If youâ€™ve been running ad campaigns online youâ€™ve likely heard of Portfolio Bidding, and maybe youâ€™re thinking of implementing it across your program. What you may not yet know are some of the intrinsics of Portfolio Bidding, and how this bidding strategy differs from the traditional Keyword-based approach.
This article explores the strategies and motivation behind adopting a Portfolio Bidding approach, why it can be a good idea to set up on your account, and some potential issues that need to be considered when setting up in this way.
What is Portfolio Bidding, and how is it different to Keyword-Level Bidding?
Portfolio bidding is a change in the way bid strategies approach hitting the target goals you set. In a traditional keyword bidding strategy, each individual keyword is bid on in such a way to ensure that each keyword hits the target goal. In essence, each keyword is isolated from the performance of other keywords in the same campaigns or account.
For some automated bidding platforms, this can be a limitation. We all know different keywords have different performance, and as such an automated bidding platform may act inefficiently when presented with a goal that is much too aggressive, or lenient for a given keyword. For example, if we are bidding to a Target CPA, a goal that is set too low for that keyword will likely result in disappointingly low volume.
Instead of manually changing targets at a keyword level, or restructuring a campaign in such a way that similarly performing keywords are grouped together, we can instead use a Portfolio Bidding strategy. With Portfolio bidding, the keywords execute in a manner that ensures aggregate performance across all keywords in a portfolio (typically a Campaign or Bid Policy) hits the target goal.
Why should we use Portfolio Bidding?
Put simply, Portfolio Bidding gives the bidding algorithm of your choice more options to optimize. To explain this in more detail, weâ€™ll introduce the concept of the volume/efficiency curve. For each keyword in your account, itâ€™s possible to draw a curve representing the relationship between cost, and the expected volume of your target metric. This curve will â€˜taper offâ€™ as spend increases, resulting in a lower marginal return for each additional dollar spent on that keyword. In economics, this is called the Law of Diminishing Returns. A given keyword will have a unique curve, which could be subtly or dramatically different when compared to other keywords in the portfolio. Below is a hypothetical example of a keywordâ€™s volume/efficiency curve.
We can see that an increase in
A given portfolio has a single campaign containing two keywords, and the target goal is to drive volume at a Target CPA of $10. The keywordâ€™s volume/efficiency curves are as follows:
As we can see above, Keyword 1 is more efficient at lower costs, but tapers off early. Keyword 2 does not have the â€˜low hanging fruitâ€™ available at a lower spend, but instead has a lower drop off at higher spend levels.
A keyword level bid policy will likely take this approach to
When optimizing to the target we can only bid in such a way that finds the highest spend level achieving the target CPA for each keyword. This is inefficient, as the marginal spend on Keyword 1 would have been better used on Keyword 2. As a keyword level bid strategy does not have insight into the performance of other keywords, this is the only option available. This example shows a spend of ~$650 total would expect to see around 65 conversions.
On the other hand, the portfolio bid policy will work in this fashion:
With portfolio bidding we can find the optimal spend distribution across all keywords, to hit our target goal with the highest volume possible. This example shows a spend of ~$600 total would expect to see around 67 conversions. We can see in this example that we can end up with a higher volume of conversions at a lower cost, resulting in improved account performance.
When should we be careful about using Portfolio Bidding?
Logically it follows that where possible, our bidding strategies should consolidate on the smallest number of Portfolios. However, we should keep in mind that Portfolio bidding is not a silver bullet for improving the performance of campaigns. Some of the situations in which discretion is advised when potentially switching over to portfolio bidding are:
- Different target goals: As you would split up campaigns under keyword-based strategies based on the target goals, if the target goal for a given list of campaigns is drastically different, or you wish to optimize to different metrics, it is best to keep them separate. However, you can still apply a portfolio based approach to each of these groups.
- Volume Requirements for certain campaigns: This is a subtle, and potentially overlooked, point when setting up this type of strategy. As a portfolio-based strategy is attempting to maximize the volume of conversions across the whole portfolio at the desired efficiency target, it is possible that it will lower the volume of campaigns that are important for other business reasons, even if they could hit a high volume at the target goal. The recommendation here would be to split those campaigns out into other bid strategies.
- A need for Manual Overrides: As a Portfolio-based automated bidding platform uses characteristics of keywords across the whole portfolio, it may behave inefficiently if there are regular manual overrides on certain keywords/campaigns. If this is something that is liable to occur for multiple campaigns (i.e. certain campaigns regularly need to be boosted due to an anticipated spike in traffic), it would be best to move those campaigns into their own policy.
For most accounts using automated bidding platforms, switching over to a Portfolio-Based Bidding Strategy makes sense. There is a good chance you will see improvements in volume at your target goal across campaigns, as a Portfolio-Based Strategy can greater leverage the characteristics of all the keywords in your campaign, and not just optimize on a keyword by keyword basis. QuanticMindâ€™s bidding platform has the ability to capitalize on the benefits of portfolio bidding at a Bid Policy level (by combining keywords with similar goals across different ad groups and campaigns), as well as further optimizations by sharing Natural Language Processing (NLP) data between semantically similar keywords.