Big data is okay?If you have to think about it in depth, it might ponder the question: What is a good Nike Air Jordan 7 digital large? Orders will bring 100,000 dollars to 10,000 dollars of orders must be better than what? It does bring 1,000 new users will certainly bring more than 500 good? Is 1000 visitors than 900 visitors to valuable it? Text / guest author Xin After defining any network marketing data interpretation, we need to NIKE KD 6 NSW find a set of Metrics (indicators) to measure goal achievement situation. And Metrics To find a suitable three steps: first, to understand we monitor the number of significance, to find the right metrics to measure the effectiveness of our aim to achieve. The second step, research data, will be null and partially peeled false. The third step is to pass the test, the auxiliary index calculation and long-term monitoring to analyze the hidden and long-term effects. In the 'Internet Marketing: The Forgotten indicators and targets,' the article, the first step we have to be a brief introduction, and today we will discuss the second and third step. First release invalid data, we have to study the data will be invalid and partially peeled false. Because monitoring tool is not perfect, manual errors, or intentionally or unintentionally cheating partners, we get the data is not completely true. You get 10000 click, may be a real user clicks to get 100,000 registered users, are all machine may generate. - So, we have to have a screening process of filtering. Because so many methods of cheating, I do not think there is any systematic approach can be used to solve this problem. There is a simple principle: to find the exception. Real data that would seem more 'real' proportional relationship between the data will be more reasonable, data on the time and geographical distribution curve will be relatively smooth. In turn, cheat traffic may ill-considered, and thus to show abnormal in some way too far. For example, if I was monitoring me on www.MarsOpinion.com click advertising data, which may appear 'abnormal' place? Flow distribution in time exception: if you usually bring 10,000 click a day, today suddenly brought 100,000, which is very questionable. Alternatively, if the daily average of 10,000 still very stable, but a closer look found five thousand were at 8:05 assigned to 8:10 over five minutes, Nike Hyperdunk 2013 which is an unusual day. Traffic anomaly in behavior: Under normal circumstances, most of the traffic through advertising should be brought directly to the site leave (bounce), the rest of the traffic will Cheap Nike Free Outlet behave differently, some people see more than a few pages, some people see a few less. If you find that all traffic is bounce, which is an abnormal, if there is no traffic bounce it very abnormal; if the majority of the flow have shown a similar behavior which is very abnormal - for example, 90 percent of all traffic to the landing page click the link at the top of the page a humble, stay next page 0 seconds, the top of the page and click on an obscure link, then leave the station, which looks very fake. Traffic anomaly geographically: if 10000 click them, 9800 from the same IP (or an adjacent section of IP), which is very strange. If the point of view, a free shipping for Beijing activities to traffic 99% of Shanxi, which is inexplicable from the geographical analysis. Traffic anomaly in origin: the original buy is Nike Air Max www.MarsOpinion.com online marketing area of advertising, but it finally careful study found that 90% of traffic from the game zone points, which is very wrong. (Flow rate value is not the same in different places, maybe they are selling you advertising publicity when 'our audience happens to be your target customers, coincident high, good effect,' but in fact they are the aspects little traffic, In order to fill the numbers had from other popular channels - Import junk traffic spice) if we monitor not click, but the registration does - such as gaming channels, entertainment channels? Similar. Time Distribution exception: only 5 per day the previous month registered the day before the settlement came 10,000 registered. Abnormal behavior: Ads that link to landing page, but all traffic is registered directly to the registration page, without landing page. Regional exception: all from the same IP segment other anomalies: All registered users do not have to fill in specific information; or similar names to all registered users call; or all registered email registered users are looked like the more common is marsopinion0001 @ gmail .com, marsopinion0002@gmail.com, marsopinion0003@gmail.com, marsopinion0004@gmail.com, marsopinion0005@gmail.com, marsopinion0006@gmail.com... just above initiate, and can not cover all the characteristics of cheating traffic. But from the perspective Marketer, we judge Nike Hyperdunk 2013 the false principle flow is certain: if traffic is significant abnormality. If he can do it in the data flow performance and real no difference, we do it is difficult to catch them out. At least, after a good first step, we have to compare those data Nike Kobe 8 System MC to the silly cheat ruled out. On average companies do this step and stopped the - ah, to be exact, many companies have not stopped to do this step. Common auxiliary indicator if you have to think about it in depth, it might ponder the question: What is a good digital large? Orders will bring 100,000 dollars to 10,000 dollars of orders must be better than what? It does bring 1,000 new users will certainly bring more than 500 good? Is 1000 visitors than 900 visitors to valuable it? Not necessarily. Because although we measure the extent (quantity) campaign reached indicators, but without careful evaluation to achieve quality indicators. As if to say that we sent two people to different apple fruit stand to buy 50 dollars, my heart would like to buy more even if anyone Nike Air Jordan XX8 who is doing (with the 'weight' do indicators). Finally, bring back New Basketball Shoes 20 pounds apples A, B brought back 10 pounds, so we believe that some of the more capable armor. - Carefully think about it, is not that a little strange? A case Apple Kevin Garnett Shoes brought back difficult to eat (unpalatable, but not fake Apple)? In case B bought is the best food you'll ever eat apples it? In this example, we would like to use 'Apple's weight' to measure the level of two competent people, and later joined the 'taste' to assist the judge. Indicators for network marketing, we can similarly introduce some auxiliary indicators to help us better understand the 'quality' of those key indicators. Supporting indicators of the first class is able to get the current that can help us measure the quality of the main indicators Nike Air Max 1 of other indicators. For example, our goal is to 'sell', in addition to the amount of orders we can also be integrated orders, average order value, shopping with the view of the number of customers and profits. Suppose A website advertising has brought 10 million dollars in sales, B website advertising to bring 80,000. Order number A site was 500, from 500 different customers, with an average order value 200; and the number of orders B site was 800, from 800 different customers, the average order value 100-- this time, according to our different strategies We are likely to make different judgments. If our goal is to 'expand the customer base,' maybe we'll define the Revised Revenue = Revenue * (1-Buyer Weight) + Buyer * Average Order Value * Buyer Weight, if the average order amount is 250 of our site, and we give. ' Shopping the number of customers, 'the weight is 50%, Nike Air Jordan 7 then we can calculate A Web site advertising to bring the Revised Revenue = 100,000 * 0.5 + 500 * 250 * 0.5 = 50,000 + 62,500 = 112,500, while B website advertising brings Revised Revenue = 80,000 * 0.5 + 800 * 250 * 0.5 = 40,000 + 100,000 = 140,000, comparing the two websites advertising effectiveness B better than A website advertising effectiveness! But conversely, if our current goal is 'profit' (such as we are doing service, the service cost for each order is similar - such as 100, a single order buy less we lose money, buy more we earn more) and profit data from the WA in order not to give, we may have to use the Revised Revenue estimates simply look, for example, we can define the Revised Revenue = Revenue - Order * 100, use this formula to calculate it, we will A website concluded that better advertising results. All in Nike Kobe VIII all, we have to be adjusted in accordance with the strategic objectives defined in its Metrics indicators in order to make it more accurate reflection of the 'target reached in the case.' As another example, such as we want to be Campaign, aims to bring traffic to the site, then, time on site, browse depth, and has reached a particular page (such as a company presentation page, product page) can also become our assistant index . For example, we can define the following Revised Traffic = Traffic * 0.1 + 0.4 * (Traffic with more than 3 pageviews) + 0.5 * (pageview of target page), this way, that gives us high quality traffic (to the east on our website stroll around the West, or interested in our reading our Company Products) marketing channels will be easier to come to the fore. For example, our advertising has brought one million visitor, 10 万 browse more than 3 pages, 80 000 (10 000 people and in front of these repeat) people read product page, while offering 1.2 million visitor QQ ad, 50000 browse more than 3 pages, 60,000 Nike Kobe 9 Elite Low (and in front of these 50,000 have duplicate) people see the product page, the number of visitors at first glance like QQ better advertising, but carefully consider mass flow design with revised traffic analysis we will find not so thing. Sina Revised Traffic = 1,000,000 * 0.1 + 0.4 * 100,000 + 0.5 * 80,000 = 180,000, while the QQ Revised Traffic = 1,200,000 * 0.1 + 0.4 * 50,000 + 0.5 * 60,000 = 150,000, Sina advertising effect better! You can see from the above example, our goals are different, it will lead us to a different 'quality' of understanding, but also will need to use different secondary indicators to help us better interpret the data. Supporting indicators can not be ignored - the long-term effects of short, it is time to time, and then measure the traffic brought before a campaign, how customers behave. Or use the above example, our advertising 1000000 Visitor, QQ advertising to bring 1.2 million Visitor. After completion of advertising, after a month we can (if WA tool you use this function, many paid tools can do, GA then seen others used that way - but I have not personally operated) look at These people's performance, such as 'How many people after the advertising period also came to our site,' 'How many people in the last seven days came to the site,' 'How many people bought something after the launch of the' ...... Maybe we'll find , 1000000 Visitor Sina among 100,000 in advertising stopped again after the website, 50,000 came in the last 7 days (described high frequency of visits), 10000 bought something; and QQ 120 million of which only 8 Wan again website, 40,000 came in the last 7 days, 7,000 to buy something. We can have more confidence in the conclusion that better marketing results Sina really bring customers to the site, have long-term positive impact on us. As another example of two very similar, first, assuming that we have a bunch of forums and engage in activities to attract the forum's website to your site is registered send Credits and small gifts, it is easy to see the good in the early stages of effects - such as registered 10,000 a day, we think of acquiring each new user value is higher than 200 dollars, so when we've asked 20 dollars a registration time to we feel quite fair. Look at the data in the settlement, that registrations looks nothing unusual, should also be true, so very happy. However, when we look at the data in a month later, we found that only 0.1 percent of the 'new user' in the end of the event later visited the site, 0.01% of people visited 7 days, 0.001% of people shopping behavior - and natural growth 'real Air Jordan Outlet new user', all aspects of data performance should be at least 200 times the data! Put it together, the value of those new users only forum we usually natural growth of new users of 1/200, which is worth 100 dollars / month, bought 20 dollars is really a loss. Why is this so? One possibility is that our partners are cheating - do very clever, we did not see it, but he is also hard to think of how to do it again says that he continues to cheat simulation after the event so that these 'new user' to visit website, so this time is that we caught. Another possibility is the 'new user' is to be entirely Credits and small gifts to attract to register, they are just the instructions of the Forum just fill in a form to submit, did not bother to understand our site is doing, after natural It will not come to visit - this is the problem of the rule set. As another example, if we and a non-competitive large sites do users to exchange, in order to attract customers, we can give 'buy 100 minus 30' discount coupons for partners to bring new registered users. Because we are very strict: only 1 traffic coming from partner to get discount coupons by registering; 2 only after the user buy things we really need to pay the cost of the coupon... So sounds only of losing business - which got 100,000 new users through this partnership, 50,000 to buy something, we will be very happy in my heart, MKT can also submit a report to the boss came pretty demonstrate how well they are doing. But, if it be sensible in a month later to check data, you may find that 50,000 'shopping new user' only 1% again visited the site, and normal data should be at least 10%. Why the gap? After you carefully search the Internet may find all the online shopping forums have posted a message saying 'www.MarsOpinoin.com the money to buy 100 get 30, by re-registering an account like this!' And countless old users (typically, this post has been the most attractive shopping marsopinion.com before, I believe the user to this site) click on the link re-register an account (this use is that they are not commonly used in email, or a new lane an email to register) took the coupon to buy his own had to buy something, put the account after buying lost. The Campaign Plan is not a good deal we could talk about, but its effect is not as good imagination is certain. We can see that the secondary index not only can help us be more precise measure of network marketing, measure its true role in 'to reach the goal', but also to a more comprehensive and accurate data filtering cheating to some extent, do both.