This is the subhead for the blog post

Let’s face it: not all clicks are created equal. Search is full of irrelevant traffic, both human and bot. In the search engine marketing (SEM)/pay per click (PPC) world, bid adjustments are often the focal point of many optimization strategies intended to target a specific audience.

Bid adjustments are a percentage increase or decrease applied to almost any aspect of an account – campaigns, ad schedules, locations, placements, device, etc. Using bid adjustments to hone in on when, where, and how your target audience converts can exponentially increase account performance. Google’s Conversion Optimizer has seen improvements recently and is a viable option for many search campaigns.  However, manual bidding can still outperform CO in many instances, especially for campaigns with drastic fluctuations in seasonality and/or data sparsity.

When researching manual bid adjustment best practices, most articles/blogs focus on using one metric to judge performance and subsequently create bid adjustments. Popular metrics for this include return on investment (ROI), cost per acquisition (CPA), conversion rate (CVR), and click-through rate (CTR). However, this methodology is too narrow-minded to be effective. There is no single metric that can serve as the ultimate source of truth for an account’s performance in a complex digital landscape.

For example, when applying bid adjustments to an ad schedule, 3am on a Tuesday could have an excellent CTR but extremely low search volume (impressions and clicks), thus making it a bad candidate for a positive bid adjustment. Same with location; one person could have converted in Gila Bend, AZ (population 1,977), making the ROI look great, but this is not an accurate indicator of future conversions. Additionally, clients will often have more than one KPI goal to optimize to.

Thus, in order for bid adjustments to be successful, they need to consider multiple metrics and prioritize those metrics based on client goals. To do this, we can use multiple criteria weighted averages to aggregate a specific dimension’s performance across different metrics. This will allow us to prioritize certain dimensions over others and systematically optimize bid adjustments.

How this works, mathematically

Weighted averages are an average resulting from the multiplication of each criterion by a weight reflecting its importance.  For our purposes, the criteria will correlate to various SEM metrics (CTR, ROI, CVR, etc.).

If we had three criteria (X₁, X₂, and X₃) and wanted to weight each .5, .3, and .2, respectively, the weighted average equation would look like this:

WA = .5(X₁) + .3(X₂) + .2(X₃)

What this equation is doing is assigning importance to the criteria through weights. Each criterion must be measured on a similar scale in order to compare apples to apples – in other words, you cannot directly compare CTR to ROI. You can include as many criteria as you want, but the sum of the weights must equal 1. It is up to you to set the relative importance of the various criteria through weights. The weights should reflect the relative value, going from most important to least important. The final outcome will allow us to rank specific dimensions to see which are the overall best and worst performers.

For more on weighted criteria, check out Ask Dr. Math.

How this works in practice

Let’s assume we are trying to optimize bid adjustments for an ad schedule (this method will work for all other dimension-specific bid adjustments such as device, location, placements, etc.). Let’s also assume that our client’s KPIs are prioritized as such: 1) maximize conversion volume, 2) minimize CPA, and 3) maximize CTR. Below is a snapshot of a fictitious data set for campaign performance by time of day and day of week:

The first step is to measure our 3 criteria (conversions, CPA, and CTR) on similar numerical scales. 0-100 works well as it is easiest to calculate (where 0 is the worst performance and 100 is the best).

CTR – since all of the CTRs above are less than 1%, we can use that as a benchmark. Working with the first row:

CTR Score = (0.0054/.01)*100 = 54

Conversions – Since all of the conversions are less than 10, we can use that as the benchmark. Again working with the first row:

Conversion Score = (2/10)*100 = 20

CPA – This is a bit tricky considering that a higher score should go to a lower CPA. Thus we will use the inverse. All of the CPAs are less than 400, so we will use that as the benchmark.

CPA Score = (1-(274.52/400))*100 = 31.37

Step two is to calculate the overall score with the aforementioned weighted average equation. You will need play around with the weights according to your KPI priorities. This will require a deep understanding of your account, knowledge of KPI goal prioritization, and common-sense intuition. If the final bid adjustments do not seem quite right, try tinkering with the weights. For this example’s purpose, we will use the following weights: .5 for conversions, .3 for CPA, and .2 for CTR.

Overall score = .5(20) + .3(31.37) + .2(54) = 30.21

Step three is to repeat the first two steps for the remaining rows of data:

Step 4 is to calculate bid adjustments. In order to do this, we need to compare the overall scores to a benchmark score. The average of the overall scores is a good benchmark to use, especially for dimensions with a lot of subsets (such as ad schedules). This will tend to result in an equal distribution of positive and negative bid adjustments. If you want to skew your bid adjustments in either direction, adjust the benchmark to above or below the average. Unlike ad schedules, dimensions with fewer subsets, such as device (desktop, mobile, tablet), will tend to need a lower-than-average benchmark. Play around with the benchmark and use your best judgment – do not forget what is and is not good performance.

The average of the entire data set’s scores in our example is = 35.84 (not all data shown in snapshots above). The bid adjustment calculation for the first row is:

Bid adjustment = ((30.22/35.84)-1)*100 = -15.69%

Because most search engines require whole numbers, we will round to -16%.

If you were working on other dimensions, you should be done! Apply your bid adjustments.

Ad schedules require a little more work as each day of the week can be segmented into six time blocks. To calculate the entire ad schedule, I set consistent time blocks for each day to fill a 24/7 schedule and averaged the bid adjustments of each block.

Conclusion

Overall, multiple criteria weighted averages allow us to factor numerous metrics when judging performance. This is particularly useful when making optimizations such as bid adjustments because it produces a holistic view of overall performance in one simple number. The most powerful aspect of this methodology is that it can be applied to everything: campaigns, ad groups, keywords, ad schedules, locations, devices, etc. Keep in mind that there are limited opportunities to experiment with your weights as there are many other factors that will affect account performance. Additionally, the bid adjustment calculation is a comparison relative to the benchmark you choose. If, for example, your account is performing great across all three devices (desktop, mobile, tablet), you will want to set a low benchmark so that each device has a positive bid adjustment.

Good luck and I hope you found this helpful! I’m happy to answer any questions, so please leave a comment.