The misunderstanding of the data

As a data analyst, do you really know the data? There are three kinds of lies in the world: Lies, terrible lies, and data. And for the electric business industry, statistical thinking one day and marketing creativity is equally important.

Is the data really worth our trust?

If your Web site is focused on selling your bags. The SEM manager told you that the ROI of this month's SEM has increased by 23% in the same budget, and your website analyst has told you that the average purchase amount for this month's site has increased by 20%, and we are convinced of the performance of the two-person. Because the data proves everything. and is it worth your while to be really happy?

Have you ever thought of the site's SEM roi 23%, a single purchase increased by 20%, what is the real reason? We have a rational vision of several possibilities: 1, whether seasonal factors affect (this topic for discussion in the next article): Summer is here, Ladies need to change some color more beautiful bags to match clothes, seasonal and lead to a single sales increase, resulting in SEM roi promotion, but also lead to the promotion of website conversion rate. 2, also may be your website analyst to the commodity carries on the more accurate related recommendation, causes the product the single time purchase amount to increase, originally the customer to the website is wants to buy a keyboard, but your website is more humane, at the same time reasonably recommended the related product. For example: mouse, headphones and other products resulting in a single sales increase. Through the optimization of the site to upgrade their own single purchase volume led to the ROI of SEM. 3, there may be your SEM manager carefully analyzed your Baidu or Google account to make the keyword more accurate, matching the more corresponding landing page to improve the roi?4, of course, there will be your site analyst to make the site's shopping process faster and more convenient. Increasing the conversion rate also increases both data. Of course, there are so many possibilities, and you can totally imagine. So the single data we can not determine is because the market outside of the reason or your staff efforts or not to work. It is also not possible to determine which department employees credit.

Can the data be fully representative of the facts? We have to think more.

Second, do we really understand the concept of data?

Average? We want to give an example to understand the concept of the mean: a city, 1000 people, annual income per capita 80W yuan, very NB income it? This data gives us the feeling that the city's living standard must be superb, a leisurely, rich life picture cannot help to emerge in front of us. But when we go to this city, we find that the quality of life here is as good as we think, and most people are still not well-off. This time, not only do we have to wonder, is there a problem with the statistical data? But it is true data. So why is this? is the people formed the culture of frugality? or what? Finally found that the city lived two "coal bosses", the annual income of the kind of billion, they are the level of the average person to pull higher. So do you believe in averages? Do you really understand the average? In China, we can abuse the statistical data of the Bureau of Statistics are not allowed. But you know the data distribution, do you still abuse it?

So how do you understand the average stay time on the site?

Three, do not subdivide we will see the truth?

We know that for the site/landing page there are two very important indicators that is the bounce rate and average stay time. Bounce rate means that the person who arrives at your page will turn off the page without doing anything, OH, no! This is definitely not a good news. The other is the average stay time is the average number of people who get to your page. Well, of course, the bigger the number, the more attractive your site is to customers. With these two values, you can clearly know which media flow is more valuable, and you can judge the site: those pages are designed to be more customer friendly, and of course you know that your products are more attractive to customers. But for SEM, you must carefully subdivide the data. For example, you Baidu sem bounce rate is 60%, the average stay time of 2S, how about? Did you take a closer look and find out that a certain plan was 96% out of the way? and to see if the bounce rate for other projects is only 50%, then your focus on optimization is found. So do not believe in the overall data, only in the eyes of "overall" we can not find problems.

Four, the difference of data samples will distort the results?

The probability of throwing a hard skin on the opposite side. So you do a test, 5 consecutive times, are positive, do you judge the probability of the positive is 100%, the opposite is zero. Is such a sample reasonable? Of course we know that the more we throw, the more likely this probability is to be 50%. But do we have a similar approach when we do A/b test when we do it five times before we cast a coin?

Have you ever tested IQ?

So your IQ is 132, which means you're smart?

Smart you, you have to make some comments!