Guest post by Mike Nelson, Four15 Digital
Are you still doing manual bidding for your display advertising? If so, you might be spending a lot of time on something that can likely be done better through machine learning. Enter tCPA, or Google’s native bidding tool (previously known as Conversion Optimizer). tCPA, which stands for “target cost-per-acquisition,” is a tool advertisers can use to set bids that capture conversions closest to the price parameters you want to pay per acquisition.
tCPA bidding uses an algorithm that intelligently evolves over time as it learns about your unique business performance in order to optimize your bids and get you the most bang for your buck. It’s designed for advertisers who are bidding to a cost-per-conversion goal, so it’s best if you’re looking to generate leads versus making bids based on position, number of clicks, ROAS, or target impression share (which have other options available within Google). Below, we share more about why you should consider using tCPA—if you aren’t already—and offer suggestions for you to build and improve your bidding strategy.
tCPA versus Other Options
tCPA is native to Google, but there are many third party options also available for display advertisers. In some cases, these 3rd parties offer additional bidding options beyond target CPA bidding and target Return on Ad Spend, such as “maximize profit.”
Although Google cooperates with these 3rd parties, you should consider them competitive options to tCPA as it pertains to bidding. Some notable options include: Search Ads 360, Marin software, Kenshoo and Quantic Mind. These tools also provide reporting and other functionality which may be useful to you, and can execute bidding across-platforms, including on Bing Ads, Facebook and others.
That said, Google has at least three major advantages compared to the 3rd parties. First and foremost, Google has substantially more audience signal data. It’s recommended to review the signals Google uses (also listed below). Those which I’ve bolded do not have bid modifiers available directly in the Google Ads UI for modification. They are also not available via the API for a 3rd party to use.
- Physical location
- Location intent
- Weekday & time of day
- Remarketing list
- Ad characteristics
- Interface language
- Operating system
- Demographics (Search and Display)
- Actual search query (Search and Shopping)
- Search Network partner (Search only)
- Web placement (Display only)
- Site behavior (Display only)
- Product attributes (Shopping only)
- Mobile app ratings (coming soon)
- Price competitiveness (coming soon for Shopping)
- Seasonality (coming soon for Shopping)
Another advantage is that Google can do actual real-time, query level bidding. 3rd parties are restricted to pushing a single bid along with bid modifiers as an account manager would do. They generally make bid updates once per day. Lastly, there is no additional cost to use tCPA, whereas 3rd parties typically have minimum fees and percentage of spend upside. You could also argue that Google spends a lot more on researching and developing its flagship product, Google Ads, than any 3rd party could (especially with its investment in tCPA).
Types of tCPA Bidding
There are two main types of tCPA bidding: “standard” (tCPA) and “portfolio” (ptCPA).
Standard tCPA goes down to the ad group and/or campaign levels which allows an advertiser to come up with many different tCPA bids. There are no upper bound or lower bound cost-per-click bid limits available via standard tCPA advertising—that is, Google can bid as high or as low for a certain auction as it calculates in order to achieve its goals. Also note that Google is trying to meet your tCPA, not come in under or over it. This means that they may bid much higher on individual queries than your actual tCPA goal if they are beating the requested target.
The purpose of the portfolio option is to have a shared goal across campaigns (although multiple portfolio goals can exist). One nice feature of this is you are able to set max and min bids for portfolio tCPA bidding. A sensible bid maximum would be something lower then your tCPA. In general, since there is a lot of fluidity between keyword and query matches now-a-days, Four15 Digital uses multiple ptCPA bids across campaigns. That said, it’s always a test and learn process!
Migrating to and Managing tCPA
Much of the information above can be found in Google’s Help forum, yet there’s little detail around how to actually roll out and test tCPA for your business. Below, we’ll dive into the key strategies we use at Four15 Digital in collaboration with various Google account representatives.
Step 1: Running a bidding experiment
Once you’ve made the decision to test tCPA bidding, it’s critical to set your experiments up for success. One of the most common reasons for advertisers not adopting tCPA due to “poor test results” is incorrect set up. Follow these 7 unique tips to set up an effective experiment:
- Identify the right success metric: A common mistake advertisers make is to measure the success of a tCPA bid strategy test based on metrics the bid strategy is not designed to hit. Considering CPCs, impressions, clicks, CTR, impression share or other metrics in performance evaluation could be counterproductive. The bid strategy is designed to do one thing: Give you the maximum number of conversions at the CPA you desire. So the only way of measuring success is to compare the conversion volume of experiment versus control at a given CPA. Ignore everything else.
- Identify the right portfolio: Unless you have an account structure where individual campaigns represent unique business units with their own business level profitability/CPA goals, you would want to go ahead with a portfolio bid strategy. A common misconception is that in a portfolio, every campaign is expected to achieve the portfolio’s CPA goal. Nothing could be farther from truth. In a portfolio, campaigns may be operating in a wide range of CPAs, but the goal of the bid strategy is to achieve the target CPA as a weighted average across all campaigns. That is, the total spend of the portfolio divided by the total number of conversions achieved. The second key criteria to keep in mind is to ensure that the chosen portfolio has a sufficient number of conversions. While there are no guidelines on a minimum number of conversions, we believe that any portfolio with less than ~50 conversions in the past 30 days would throw extremely unreliable results. The more conversions you include, the more reliable your test. The volume of conversions dictates, to an extent, the length of the test period. The more the conversion volume, the sooner you will achieve statistically significant results with high confidence intervals and vice versa.
- Set up Drafts and Experiments: Drafts and Experiments is one of the most reliable ways of A/B testing your media strategies. D&E lends itself very well for smart bidding tests. It’s typically a good idea to split the test 50/50 so that the results are easily comparable. Two key things to keep in mind are:
- Do not change anything in one version without making the same change in the other. Ideally, just stay away from making any changes at all during the test period.
- Do not compare individual experiment campaign performance against its corresponding control campaign if you are running a portfolio test.
- Set up the CPA target based on historic values and not ambition: This is by far the most important best practice to follow if you want to give your experiment the best chance at success. Make sure you set up the CPA target equal to the actual CPA you’ve achieved during the past 2-4 week period. Once the bid strategy has achieved relative stability, you can slowly move from the current target CPA towards your desired target CPA by changing the tCPA by 10-15% every few days.
- Have a minimum test period of 3-4 weeks: As mentioned in tip #1, the greater the conversion volume, the sooner you can measure the bid strategy performance with high confidence. However, irrespective of the size of the test, we recommend running the test for at least 3-4 weeks. This gives the results higher fidelity and the bid strategy a genuine chance to learn and improve performance. You will notice that the first few days of the bid strategy are marked as “Learning.” We strongly recommend that you do not make any changes to the control or the experiment campaigns during the test as this would directly impact the bid strategy performance. Post learning period, we still advise to not make drastic changes to the campaigns or the business that the campaigns represent during the test period. This requires planning ahead of time and selecting the appropriate portfolio. Do not use portfolios that are already running other media/business tests or are expected to change during the test/learning period as this might muddle the results. Importantly, do not add/remove ad copies, keywords, etc. during the test period. Just leave it be and only make tCPA adjustments. Remember, the bid strategy is always learning (even after the learning period has ended) thanks to its machine learning algorithm. There have been instances when the bid strategy dramatically improved the performance 2-3 months after going live!
Scaling tCPA Strategy
Once you have run a successful test, it’s time to adopt tCPA across your entire account or MCC. A few things to keep in mind:
- Portfolio Selection: Assuming you are using a portfolio bid strategy, the first key decision to make is how you want to carve out your account into multiple portfolio strategies. As a rule of thumb, we recommend that you have a small number of large portfolios instead of a large number of small portfolios. Beyond this, the main influencer of this decision is your business objectives. If you have multiple products/services/business units with unique profitability/CPA goals (which you should), it is critical that each portfolio exhaustively and exclusively represent these products/services/business units and each portfolio tCPA reflects their respective goals. Conversely, if there are multiple products/services/business units but they have a common goal of achieving maximum profitability or a certain CPA, then this heterogenous group of campaigns can all be put together in a single portfolio.
- Pace of Scale: Depending on how fast you want to scale, you can adopt ~25% of your account spend every month over 3-4 months or you can adopt the whole account in one go. The latter is a good idea if your test was large enough for the results to be statistically significant.
- Goal Setting: Once again, it is critical that you set up the tCPA based on the actual CPA observed over the past 2-4 week for a given portfolio. Refer to tip #4 in the previous section for more details. Consider goal-setting and testing your desired CPA once your portfolio has achieved relative stability, typically within 1-2 months of testing.
Prioritize testing tCPA with these tips so you can reserve your time on other aspects of Google Ads (aside from bidding) and achieve greater results.