fbpx Skip to content

 

How we beat Pmax using an overlooked strategy

It used to be the case that our prospects would always ask about segmentation.  Thought leaders had entrenched the idea that very granular optimizations lead to the best results.  Often, they were right.

 

In the example above, if someone failed to segment the campaigns and bid on the channel based upon it’s overall ROAS of 5 being higher than their goal of 4, they would keep spending.  And that would only generate a 3 ROAS assuming brand is maxed out, causing them to lose money.

 

Segmenting brand and non-brand search terms in a text or shopping ad campaign leads to better decision-making.

Segmenting bids by keyword leads to better decision-making.

Segmenting Facebook audiences between New vs. Returning leads to better decision-making.

 

As ad automation has increased, advertisers have been bombarded with a narrative that the signals that ad networks have access to enable them to drive superior results.  This narrative mostly came from Google, but then some agencies and advertisers began testing automation and seeing superior results and jumped on the bandwagon.  The noise is now deafening and everybody wants to “test it,” if they haven’t already adopted it.  But what accompanies adoption is the elimination of segmentation.  That’s a big cost for a number of reasons.  The biggest is incrementality, but we won’t cover that here.  Instead we’ll focus on how we were able to use segmentation to beat Pmax.

 

If you’ve read our recent blog entry (https://conversionpath.com/2024/03/04/how-we-beat-pmax/ ), you would have learned two things:

  1. Pmax (Google’s hottest automated campaign type) is not even close to an apples-to-apples test against standard shopping campaigns
  2. We beat it using standard shopping campaigns.   By a LOT.  Despite it’s taking credit for brand text and remarketing revenue (that our campaigns didn’t do).

How did we do it?

For one, we used our GS+ keyword bidding algorithm, which is a powerful competitive advantage.

But here’s another thing we did using the idea of segmentation paired with knowledge about how ad tracking fails.

We know that different search terms indicate that visitors are at different stages of the buying funnel.  Someone searching for guitar parts has likely not yet determined what they are buying and where they are buying from.  Someone searching for a specific product name on the other hand, likely has.

                                    Funnel Stage    Volume            ROAS

Guitar kits                     Upper               High                 Low-Mid

Product xyz                   Lower               Low                  High

Guitar kits receives many more searches, is earlier in the buying cycle (upper funnel), and has a lower ROAS.  But here’s the important thing we know about this search term.  Because it is upper funnel, the visitor is not likely to buy when they click this term.  They are likely to peruse the site and do some research.  When they ultimately come back to buy, they may come through a brand ad or a specific product search.  The brand ad or product search will then steal the credit for driving the sale that the brand ad absolutely did not drive.  This is incredibly common that upper funnel terms are under-credited for the sales they drive.

Here we have a representation of guitar kits receiving a lower ROAS than the lower funnel term Part xyz.”  Most advertisers avoid or underbid the upper funnel term (in this case guitar kits) because it offers a lower ROAS, but this is the growth opportunity that has volume.  We apply an attribution factor to terms based upon their stage of the funnel, knowing that upper funnel terms have credit taken from them.  In the case below, we applied an attribution factor of 40% lift to the Guitar kits term to calculate an expected ROAS of 5.5.  This provides a more reasonable estimate of sales driven by this keyword, and enables the keyword opportunity to be grown.

 

How did we execute this?

  1. Estimate a lift factor

We analyzed click path reports to see how often upper funnel terms are followed by brand ad clicks.  We know that click path reports have a high failure rate for a variety of reasons.  By applying what we can see in click path reports, along with an assumption of failure rate, we come to a lift % estimate, in this case 40%

  1. Ramp up bidding on search term based upon expected ROAS

For shopping campaigns, we use our GS+ keyword bidding algorithm to achieve higher levels of coverage, which can be all the way to dominating the search results page.

  1. Accountability

We expect to see growth.  Often, that growth is seen in product name and other lower funnel searches.  But it’s also in brand ads.  Outside of the ad tracking environment, we should see lift in organic, direct, etc.  But whether or not a measured lift will be seen depends upon the scale of the individual search term.

The measurement is admittedly imprecise.  But you know what, all digital advertising measurement is imprecise, which is why our company built an attribution software platform.  If companies wanted to avoid investing in things that are imprecise, they would never invest in digital advertising.  What this strategy is, is logical.  It makes sense.  And thus it benefits those who use it, despite the difficulty in proving value.

While we can’t pinpoint the specific sales increase at a click level using this strategy, we can tell you it’s one of the significant contributors to how we beat Pmax performance:

 

Set it and forget it

QUESTIONS? We’d love to help. Fill out the form to get in touch now!

Read more

Pmax Implications

QUESTIONS? We’d love to help. Fill out the form to get in touch now!

Read more
QUESTIONS?
We’d love to help.
Fill out the form to
get in touch now!

    Back To Top