- Mark as New
- Follow
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report
Best Practices for Portfolios: Portfolio Budget and Configuration Strategies
Advertisers with Media Optimizer Premium Only
Use the Appropriate Budget Strategy for the Portfolio
Consider the appropriate use case for each budget strategy:
Strategy | Use Case | Pros | Cons | Considerations |
Daily | Similar traffic trends occur throughout the week. | This strategy is easy to manage. | This strategy can't exploit day-of-the-week trends. | The budget may fluctuate slightly over a week, but the seven-day average will be close the daily budget. |
Monthly | The advertiser has a monthly budget and doesn't want to change it aggressively within a month. | This strategy is easy to manage. | Spend may be higher or lower at the end of the month, and this strategy can't exploit day-of-the-month trends. | The advertiser must be flexible regarding the daily spend. |
Day of Week | Inherent day-of-the-week (DoW) trends occur, or the advertiser has business constraints on specific days of the week (for example, when customer service is not available to process offline applications over the weekend) | This strategy can leverage DoW trends, and it provides more budget control than Daily. | It's difficult to determine the appropriate budget to exploit DoW trends. | The optimization capability must gather enough historical data to create DoW factors. |
Day of Month | Inherent day-of-the-month (DoM) trends occur. | This strategy can leverage DoM trends, and provides the most budget control of the budget-based strategies. | It's difficult to determine the appropriate budget to exploit DoM trends. | The optimization capability must gather enough historical data to create DoM factors. |
Return on Investment (ROI) | The advertiser has an ROI target, and the portfolio objective will include only revenue (or any other transaction properties will be weighted very low in comparison). If the advertiser is providing data via a feed, the feed delivery must be reliable. | The ROI target determines the budget; you can optionally set a maximum daily budget as a precaution. | This strategy provides less budget control. In addition, spend can be affected by any inaccuracy in the revenue model, and performance is affected when orphan transactions or other data issues occur. | Use when the traffic trends don't vary much. For advertisers sending data via a feed, use only when there isn't a large delay between the time the advertiser receives revenue and the time they send us the revenue and revenue per transaction data.
Best Practice: Use when the model accuracy is at least 85%, and set the minimum and maximum daily budgets to provide some budget control. In addition, set a low revenue model half-life to prevent volatility in performance. |
Cost per Transaction (CPT) | The advertiser has a CPT target (where CPT = Cost ÷ Weighted Revenue), and the portfolio objective will include only orders (or any other transaction properties will be weighted very low in comparison). If the advertiser is providing data via a feed, the feed delivery must be reliable. | The CPT target determines the budget; you can optionally set a maximum daily budget as a precaution. | The same as for ROI | The same as for ROI |
Marginal Cost per Transaction | The advertiser has a marginal CPT target (where marginal CPT is the change in cost to acquire one additional unit of weighted revenue). If the advertiser is providing data via a feed, the feed delivery must be reliable. | The marginal CPT target determines the budget; you can optionally set a maximum daily budget as a precaution. | The same as for ROI | The same as for ROI |
Optimize Your Budget Across Portfolios
When you have multiple portfolios, the Spend Recommendation Tool can help you to identify the optimal budget distribution across all of them so you can maximum revenue for the entire portfolio set.
Set a Maximum Bid for Each Portfolio
Most advertisers should configure a maximum bid for each portfolio, particularly when they receive smaller amounts of revenue per transaction. Sometimes the ad networks provide inaccurate traffic estimates for new keywords and ads, and you don't want to spend more for an ad than it will produce. For most portfolios, use the maximum expected cost per lead as a guide for setting the maximum bid. For Yandex-specific portfolios, initially set the maximum bid to two times the average cost per click and change it if necessary based on performance.
If you do set a maximum bid, make sure you monitor performance over time; it's likely that you will need to change the value eventually based on market changes.
Let Media Optimizer Adjust Campaign Budgets
Under most circumstances, you should let Media Optimizer determine the optimal bids by enabling the portfolio's spend management option to auto adjust campaign budget limits, with a multiple of two (2) or less to prevent volatility during the initial weeks of the launch. For portfolios with display campaigns, the optimal multiple value is 1.3. For portfolios with Facebook ad sets, we recommend 1 as the multiple value.
Don't manually bid on keywords unless it is absolutely necessarily. If it is necessary, consider temporarily setting manual bids by creating and posting campaign bulksheets, rather than setting manual bids permanently. Any bid changes resulting from the posted bulksheet data are applicable for just one day, and after that Media Optimizer will resume setting bids according to its own optimization strategy.
Avoid Large Budget Changes Unless They're Necessary
Typically, you shouldn't change the budget by more than 25%; this allows the optimization capability to learn gradually. However, when you need to make significant budget changes, the best practice is to let the optimization capability determine the optimal bids by enabling the portfolio's spend management option to auto adjust campaign budget limits.
In any case, you should expect to see volatility and possibly some overspending in the first few days after a large budget change, especially when the historical spend is much lower or higher than the new budget, as any keywords are bid to their previous positions. As soon as the optimization capability has new data on these positions and is able to create better forecasts, the bidding technology will adjust and the portfolio will stabilize.
Minimize Portfolio Limits
The optimization capability provides optimum performance when portfolios have no bid or position limits. When you set limits, ROI may decrease. Nevertheless, if you have smaller amounts of revenue per transaction, configure a maximum bid for each portfolio to ensure that you don't spend more for an ad than it will earn.
If you need to set portfolio limits, use historical data as your basis. For example, if you use a maximum bid for the portfolio, use the historical maximum CPC.
(Portfolios with Search Campaigns Only) Set the Learning Budget to at Least 10%
This will ensure that untested keywords are bid up until they receive impressions (for search campaigns, impressions on the first page of search results).
Set Appropriate Model Half-lives
Data half-life parameters control how quickly the optimization capability reacts to changes in market conditions by giving a higher priority to the most recent N number of days when evaluating bid units for bidding purposes. A smaller half life value makes the optimization capability react to changes more quickly. Each portfolio has a cost model half-life and a revenue model half-life.
For the cost model half-life, use seven days or more unless you expect volatile user behavior. For example, the Facebook and display marketplaces tend to be volatile, and you may need to use a half-life of just a few days for portfolios that include those types of campaigns.
For the revenue model half-life, use the greater of these two values: a) the average number of days between clicks and conversions or b) 30 days unless you expect volatile user behavior. For example, the Facebook and display marketplaces tend to be volatile, and you may need to use a half-life of just a few days for portfolios that include those types of campaigns if conversions occur quickly after clicks.
Note: You may lose information on some tail terms if the revenue model half-life is too short.
Note: Various factors can unexpectedly affect the performance of a portfolio, making the forecasting models inaccurate, and you may need to adjust the portfolio's half-life values accordingly.
Views
Replies
Total Likes