This is a guide that focuses on data-driven optimization of Yahoo Gemini campaigns.
Examples are for illustration purposes only. There is a period of “data discovery” that you will go through figuring out what your upper and lower limits on cost before making an adjustment.
Requirements to Successful Optimization
There are two cardinal items that must be setup correctly for you to succeed with Gemini ads.
- Campaign Tracking
- Conversion Tracking (Pixel, or "Dot Tag" Setup)
If you don’t have these setup correctly, you are dead in the water. I cannot stress this enough, do not implement any of this without proper tracking in place and verified. Trustworthy data is the cornerstone to success.
First a general rule of thumb is to apply optimizations at the campaign level; not the account level. There are certainly exceptions to this rule; however, this guide is based on campaign adjustments.
Second, we recommend allowing campaigns spend at least $10 before making any optimizations.
Third, this guide is based around cost per action “CPA” rules, such as:
- Click to landing page
- time on site
- watched % of video
- email submit
- lead generated
Literally anything as long as you are tracking the action.
These are guidelines to get you started in the right direction, adjust according to your style, workflow and budget.
The example rules below break down into two components: action and conditions. The action is the change you want to make such as changing a bid. The conditions is the criteria for the action to take place.
Now let us dive in.
Decrease Bid on Sites Above CPA
Star by lowering the bid on a site to see if that gets it to the CPA goal. In this rule we are lowering the bid 25% after 2x the goal CPA is spent. You might also use multiple tiers (two rules) of this kind, such as, a 20% decrease after CPA goal spend, then a second rule to decrease another 20% after 2x CPA goal spend.
Update Site bid - decrease by 25%
Publisher spend is greater than $100 (2x $50 CPA goal) & Site CPA is greater than $100 (2x $50 CPA goal)
Example rule in Brax
Increase Bid on Sites Below CPA
This rule is aimed at getting more volume from a site placement that performs well by bidding slightly higher. The idea being you will be in a higher position on that publication which can subsequently increase CTR. You want to exercise caution and increase the bid slowly so a profitable placement does not turn upside down.
Increase Site bid 10%
Site spend is greater than $100 (2x $50 CPA goal)
& Site CPA is less than $40 (20% below $50 CPA goal)
Example rule in Brax
Frequently Asked Questions
Use these rules as a starting point to streamline your campaign optimization efforts leveraging performance data.
Many of the questions don’t have a straightforward answer without insight into your business metrics so consider it general guidance. Budget is a big factor to determine how long to look back (last 7 days vs 14 days) and minimum spend before optimizing.
How many days back should I look?
I recommend at least 7 days at a minimum; with up to 30 days if your conversion window is longer and budget allows. Small budgets need longer time frames usually.
The first time you run rules, try looking back 60 and 90 days for sites that have never performed. Repeat this once a month to once per quarter.
What should my minimum spend for campaigns/ads/publishers before I pause/block/adjust?
I recommend a campaign spend of $10, ads & publisher spend of $5 at a minimum. You should have a CPA goal (or similar replacement) and base it on 3x the cpa goal. For example, a $10 cpa goal uses a minimum of $30 for campaigns/ads/publishers before adjusting. The common exceptions are obvious fraud or poor performance.
How often should I run a rule?
This really depends on how much you are spending; because a small budget won’t have enough new data each day. On the other hand, spending thousands per day may warrant running every day.
A good rule of thumb is once a day to catch new placements that need blocked and once a week for bid adjustment rules.
Usually the best approach to Gemini ad optimization is to start with a manual process to work out the kinks. Then progress to a more automated process.
If you have not done the optimization manually yet, automating it will likely make a mess.
Make your operation efficient with the rules above. Once you have defined, tested and trust rules then you are ready to automate.