How to judge the effect of protection, before and after
Contents
Share
The documents that were read do not set a way to measure the effect of protection on ad spend before and after. They describe the decision on a visit and an allow if the script did not get an answer in time. A comparison of "we spent this much, then turned protection on and spent less" does not follow from them as a method. The absence of a method is not a measurement and not a figure for the effect on the cost of a lead.
Something else can still be compared: a channel's spend and that channel's revenue in end-to-end analytics, and the decision the check would have recorded in protection while the site is still in observation. Click fraud does not by itself provide that column.
What end-to-end analytics assembles
Metrica's help describes end-to-end analytics as a comparison of channel cost with channel revenue, and as the client's path from an ad click to an order or a purchase. To assemble it you mark ads with UTM, upload spend for the marked channels over a period, store in the CRM the ClientID obtained by the getClientID method, and pass clients and orders from the CRM into Metrica. On this page the help calls ClientID the identifier of a unique visitor, which Metrica assigns itself.
This is a way to see a channel's revenue, not a method of "ad spend before and after protection." The same data is read in the reports "CRM order sources," "Sources, expenses, and ROI," and "Visitors and clients." Goals used to pass order statuses can be added to other reports so the funnel stages are visible. The report names do not promise a "bots" column. ROI is not expanded into a formula on the page, so you cannot substitute your own formula of savings from protection into it.
Invalid clicks are not named as a separate column in these reports. End-to-end analytics does not subtract click fraud. Who clicks ads is not visible in the page's ROI. A refund for click fraud is not something the check calculates: it does not run the cabinet, does not count money, and does not build dashboards. It receives the site list ready-made. Reading the log is the cabinet's job.
Observation and protection
A site has an observation mode and a protection mode. In observation the check is computed in full, but the visitor is always allowed outward, a repeat-visit ban is not set, and the log receives both the actual allow and the decision that would have been made in protection. In protection the decision is applied to the visit.
The practical order is this. First turn on observation and read in the log which decision would have been applied, while real visitors still pass. Then turn on protection and compare the decisions that were applied with that log, not the account's spend "before" with spend "after": the documents do not set the second comparison. The ad platform still charges the money. The Binetro cabinet shows the log, not a recalculation of the platform's clicks.
The script gives the check result to the page. If a Yandex Metrica counter is set, the result can go there too. That mark is not Metrica's own definition of robots and does not rewrite the ad platform's report. A rise or fall of robots in a Metrica report after protection is turned on is not explained by this promise.
How not to substitute one comparison for another
Assemble end-to-end analytics if you need channel revenue: UTM, spend for the period, ClientID from
getClientID, clients and orders from the CRM.Read "Sources, expenses, and ROI" as a comparison of cost and revenue. There is no "bots" column in the report name, and the page does not give an ROI formula.
Do not subtract a click refund from this report. The check does not count money.
In observation, read the pair of records: the outward allow and the decision that would have been made in protection. A repeat-visit ban is not set in this mode.
In protection, compare the applied decisions with the observation log. Ad spend before and after is not measured by that pair.
Do not read a mark in the Metrica counter as Metrica's robots or as an edit of the Direct report.
An example, marked as a teaching example and not as a measurement. For a week a channel showed uploaded spend and revenue in "Sources, expenses, and ROI." In the same week the observation log recorded how many visits would have been blocked in protection mode. The two numbers answer different questions: the first is about the channel's money in Metrica, the second is about the check's decisions. The documents do not propose subtracting one from the other.