Regardless of your campaign structure, it’s inevitable that some campaigns or keywords will have “low data confidence,” for which it’s difficult to make smart optimizations.
That’s where the AdLabs algorithm employs a Data Confidence Hierarchy
— think of it as “ladder”we climb until we find a level of aggregated, reliable data.
In order for us to consider a campaign as having high-confidence data for placement adjustments, each placement within that campaign must meet 2 criteria:
Have at least a certain number of clicks
Have at least one order
If each of the placements within a campaign have a sufficient number of clicks and orders, we can consider the Conversion Rates in that campaign to be “reliable” and usable for placement calculations.
To determine the number of clicks needed, we reference the average Clicks-to-Conversion (aCTC) of the parent entity.
In the event of insufficient data, we cross-reference a larger data set with more data confidence.
When campaigns are in an Optimization Group, that group becomes the first reference point and the placement performance for the entire group of campaigns is aggregated to create a higher number of clicks and orders for each placement.
In order for the Optimization Group’s placement data to be used, the aggregated placements for the entire Opt Group must also have high-data confidence (i.e., each placement type must have at least as many clicks as the Profile’s aCTC and at least one order).
If the Optimization Group placements are also found to be lacking data confidence, then the final reference point is the Profile level placements (i.e., the aggregated placement performance for the entire advertising account).
There are 2 primary conditions where a keyword has insufficient data for an informed bid:
High Spend, No Sales
Low Visibility
Similar to how we identify data confidence for campaign placements (detailed above), we climb the data confidence hierarchy until we have at least 1 order with a sufficient number of clicks.
When a keyword doesn’t have a sale, we do not have a “Revenue Per Click” from which to calculate what the bid should be.
What is Revenue Per Click (RPC) Bidding?
Our method for bid management and why it makes sense.
In these instances, we calculate the bid from the anticipated RPC.
We calculate this by climbing our data confidence hierarchy until we find a reliable datapoint, moving from Ad Group, to Campaign, to Optimization Group, until finally we reach the Profile level.
Once we find an entity with enough clicks and at least 1 order, we can use that entity’s benchmarks to estimate the keyword’s anticipated RPC as follows:
This allows us to appropriate calibrate the keyword’s bid to reduce non-converting spend so that if and when the keyword converts, its CPC will hit the Target ACOS.
For Low Visibility keywords (which have fewer clicks than the Benchmark aCTC), we need to increase bids to a place where we can get more data.
However, we don’t want to increase bids to infinity and beyond, so we need to have some safeties in place.
This is where Smart Bid Ceilings come in:
What are "Smart Bid Ceilings"?
How AdLabs protects you against wasted spend and overspend
The calculation of the Smart Bid Ceiling is dynamic, based on selected time frame and data confidence.
AdLabs will only increase bids up until the maximum affordable CPC, as determined by the “Benchmark RPC” according to the data confidence hierarchy named above.
We hope this article brings additional clarity to the AdLabs bidding algorithm and helps you use our system to grow sales profitably!
If you feel we are missing any information, please be sure to let us know :)
