How to evaluate the results of performance campaigns

In this article, you'll find recommendations for analyzing statistics that are appropriate for the following types of campaigns:

  • “Text & Image ads”
  • “Smart banners”
  • “Search banners”
  • “Dynamic ads”
  1. Which metrics to track
  2. How to create a report
  3. How to make decisions based on statistics

You can view ad campaign reports after your ads start displaying. Study the statistics regularly to learn how effectively your ads attract customers and to find weak points in your ad campaign and fix them.

Which metrics to track

To find out whether performance advertising is effective, you need to know more than just how much traffic it has brought to the site. The main point is whether you are gaining customers and revenue. Instead of looking at the CTR, the CPC, or the ad display position, focus on metrics related to the conversion of clicks into real orders.

In order for conversion metrics to be available in a report, you need to connect Yandex.Metrica.

Settings to configure Metrics available in the report
Install a Yandex.Metrica tag on every page of your site and link the tag to the ad campaign Engaged sessions, cost per engaged session, engaged session rate (%)
+ Set up Yandex.Metrica goals (see recommendations below) + Conversion rate (%), Goal cost, Conversions
+ Submit the conversion value (see recommendations below) + Revenue, Profit, ROI
Engaged sessions, cost per engaged session, engaged session rate (%)

Engaged sessions metrics show how interested site users are in your products and services. This is calculated using machine learning algorithms. These metrics help you evaluate the effectiveness of advertising if you have connected Yandex.Metrica but have not set up goals.

Learn more about engaged sessions

Conversion rate (%), Goal cost, Conversions

What they show. The Conversions metric represents the number of converted sessions that advertising brought to the site, Conversion rate (%) shows these sessions as a percentage of the total number of sessions, and Goal cost shows the average expenditure to obtain one converted session.

How to configure the settings. To get conversion data in reports, set up goals for your advertising campaign in Yandex.Metrica — that is, the key actions that you want users to perform.

We recommend creating several goals that correspond to different levels of user engagement, for example:

  • For online stores — “Viewed 2 or more products”, “Added an item to the cart”, “Placed an order”.
  • For a site offering services — “Viewed 3 or more pages”, “Visited the services catalog”, “Clicked ‘Call Me’, “Submitted an online request”.
Learn more about setting up goals
Revenue, Profit, ROI

What they show. Conversion value is your estimate of the revenue you receive when the goal is reached. Once you determine the conversion value, you can find out how much revenue and profit you receive from advertising and its return on investment, i.e., the payback from each dollar or euro:

How to configure the settings. In Yandex.Direct, you can pass the conversion value in two ways:

  • Specify the priority goals and their conversion values in the campaign settings. Learn more about priority goals

    For example, if 100 requests on the website brings you 5,000 euros, the conversion value for the goal “Submitted an online request” would be 50.

  • If your main advertising objective is for purchases to be made on the site, enable e-commerce data collection in Yandex.Metrica to pass the amount of each order.
Note. If you know the approximate value of each conversion but do not submit them to Yandex.Direct, you can estimate the revenue, profit, and ROI independently using data on the number of conversions and their cost in the report.

How to create a report

Follow the View statistics link on the campaign page. To view statistics for all campaigns, go to the My campaigns page and click on Statistics for all campaigns.

In the Report wizard, create a report with the following parameters:

Dimension
Choose “For selected period”.
Goal data
Specify the goal you want to find out conversions and profit for.
Attribution

Select “First click” to attribute conversions to the ads and keywords that initially brought a user to the site. This attribution model is appropriate for sites with deferred conversion where users take a long time to make purchasing decisions.

Select “Last non-direct click”to attribute conversions to those ads and keywords that brought the user to the website before completing the goal (that is, there were no transitions from search and from other sites between clicking on the ad and conversion).

Learn more about attribution models

Columns

Select metrics:

  • Clicks, Average bid per click, Total expenditure
  • Engaged sessions, Cost per engaged session, Engaged session rate (%) — if a Yandex.Metrica tag is installed.
  • Conversion rate (%), Goal cost, Conversions — if, in addition, goals are set.
  • ROI, Revenue, Profit — if, in addition, conversion values are transmitted when goals are completed.

In the report, you will see:

  • The number of conversions and how much profit and revenue was produced by advertising.
  • How much you spent for the period overall and on average per conversion.

How to make decisions based on statistics

To investigate which factors affect ad conversion, add different cross sections to the report and see how the data for these cross sections differ.

For example, by adding the cross section Formats, you may see that ads with video extensions bring more conversions and the price of these conversions is lower. In this case, you can get additional conversions by setting bid adjustments for displaying ads with video extensions.

In this example, the CPA for ads with video extensions is 16.5 euros. If you are satisfied with a CPA of 50 euros, you can increase the rate for video extensions by 50 / 16.5 ≈ 3 times, that is, add 200%.

Similarly, you can compare conversion costs and ROI by city, gender and age of users, and so on.

How to compare campaigns

To compare campaigns that have different settings, conduct anexperiment. For example, you can compare a campaign with a manual bid management strategy and a campaign with automatic bid management. You can also compare a campaign that uses autotargeting and a campaign without autotargeting. An experiment allows you to exclude the impact of other factors: randomly divide your audience into segments and assign each of your campaigns to a different segment. You can use the Calculator to analyze the experiment results.

Learn more about experimenting with campaigns

Compare which of the campaigns is the most profitable.

How to compare keywords

Add the cross section Impression criteria to the report — it displays the keyword text. Sort the report rows by the Clicks column value. Study the statistics for the most popular keywords — for example, the 100 keywords with the most clicks. Note how many conversions and how much revenue each keyword brought, as well as the cost of these conversions.

Exclude non-relevant queries

If the ROI or the cost of conversion for a keyword is below the optimal value, check the keyword using the Search queries report. In the search queries report, specify the same parameters as for analyzing the overall effectiveness of the campaign in the Report Wizard. Add goal and keywords filters.

You will see real search queries entered by a user that match your keyword. Sort the report rows by the Clicks column value. This way you can identify non-relevant queries that don't lead to conversions. To exclude impressions for these queries, add this query to negative keywords for the campaign or ad group (you can do this directly on the report page), or add negative keywords to the keyword.

Adjust bids

If you manage your bids manually, increase the bid for high-yield keywords and decrease the bid for low-yield keywords.

To calculate the most efficient bids according to the statistics, use the Average bid per click indicator: it shows the average bid for the keyword with all the factors and adjustments. You can also calculate the bid yourself.

The optimal bid in search is

.

As follows from the VCG auction rules, the difference between conversion revenue and advertising expenses is at the maximum at this bid rate.

The Yandex Advertising Network and external networks have a GSP auction, and the optimal bid is

, where

  • is a coefficient from 0 to 1 depending on the type of business. The median value of α is 0.8.
How to compare ads

Add an Ad ID/Banner № cross section to the report and compare the stats on the ads.

As a rule, ads that advertise the same product and have a similar message meaning will have a similar conversion rate. Among these ads, it's sufficient to choose the one that is most clickable. This selection is done automatically if you place ads in one group: as soon as enough statistics have accumulated, the system will automatically identify the most clickable ad for each keyword and display it more often. Learn more about experimenting with ads

How to compare redirect pages

To compare two pages that a user lands on when they click on an ad, conduct an experiment. Create two ad groups, each in its own campaign, with identical content and settings, with the exception of links in ads. The redirect pages must have the same tag installed. Learn more about experimenting with campaigns

Compare the campaigns' conversion rate to find out which of the redirect pages is better in getting users to a purchase.

How to compare advertising platforms

It does not make sense to analyze statistics by platform and manually disable impressions on ineffective platforms. Instead, we recommend using automatic strategies: Optimize conversions or Optimize ROI. If you apply other strategies, specify your priority goals in the campaign settings. The system automatically optimizes the bid in accordance with the conversion forecast: the less conversion traffic coming from an ad platform, the cheaper the clicks will be.

Account for statistical significance and margin of error

In order to draw sound conclusions from statistics, a sufficient amount of data is needed. For example, if a report line contains one or two conversions, their price in the report may differ significantly from the true value, which is the one you would see if there were very many conversions.

When comparing statistics for two cross sections, consider the statistical error – the possible deviation of the data obtained from the true value. If the data differ by less than the margin of error, it cannot be concluded that in one cross section the results were better than in the other.

Reducing the margin of error by half requires four times as many conversions. To increase the number of conversions in the report, enter a longer period or choose a different goal.

How to calculate the margin of error

With a 68% probability, the true conversion values and conversion costs are in the range

and

Here the coefficient is , where is the average conversion cost and is the standard deviation. In most cases, it can be assumed that .

The table shows the relative error depending on the number of conversions.

Conversions Deviation from true value
Conversion rate (%) Conversion cost
25 ± 20% ± 40%
100 ± 10% ± 20%
400 ± 5% ± 10%
2500 ± 2% ± 4%
10,000 ± 1% ± 2%