Avatar

Level 10

Best Practices for Portfolios: Improve Cost and Revenue Model Coverage

Advertisers with Media Optimizer Premium Only

The optimization capability uses historical data to create cost and revenue forecasting. The reliability of forecasting is dependent on the breadth and quality of the historical data available for optimization. You can identify how many of the portfolio's bid units have cost model coverage and revenue model coverage on the Modeling tab of the portfolio settings.

The following actions will help improve model coverage for a portfolio:

  • Let the optimization capability automatically adjust spending limits for the campaigns in the portfolio to maintain spending for the entire portfolio, based on expected daily spend, to ensure that no individual campaign's budget is too low. To do this, go to the Spend Management tab of the portfolios settings page and enable the option to auto adjust campaign budget limits; for instructions, see "Edit Portfolio Settings."

  • Set a reasonable learning budget for the portfolio, which helps expose bid units that have not received enough clicks to be thoroughly evaluated and for which the optimization capability has therefore not built cost models. In addition, spend a higher percentage of the learning budget on bid units that have never received any impressions. To do this, go to the Modeling tab of the portfolios settings page and select the options to enable learning and to bid up bid units with zero impressions; for instructions, see "Edit Portfolio Settings." Increase the bids slowly, and monitor the model coverage for the portfolio after you do so. The optimization capability will be able to create cost models as soon as the bid units get enough impressions.

  • If you don't want to use the learning budget to help existing bid units without cost models to receive exposure, then manually adjust bids for individual bid units that have never received any impressions or are in very low positions. To identify these bid units, run a keyword-level report. You can temporarily set manual bids for any campaign in the portfolio by creating and posting campaign bulksheets; any bid changes resulting from the posted data are applicable for one day, and after that the optimization capability will resume setting bids according to its own optimization strategy. For instructions on using bulksheets, see "About Managing Search Campaign Data via Bulksheets." Bid up the bid units in increments until they get some impressions. The optimization capability will be able to create cost models as soon as the bid units get enough impressions.

  • For new keywords and ads, set initial bids to at least 25% of the average CPC. Adding new keywords with reasonable initial bids helps them to get exposure as soon as they are added so the optimization capability can test them effectively. The optimization capability will bid down any low- or non-performing keywords and ads and will bid up productive ones.

    • For search campaigns, you should expose most of the keywords that map to relevant broad match searches by setting reasonably high bids for them (appropriate to the marketplace) until they have received a non-trivial number of impressions and clicks at Positions 1 and 2 for at least a couple of days in the past 30 days without hitting the campaign budget caps.

    • For display and social campaigns, whose ads have shorter life spans, you should expose all of the ads by setting reasonably high bids for them (appropriate to the marketplace) until they have received a non-trivial number of impressions for at least a couple of days without hitting the campaign budget caps.

  • If you think that low revenue model coverage may be due to inherent low revenue potential, consider pruning ineffective bid units, adding new bid units, and (for search campaigns) adding negative keywords.