earnings report for 2019. The total revenue was $303.6 million. The net profit attributable to Sogou was $21.3 million, compared with $38.2 million in the same period last year. On the one hand, the absolute value was low. There has also been a significant drop. On the day of the release of the earnings report, the stock price of Sogou, which was not so prosperous, fell by 11.6%.
Sogou or is in a quagmire of growth stagnation. At this time, whether it is necessary to “stabilize the military” for internal, external and investor investors, the CTR report may be the best placebo in September. The mouth is thick. However, after the integration of the input method and the AI, it is necessary to take more leadership products into the future, and the placebo should not be used as an anesthetic, which blinds the pursuit of technology and the pursuit of the future.