Data-driven web marketing and site operations notes

Intermediary transaction http://www.aliyun.com/zixun/aggregation/6858.html ">seo diagnose Taobao guest cloud host technology Hall

Today, in the fence net, I attended the second meeting of the Shanghai product manager. The "Mars Ren Jinxin" Talk of "data-driven network marketing and web operations" was heard at the meeting, and it was a huge benefit, with no paper and pens at the time, but only memory to record data-driven web marketing and web operations. something of value.

Now many companies or websites are in the data driven, data driven really good use is really good, but if you use the wrong, it is really wrong!

First, understand how the data came to ensure the authenticity of the data

For example, a company for email marketing, here's his data:

1. A total of 1 million messages issued;

2, mail arrival rate of 99%;

3, the mail opening rate 19%;

4, the mail link clicks 19%;

5, the ratio of new users is 80%;

6, came to the site after the site to stay for 19 seconds;

Judging from this data, there are usually several conclusions:

1, the message arrival rate is very high, almost no optimization space;

2, the mail opening rate is not high, may be the message title is not very attractive;

3, the link in the mail is not high, the message content users are not interested; There is also a great space for optimization;

4, the proportion of new users can also;

5, the time to stay on the site is very short, to show that the things users do not like, optimize the space is still very large;

So, before making this conclusion, if you ask about the method of calculating the data, you may get a completely different conclusion:

1, issued a total of 1 million messages:

What kind of issue is it? is the message sent out from your software? is the email address correct? Is there a mailbox that will not open for decades?

Well, the result may be that the value is exaggerated; maybe 1 million emails, only 100,000 are active mailboxes, if so, then the opening rate will be far underestimated;

2, the message arrival rate 99%:

We continue to cross-examine; the server is responding, even if it arrives? After arrival may be sent directly to the junk mailbox, there may be no spam mailbox, it was deleted by the server and so on;

So this value must be asked, the final result may be that this value is very large, in fact not;

3, Mail opening rate 19%:

The opening rate of the message tracking, usually in the Mail add an IMG, this IMG is actually a track code, so long as someone opened the message and executed this IMG display, then track code will be counted as open;

And what about the actual situation?

We know that some mailboxes in the open, for the sake of security, will not display the pictures in the mail, then will not be counted; In some cases, we use Outlook to send and receive mail, this statistic code will not be counted;

So, the opening rate of 19%, should be counted small.

4, the link in the message clicks 19%:

Link clicks is to come up with the mail opening rate calculated out, the former is a problem, then this is not to mention;

5, the ratio of new users is 80%:

Statistics of new users of the method, now the site is basically in the user came to the site, in the user's cookie record, then next time, first look at the user's cookie there is no record, if there is, counted for the old users, if not, as a new user;

The actual situation is that many browsers, there are some features in the closed, direct clear cookies, then the new and old users will have problems;

So, this 80% of new users, is the big;

6, visit the site after the site to stay for 19 seconds:

How does the site stay time count? The last page is open time minus the opening time of the first page; there may be other algorithms, but with the first algorithm, the time is obviously small; because the last page looks at the time is not counted;

So, this time is small;

Said, we look back to the beginning of the conclusion, is not completely wrong!

So, the data driver is driven to the wrong direction. So, before the data is driven, it's important to understand how these data come from, and the algorithms to know these technologies, otherwise, just look at the data do not know how to come, may lead you to a wrong direction up;

I remember that there was a picture on the PPT is very image, that is, pull the bow archery, the results are shot to the person himself, because the direction is wrong.

Ii. What kind of data does the business need?

Another example, in the market launch, a total of two sites, look at the data:

A website:

Put RMB: 1 million

Number of users bought: 2 million

Buy back pv:4000 million

Registered users: 100,000

As a single user: 20,000

B Website:

Put RMB: 1 million

Number of users bought: 1.5 million

Buy back pv:5000 million

Registered users: 80,000

As a single user: 25,000

First ask a question, these two websites which effect is good? If I think that the effect is good, I will give the other family 1 million also to the top of the house;

This time, it is more difficult, in general, will be before the launch, first set a goal, for example, my launch is to PV; if so, I will eventually choose b site, because the B site PV is the highest, the same 1 million can buy more PV; But the practice is better, there are many goals to calculate the effect;

For example PV accounted for 1 points, into a single account of 5 points, registration accounted for 3; then, according to buy back the situation, calculate a value out, so as to calculate the effect;

I remember there was a picture on the PPT that was a dice, and there were multiple faces.

Third, deep digging data

Another example, in the market launch, a total of two sites, look at the data:

A website:

Put RMB: 1 million

Number of users bought: 1.5 million

As a single user: 20,000

B Website:

Put RMB: 1 million

Number of users bought: 1.5 million

As a single user: 20,000

The last look at the data, two sites like, then is not really going to think is the same? Of course not;

To dig deep into the data, I draw a diagram to help understand:

If you look at the data on both sites, you will find that different, I can cite an example:

1, can see the amount of the order or we can get the number of profits, may be the two site orders are the same, but the amount of each single may be poor a lot;

2, can see the data a little bit longer, only to see the situation is so, if you look at the three-day data, may be different;

3, see B site in the landing page loss is very large, if we optimize the landing page, the situation will be very different!

Therefore, the data must be dug deep, look more detailed;

Above is today only down feeling very good several points, the record, also shares to everybody!

Article Source: http://www.zishu.cn/10/931.html