DoubleClick Search Bid Strategies – How and When to Test
Published: April 10, 2018
Author: Alexandra Shanahan
When it comes to 3rd-party bidding platforms, DoubleClick Search is quickly rising above its competitors thanks to its powerful bid strategy algorithms and plug-and-play compatibility with the rest of the Google/DoubleClick ecosystem. More and more of our clients are testing the platform or making the switch over from other tools.
How do you know if DS is right for your marketing campaigns? Furthermore, how do you begin testing, and what are the best practices for testing? We’ll aim to answer those questions in this blog post.
When Should I Use DS Bid Strategies?
There are three main questions you should ask yourself before starting to consider DS:
- Are you a large advertiser? DS makes the most sense for enterprise clients across multiple engine accounts. In addition, if you spend a significant amount on other search engines (Bing,Yahoo, Baidu, etc.) you can leverage Google’s technology for those accounts.
- Have you tested Google AdWords Smart Bidding? If you have tested tCPA (Target CPA) and/or Conversion Optimizer and had success, you might not need DS. DS is designed to handle large accounts with offline conversions and latency. If you have trouble with pre-packaged bid strategies like tCPA/CO, DS’s custom features could help.
- Are you willing to pay for the algorithm and other campaign management tools? Like other 3rd-party platforms, DS does come with a fee, typically a percentage of media spend.
If the answer to all these questions is yes, it’s time to consider onboarding DS.
Bid strategies are the primary reason to consider DS, but there are also other features that can be handy for enterprise accounts.
- Executive Reports – create custom visualizations to present and analyze your campaign data
- Social Reporting – pull in performance data from Facebook, Instagram, and Twitter
- Business Data – uses information about how you organize your business to provide reporting insights and automate campaign management
- Budget Management – organize and track budget pacing based on your business structure
- Ad Copy and Landing Page Testing tools
You can download this whitepaper for more detailed information on how to set up DS.
How Should I Test DS Bid Strategies?
We recommend that you test DS using AdWords Drafts & Experiments to set up an A/B test between DS bid strategies and your current method of bidding (automated AdWords or manual bidding). This will allow you to split your campaign’s potential traffic into two parts – an experiment and a control. We typically use a 50/50 split to test, but you can set up the split that best suits your needs.
In order to run a bid strategy on DS you need at least 30 conversions per day in each bidding portfolio. You need to pick campaigns to test that have enough volume to be split into two – so if you’re using a 50/50 split, each arm of your test needs 30+ conversions per day, so make sure the campaigns you choose to test have enough volume to start (60+ total conversions per day).
- Decide on test conditions:
- Conversion target(s)
- Bid strategy type based on your goals:
- Conversion actions/transactions
- A combination of ROAS and CPA-based goals
- Volume split
- KPI (e.g. increase spend at same CPA)
- Select campaign(s) (60+ total conversions per day, 4+ weeks of history)
- Add to a new DS label: “<experiment_name>: DS Bid Strategy group”
- Create a new paused DS Strategy
- Add the paused strategy on to your test campaigns (it will not begin optimization until enabled)
- In AdWords: create a draft and run an experiment for each of your chosen campaigns
- Experiment name = “<original campaign name>_experiment
- Repeat for every campaign
- Run engine sync in DoubleClick Search
- New experiment campaigns will appear in DS as the experiment name ( they will be added to the paused bid strategy by default)
- Remove new experiment campaign from the bid strategy
- Add newly synced experiment campaigns to a new DS label: “<experiment_name>: <legacy bid strategy> group”
- Wait 7 days to allow traffic to normalize: experiment runs with paused bid strategy
- If the volume split is not even after waiting 7 days, wait longer
- Launch Bid Strategy
- Activate bid strategy in DoubleClick Search
- Activate bid strategy in AdWords (if testing AdWords automated bidding)
- Allow algorithms roughly 7 days to calibrate/learn
- This period will be excluded from performance analysis
- The bid strategy overview tab will show when the strategy is out of calibration
- Monitor performance for 6 weeks
- This is the period where DoubleClick Search Bid Strategies and your legacy bid strategy will be compared to each other
- Relax constraints (min/max bid limits and budget) to make sure they are not limiting
- Avoid unnecessary target adjustments
- Pull Results & evaluate performance
- If DS is the winner, roll out the strategy to the rest of your campaigns!
Test Setup Notes
- Both arms of the test must be on Manual Bidding for at least a week before starting the test – even if you were running automated bidding before on AdWords. This is so both arms of the test are on an even playing field before being opted into automation.
- Don’t include RLSA in bid strategy during split testing; this will skew the results.
- Make sure modifiers that aren’t supported in DS (age, gender, etc.) are set the same in both arms of the test before launching the bid strategy. Continue changes in the manual arm as normal, but leave as is for the DS arm during the testing period.
- For a clean A/B test, make sure you’re optimizing to the same KPI in both arms of the test.
- Don’t make changes to the automated side of the experiment during the calibration period (especially the tCPA target). You can update the campaign budget if you are constrained.
- Monitor the bid strategy health report during the experiment.
And there you have it! We’ve found DCS to be a very effective part of SEM strategy, especially for larger clients, and recommend that you test it soon if you haven’t before. If you have any questions, feel free to drop a comment!