Testing a period in context targeting

Search period is the time during which search queries are counted, that is, how many days ago the search query was taken into account.

Please note that a short period of search increases efficiency, but reduces the coverage.

Usually, the search period corresponds to the sales cycle of the advertised product, but it is better to conduct an experiment to determine it more accurately.

How to select a search period

To select the most effective search period, create two campaigns with the same settings and a different search period.

If the list has more than 10 million unique users and the campaign's geography settings are wide, then a comparison is appropriate:
  • 3 days / 3-7 days,
  • 7 days / 7-14 days,
  • 14 days / 14-30 days.

If the list includes 1-10 million unique users or the campaign's geography settings are narrowed (for example, only the Moscow region and the Leningrad region are selected), a comparison is appropriate:
  • 7 days / 7-14 days,
  • 14 days / 14-30 days.

If the coverage of less than 1 million unique users or geo in the campaign is narrow (for example, a regional city or local advertising), a split of 14 days / 14-30 days is appropriate.

To make testing as efficient as possible, make sure that you have set a fast allocation and that the rate is high enough. Then expand the search period.

How to compare campaigns

    Let's compare two campaigns:
    • for users who searched for the product no later than two weeks ago (0-14 days),
    • for users who searched for the product 14-30 days ago.
    In order for the split test to be correct, the settings in the test campaigns must be the same, and the campaigns must work simultaneously, not sequentially. The difference between campaigns should only be in the segment.
    Step 1. Create a data source "Context targeting": load a list of keys and select a period (more about how to create a source). In our case, the period is 14d. Verify that the checkbox "Automatically create a segment" is checked.

    Then create a second similar source, but with a different period, in our example — 30d.
    Creating a list of search phrases
    So we get two lists with the same keys and negative phrases, but different search periods.
    Two similar lists with different search periods
    Step 2. Now we need to get a list that takes into account the period not from 0, but from 14 to 30 days. To do this, create a segment where we subtract the list of 14d from the list of 30d.

    1. Open the "Segments list" section and select the automatically created 30d segment — [auto] Wardrobe 30d.
    Created segment in the segment list
    Editing a segment
    2. In the list, select The "Crossing conditions" setting and click "Add audience segments". In the window that opens, select the source "Contextual targeting" and find the list for 14 days (30D wardrobe). Note "Not in the list".
    Exclude the list for 14 days
    The segment page will look like this:
    Setting up a segment from 14 to 30 days
    Save the segment.
    Step 3. Start two campaigns: with a 14-day segment (in our example [auto] Wardrobe 14d) and with a 14-30-day segment ((in our example [auto] Wardrobe 30d -14d).

    To do this, when creating a campaign, open the "Detailed audience setup" and select segment targeting.
    Creating campaign with a search period of 14-30 days
    Campaigns will be displayed in the list of campaigns.
    Running campaigns
    To collect sufficient statistics for conclusions, it is recommended to run the test for at least 3 days and each campaign should score at least 10 conversions (more is better).

    The difference in efficiency between periods is recognized if their conversion rate differs by more than 20%. If the difference is smaller, the effectiveness of the periods is recognized as similar and they can be used together in the same campaign.
    Was this article helpful?