Web analytics and Chinese comics: how to make sense
- Holmes, web analytics shows that we have a high bounce rate. We turn up!
“Watson, but what does this have to do with payback?”
Web counters generate a huge number of digits, summaries and graphs, giving rise to the unbearable ease of being a marketer. So he read 10 tips on how to increase conversion in Yandex.Direct, looked at his charts and began to redistribute the budget until it runs out. Man, stop it! You did everything wrong.

Experience with clients and colleagues has shown that many people read metrics incorrectly and, as a result, make the wrong decisions. As a result, they simply master the budget, instead of making it work on a high ROMI. We came to talk about what is hidden behind the zeros and columns.
All examples describe real events, no abstract comparisons, and tips are easy to scale for small to medium budgets.
Go.
Client: I saw in the web counters a bounce rate above 50% for a number of ads and requests, this is a lot. Let's refuse to work with these requests, we are losing money!
Web Analyst:A high bounce rate still does not mean anything. For all topics and sites, the concept of a "high failure rate" is unique, and you can only understand what your number is by trial and error. The main evaluation criteria are payback and conversions. Let's take a closer look at the ads in this campaign. We worked with semantics containing the names of competitors' brands, and the campaign had the highest bounce rates - this is logical and predictable. However, such requests have low rates and they cost a penny (literally a penny), and revenue is included in the top 10 most profitable campaigns. Conclusion: inquiries should not be refused, it may make sense to rewrite the text of the ad to drop out the target audience.

Client: Why do you offer me to give contextual ads on request, where is the name of my brand? They will find the company anyway, it will be issued. I think this is a waste of money.
Web analytics: Indeed, it is still considered strange to buy contextual advertising for your branded search queries, because the benefits of this are not obvious. And you buy, while your brand traffic is not intercepted by competitors. Firstly, the situation from the first example is very revealing: if you do not take your place, this will be done by a competitor and will lead away “cheap” customers who actually looked for you.
Secondly, clients of companies from certain subjects may not understand where the search results are in the search results, and where are the ads (for example, embroiderers).
Thirdly, on the page of search results from 10 places of presence. It is important to fill them as much as possible. First of all, this is a free presence due to organic issuance (including through social networks and “Images”). You have secured 4–5 places. But why not take another place? Brand traffic, again, is inexpensive, but as efficient as possible. Total, a minimum of 50% of your issuance, and the chance that they will come to you is 1: 2 in your favor.

Client: We ordered and paid priority placement on Yandex Maps for three months - of course, there is traffic, but - the income is minimal, conversion is low, ROMI is negative. I see no reason to develop this tool.
Web Analyst:Evidence of geolocation performance does not need to be looked for in the meters - they are not there. People use maps for the same thing as 100 years ago - to get there, get there, find yourself. Indirect indicators of successful work with geolocation: we observed an increase in live traffic to online and offline stores (online traffic grew by 3.6 times in a month), tracked the effectiveness of the channel through call tracking using phone numbers delivered specifically for geolocation cards. Bottom line: we began to work more closely with traffic through geolocation, we encourage customers to leave feedback in cards, periodically change the description for current promotions.
The growth of traffic on the channel "Maps"

Client: I looked at web analytics and saw that email is among the top three tools with the lowest conversion. We are sending out something, but people are not very active in buying. Let's turn off this channel, concentrate on Direct.
Web Analyst: The tangible returns in numbers from email and display ads will always be dubious. These are tools from a fog which are almost not visible in counters. For example, the other day we turned off a newsletter dated January 16. Rechecked, indeed from January 16th. The buyer can read the letter, postpone it, change his mind, forget, remember, save up and buy. Buy from another device, from another mail, at another time, and the counter will not show that it was sent by the mailing list.
It’s even more complicated with the media: it is shown mainly for the sake of stimulating purchases. This is an image advertising that works for qualitative indicators (the formation and retention of interest, loyalty, etc.), and not for quantitative ones. But the exhaust in numbers can be tracked indirectly: with the launch of display advertising and email marketing, organic brand traffic should confidently grow. On our graph from Wordstat we can clearly see how brand traffic began to rise with the beginning of work in October to increase recognition.

In addition, you must consider the features of the buying process in each product category. For example, in the topic “large household appliances”, people place orders mainly on weekends, carefully studying offers, consulting with their families, planning a budget place and delivery time, etc. A long maturity period of customers does not allow to accurately track all orders: a letter with the offer may come to the mail of the father of the family, and the order will be made in 2 weeks from the grandmother's phone. However, knowledge of the characteristics of the audience still allows for more precise adjustment of advertising campaigns. For example, we began to announce stocks 10 days before they start, so that the client “ripened” a certain time before the purchase. Indeed, the “tail” of discoveries and clicks stretches stably for 2-3 weeks.
Schedule of heterogeneous daily activity

3 month tail

Client: Why do people spend so little time on the site? Here is my friend’s average viewing depth of up to 20 minutes! Let's redo the site, it’s bad, people don’t like it.
Web analytics: Depth of view is another indicator that can be interpreted in different ways, as well as failure rate. Poor usability can be hidden behind a large viewing depth: people persistently look for a feedback form throughout the site and do not find it. And in a parallel universe, customers find the information they need in 30 seconds on the first page and, satisfied, leave. Meanwhile, the counter registers a small depth of views.
Client: I calculated ROMI for contextual advertising, this tool is unprofitable, let's abandon it.
Web Analyst:It is worth digging deeper and separate the grain from the chaff. It’s possible (as it was with us) that you should turn off not the entire campaign, but individual unprofitable ads, or maybe just some separate request in this ad. It is highly likely that unprofitable ads pumped a lot of money out of the budget, sharply overturning the payback of the entire campaign minus. This problem is not solved by analytics, but by continuous testing of the entire pool of ads for each campaign. We started with a pool of 34 ads, during testing and rigorous selection we turned off 31 low-performing ads, leaving only 3 for the entire campaign, but with ROMI up to 755%. The budget was redistributed between them, and the campaign began to generate revenue.
PSWhat do we mean by all this? Analytics needs analytics. Namely, in competent analytics, which he will look with his eyes, he will think with his head and will refer to logic, and not to the figure in the summary. Because a figure in itself means little.
Last tip on how not to do:
If you are faced with a strange or one-way interpretation of the data from the counters, welcome to the comments - share your impressions and misconceptions.
“Watson, but what does this have to do with payback?”
Web counters generate a huge number of digits, summaries and graphs, giving rise to the unbearable ease of being a marketer. So he read 10 tips on how to increase conversion in Yandex.Direct, looked at his charts and began to redistribute the budget until it runs out. Man, stop it! You did everything wrong.

Experience with clients and colleagues has shown that many people read metrics incorrectly and, as a result, make the wrong decisions. As a result, they simply master the budget, instead of making it work on a high ROMI. We came to talk about what is hidden behind the zeros and columns.
All examples describe real events, no abstract comparisons, and tips are easy to scale for small to medium budgets.
Go.
Bounce rate
Client: I saw in the web counters a bounce rate above 50% for a number of ads and requests, this is a lot. Let's refuse to work with these requests, we are losing money!
Web Analyst:A high bounce rate still does not mean anything. For all topics and sites, the concept of a "high failure rate" is unique, and you can only understand what your number is by trial and error. The main evaluation criteria are payback and conversions. Let's take a closer look at the ads in this campaign. We worked with semantics containing the names of competitors' brands, and the campaign had the highest bounce rates - this is logical and predictable. However, such requests have low rates and they cost a penny (literally a penny), and revenue is included in the top 10 most profitable campaigns. Conclusion: inquiries should not be refused, it may make sense to rewrite the text of the ad to drop out the target audience.

Brand traffic
Client: Why do you offer me to give contextual ads on request, where is the name of my brand? They will find the company anyway, it will be issued. I think this is a waste of money.
Web analytics: Indeed, it is still considered strange to buy contextual advertising for your branded search queries, because the benefits of this are not obvious. And you buy, while your brand traffic is not intercepted by competitors. Firstly, the situation from the first example is very revealing: if you do not take your place, this will be done by a competitor and will lead away “cheap” customers who actually looked for you.
Secondly, clients of companies from certain subjects may not understand where the search results are in the search results, and where are the ads (for example, embroiderers).
Thirdly, on the page of search results from 10 places of presence. It is important to fill them as much as possible. First of all, this is a free presence due to organic issuance (including through social networks and “Images”). You have secured 4–5 places. But why not take another place? Brand traffic, again, is inexpensive, but as efficient as possible. Total, a minimum of 50% of your issuance, and the chance that they will come to you is 1: 2 in your favor.

Geolocation
Client: We ordered and paid priority placement on Yandex Maps for three months - of course, there is traffic, but - the income is minimal, conversion is low, ROMI is negative. I see no reason to develop this tool.
Web Analyst:Evidence of geolocation performance does not need to be looked for in the meters - they are not there. People use maps for the same thing as 100 years ago - to get there, get there, find yourself. Indirect indicators of successful work with geolocation: we observed an increase in live traffic to online and offline stores (online traffic grew by 3.6 times in a month), tracked the effectiveness of the channel through call tracking using phone numbers delivered specifically for geolocation cards. Bottom line: we began to work more closely with traffic through geolocation, we encourage customers to leave feedback in cards, periodically change the description for current promotions.
The growth of traffic on the channel "Maps"

Email and Media
Client: I looked at web analytics and saw that email is among the top three tools with the lowest conversion. We are sending out something, but people are not very active in buying. Let's turn off this channel, concentrate on Direct.
Web Analyst: The tangible returns in numbers from email and display ads will always be dubious. These are tools from a fog which are almost not visible in counters. For example, the other day we turned off a newsletter dated January 16. Rechecked, indeed from January 16th. The buyer can read the letter, postpone it, change his mind, forget, remember, save up and buy. Buy from another device, from another mail, at another time, and the counter will not show that it was sent by the mailing list.
It’s even more complicated with the media: it is shown mainly for the sake of stimulating purchases. This is an image advertising that works for qualitative indicators (the formation and retention of interest, loyalty, etc.), and not for quantitative ones. But the exhaust in numbers can be tracked indirectly: with the launch of display advertising and email marketing, organic brand traffic should confidently grow. On our graph from Wordstat we can clearly see how brand traffic began to rise with the beginning of work in October to increase recognition.

In addition, you must consider the features of the buying process in each product category. For example, in the topic “large household appliances”, people place orders mainly on weekends, carefully studying offers, consulting with their families, planning a budget place and delivery time, etc. A long maturity period of customers does not allow to accurately track all orders: a letter with the offer may come to the mail of the father of the family, and the order will be made in 2 weeks from the grandmother's phone. However, knowledge of the characteristics of the audience still allows for more precise adjustment of advertising campaigns. For example, we began to announce stocks 10 days before they start, so that the client “ripened” a certain time before the purchase. Indeed, the “tail” of discoveries and clicks stretches stably for 2-3 weeks.
Schedule of heterogeneous daily activity

3 month tail

Viewing depth
Client: Why do people spend so little time on the site? Here is my friend’s average viewing depth of up to 20 minutes! Let's redo the site, it’s bad, people don’t like it.
Web analytics: Depth of view is another indicator that can be interpreted in different ways, as well as failure rate. Poor usability can be hidden behind a large viewing depth: people persistently look for a feedback form throughout the site and do not find it. And in a parallel universe, customers find the information they need in 30 seconds on the first page and, satisfied, leave. Meanwhile, the counter registers a small depth of views.
Unprofitable campaigns
Client: I calculated ROMI for contextual advertising, this tool is unprofitable, let's abandon it.
Web Analyst:It is worth digging deeper and separate the grain from the chaff. It’s possible (as it was with us) that you should turn off not the entire campaign, but individual unprofitable ads, or maybe just some separate request in this ad. It is highly likely that unprofitable ads pumped a lot of money out of the budget, sharply overturning the payback of the entire campaign minus. This problem is not solved by analytics, but by continuous testing of the entire pool of ads for each campaign. We started with a pool of 34 ads, during testing and rigorous selection we turned off 31 low-performing ads, leaving only 3 for the entire campaign, but with ROMI up to 755%. The budget was redistributed between them, and the campaign began to generate revenue.
PSWhat do we mean by all this? Analytics needs analytics. Namely, in competent analytics, which he will look with his eyes, he will think with his head and will refer to logic, and not to the figure in the summary. Because a figure in itself means little.
Last tip on how not to do:
If you are faced with a strange or one-way interpretation of the data from the counters, welcome to the comments - share your impressions and misconceptions.