5 Steps to Implement a Successful Bidding Strategy in Paid Search
When you set up bidding in paid search, it’s important you put a strategy in place prior to implementation. The most relevant questions that need to be addressed when developing this strategy are as follows: How many bidding strategies do I need? What cookie window should I use? Should I use a revenue-attribution model and […]
When you set up bidding in paid search, it’s important you put a strategy in place prior to implementation. The most relevant questions that need to be addressed when developing this strategy are as follows:
- How many bidding strategies do I need?
- What cookie window should I use?
- Should I use a revenue-attribution model and move away from a last-click model?
- How can I get the most of long-tail/low-volume keywords?
In this post, I’ll try to answer these questions in a fairly general way while keeping those mobile and location bid modifiers aside since they are layered on top of the base bids.
1. Ensure Your Account Is Ready For Bid Automation From An Editorial Standpoint
While this is not directly related to the actual bidding strategy, you definitely want to make sure your Quality Score is optimum before you attempt to scale up your account. Indeed, since Ad Rank= Max Bid*Quality Score, you want to first maximize Quality Score in order to minimize the average cost-per-click at more aggressive positions. So, you want to address the account structure and test a bunch of ad copy first.
Also, you want to make sure your ad group structure has been stable for at least a couple of weeks so you can collect enough historical data at the keyword, ad group and campaign levels.
2. Break Down Your Paid Search Program Into Multiple Keyword Buckets
Unless your paid search program is very basic, you’re most likely bidding on different kinds of keywords with different performance levels — at the very least, branded vs. generic keywords. You definitely do not want to optimize your branded keywords just like your generic keywords since the average positions and CPCs are typically very different.
More specifically, those buckets of keywords can be defined based of the current efficiency level and the number of clicks needed to make statistically significant bid decisions, which is correlated to the average conversion rate.
For instance, we could have something as follows:
- Bucket #1/Brand keywords: $1 CPA, 10 clicks to get a conversion
- Bucket #2/Generic keywords: $10 CPA, 25 clicks to get a conversion
- Bucket #3/Competitors’ keywords: $20 CPA, 30 clicks to get a conversion
Once those buckets are defined in your account, you can further delve into consumer behavior, lifetime value (LTV), and actual bid rules for each individual bucket.
3. Analyze Consumer Behavior Trends For Each Bucket
You want to determine the average time lag from impression/click to conversion and the average number of clicks per sale for each individual bucket. (See this article for more detailed information on multi-channel attribution models.)
More specifically, I’ve found it makes sense to determine how many days and clicks it takes to get 90% of all conversions. For example, you could very well see the following:
- Bucket #1/Brand keywords: 90% of all conversions occur within 60 days and 1.5 clicks. You want to use a 60-day cookie window and a custom revenue-attribution model reflecting multiple touch points.
- Bucket #2/Generic keywords: 90% of all conversions occur within 90 days and 2.2 clicks. A 90-day cookie window and a custom revenue-attribution model seem relevant.
- Bucket #3/Competitors’ keywords: 90% of all conversions occur within 90 days and 2.5 clicks. You want to use a 90-day cookie window and a custom revenue-attribution model.
In AdWords, you cannot really do anything about those trends since the default cookie window is set to 30 days, and the only revenue-attribution model available is a last-click model.
However, when using more sophisticated technologies, you’ll be able to set custom cookie windows and revenue-attribution models for each individual bucket of keywords, which is by far a more sophisticated way to optimize assisting/upper-funnel keywords.
4. Determine The Lifetime Value For Each Bucket
Obviously, branded keywords tend to outperform generic and competitors’ terms at first sight; however, the LTV can be significantly greater for generic terms. Depending on the reporting suite you are currently using, you might be able to pull user-centered reports as opposed to keyword-centered reports. Instead of analyzing the number of keywords/clicks involved in a conversion, you want to determine the number of conversions and associated amount of revenue by unique user.
For instance, you can expect these types of trends in your account:
- Bucket #1/Brand keywords: $80 revenue by unique user
- Bucket #2/Generic keywords: $120 revenue by unique user
- Bucket #3/Competitors’ keywords: $100 revenue by unique user
In conjunction with the average order value (AOV), those LTV numbers help adjust your CPA/return on advertising spend (ROAS) targets from a bidding strategy standpoint. If you are able to determine that LTV is 20% greater on generic keywords, you should definitely tolerate a higher CPA or lower ROAS in order to maximize revenue volume in the long run.
Also, high LTVs are usually correlated with high percentages of new customers vs. repeat customers. So if you are able to distinguish new from repeat customers, you definitely want this factored in.
5. Build Out Your Bid Rules Logic For Each Bucket
Now that we have a better understanding of each bucket of keywords in terms of current efficiency and consumer behavior, we can actually build a custom bid rules logic:
In this example, you would want to build at least one bid rules logic for each bucket, where the bid rules are aiming at LTV-adjusted CPA targets and using the click thresholds and cookie windows determined earlier. Also, if the technology you are using allows it, you can address those long-tail keywords by clustering the data from the keyword level up to the ad group, campaign, or even account level.
For instance, for those high-volume keywords sitting in bucket #2:
- If last 7 days keyword-level clicks>25 and conversions=0 then bid down by $0.10
- If last 7 days keyword-level clicks>25 and CPA<$14.00 then bid up by $0.05
- If last 7 days keyword-level clicks>25 and CPA>$16.00 then bid down by $0.05
For those long-tail keywords, this 25-click threshold won’t be reached at the keyword level so you want to use the ad group, campaign or even account level instead. You can also use longer time ranges in order to collect more data:
- If last 90 days ad group-level clicks>25 and conversions=0 then bid down by $0.10
- If last 90 days ad group-level clicks>25 and CPA<$14.00 then bid up by $0.05
- If last 90 days ad group-level clicks>25 and CPA>$16.00 then bid down by $0.05
Generally speaking, you want your bids to reflect recent keyword-level performance while looking at longer time ranges and less granular statistics levels to address those long-tail keywords — so, you’re likely to end up with the following cycles of bid decisions:
Those are just basic examples — in reality, you’ll most likely end up with more buckets and/or more cycles of bid decisions. However, this is a reliable logic in general.
A successful bidding strategy is not just the result of a performing bidding algorithm. Instead, it has to do with understanding consumer behavior, adjusting your goals accordingly, and then actually automating the bids within a well-organized account. Also, you should obviously learn how to easily tweak your bidding strategy to get the most of seasonal trends and special promotions.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.