Automate it
Last year, I have been actively engaged in the automation of advertising campaigns in contextual systems and caught myself thinking that I began to pay attention to the automation and robotization of processes wherever I came. Maybe the profdeformation?
It's amazing how much has been automated by people lately, some things have entered our lives so tightly that their absence already seems unnatural. Automatic money transfers alone save a lot of time, not to mention automated apartment cleaning, cooking, etc.
The described examples have one common flaw, each of them is automated with standard operations that are practically not amenable to personalization and other changes.
')
Such universalization is unacceptable in advertising, because “every client is unique” and “an individual approach is required for every client,” and this significantly complicates the creation of automated tools for a wide audience.
In Garpun, we tried to solve this problem and here's how.
To begin with, Garpun is a service that allows you to create and update large advertising campaigns in Yandex.Direct and Google Adwords. By large, I mean campaigns with more than 10,000 advertisements, of which at least 20% is updated daily. Of course, first of all, we are talking about advertising campaigns of online stores, portals and services for booking tickets, but this is not all users of the service.
The first problem we faced was the lack of a standardized format for exchanging data in various industries. Online stores, with a standard file for Yandex.Market, have greatly facilitated the work at the start, but since we focused on a wider audience, we had to teach the tool to work with arbitrary XML.
At the interface level, it looks like this:
The user selects arbitrary objects from one or more XML and links them together.
At the exit, it turns out a flat table, which contains redundant information for creating an advertising campaign. Not very economical in terms of resources, but allows users to use arbitrary XML with data. (The screenshots show the processing of XML site for the sale of films.)
Working with arbitrary XML leaves a mark on the quality of the transmitted data. For example, in one variable may contain the name of the product, specifications, product, color, listing of objects of the same type and much more.
To get around this problem, we have implemented filters and functions for working with source data at the interface level.
Now, a sophisticated user can independently remove extra information, apply filters, or group data by an arbitrary key, for example [Actor + Genre]
After the data is grouped and cleaned up, it's time to start creating ads and keywords. With the creation of the announcement, everything is quite simple, but the process of forming the semantic core of the process is sacred. How many people, so many opinions. We tried to argue, proving our approach to key creation, and then just made a constructor.
Using the key designer, an inquiring mind can create an unlimited number of query generation rules based on XML data, embedded reference books, temporary variables, and 14 functions. But, as they say, great opportunities - a big responsibility. Therefore, all key requests, before pouring into contextual systems, pass an obligatory check for the presence of impressions. If according to the information from Yandex or Google there was not a single key show for 30 days, it will not be loaded into the systems. We cannot protect ourselves from trash generation, but we can protect user accounts.
That's all.
In the next article, I will explain how we learned how to create structured XML with data based on a site map. As it turned out, not all site owners or advertising managers can make the simplest XML with information from the catalog. With this and struggled.
It's amazing how much has been automated by people lately, some things have entered our lives so tightly that their absence already seems unnatural. Automatic money transfers alone save a lot of time, not to mention automated apartment cleaning, cooking, etc.
The described examples have one common flaw, each of them is automated with standard operations that are practically not amenable to personalization and other changes.
')
Such universalization is unacceptable in advertising, because “every client is unique” and “an individual approach is required for every client,” and this significantly complicates the creation of automated tools for a wide audience.
In Garpun, we tried to solve this problem and here's how.
To begin with, Garpun is a service that allows you to create and update large advertising campaigns in Yandex.Direct and Google Adwords. By large, I mean campaigns with more than 10,000 advertisements, of which at least 20% is updated daily. Of course, first of all, we are talking about advertising campaigns of online stores, portals and services for booking tickets, but this is not all users of the service.
Our information is all
The first problem we faced was the lack of a standardized format for exchanging data in various industries. Online stores, with a standard file for Yandex.Market, have greatly facilitated the work at the start, but since we focused on a wider audience, we had to teach the tool to work with arbitrary XML.
At the interface level, it looks like this:
The user selects arbitrary objects from one or more XML and links them together.
At the exit, it turns out a flat table, which contains redundant information for creating an advertising campaign. Not very economical in terms of resources, but allows users to use arbitrary XML with data. (The screenshots show the processing of XML site for the sale of films.)
Tresh and waste
Working with arbitrary XML leaves a mark on the quality of the transmitted data. For example, in one variable may contain the name of the product, specifications, product, color, listing of objects of the same type and much more.
To get around this problem, we have implemented filters and functions for working with source data at the interface level.
Now, a sophisticated user can independently remove extra information, apply filters, or group data by an arbitrary key, for example [Actor + Genre]
The key to all doors
After the data is grouped and cleaned up, it's time to start creating ads and keywords. With the creation of the announcement, everything is quite simple, but the process of forming the semantic core of the process is sacred. How many people, so many opinions. We tried to argue, proving our approach to key creation, and then just made a constructor.
Using the key designer, an inquiring mind can create an unlimited number of query generation rules based on XML data, embedded reference books, temporary variables, and 14 functions. But, as they say, great opportunities - a big responsibility. Therefore, all key requests, before pouring into contextual systems, pass an obligatory check for the presence of impressions. If according to the information from Yandex or Google there was not a single key show for 30 days, it will not be loaded into the systems. We cannot protect ourselves from trash generation, but we can protect user accounts.
That's all.
In the next article, I will explain how we learned how to create structured XML with data based on a site map. As it turned out, not all site owners or advertising managers can make the simplest XML with information from the catalog. With this and struggled.
Source: https://habr.com/ru/post/231069/
All Articles