Paid Search’s Point of Diminishing Returns

“So, how much should we spend in Search?”

Years ago, when I was a siloed search engine marketer, that answer was unequivocally: “As much as you can spend!”. Return on Investment was through the roof and paid search was setting the standard for transparent measurement. “You only pay when someone clicks and they have to not only type in one of your keywords but feel the creative is relevant to them before clicking. That’s double opt-in qualified traffic.” Even though most accounts work on the Pareto principle (i.e. 80/20 rule where 80% of your traffic/costs/conversions are on 20% of your terms), the non-converting terms were tolerated if the account overall was kicking butt.

Folks, that was a bubble and that bubble has long been popped.

Positive ROI is no longer a given with SEM. Cost Per Clicks have risen exponentially, click fraud still exits, Quality Score measures make it more costly to run outside of your most relevant terms, and the slumping economy have forced advertisers to scrutinize online marketing’s favorite son as close as any of its other children.

The fact remains that paid search is a pull medium. It relies on users to actively search on keywords in your account and only a portion of those users even click your ads. Thus, for any paid search list of keywords, there’s a finite amount of volume available. Not only that, but if only 20% of your keywords are the truly efficient converters, there’s even less “good inventory” in search.

So, with this understanding, let’s say you had a $100K monthly budget for paid search and doing your data dive one day, you realized that the converting keywords consistently reached their volume at $60K at a fantastic Cost Per Action (CPA). However, was the other $40K you were spending on search was very inefficient?

In the chart below, assume that spend has been prioritized to the “top converting” words first. At $60K, you’ve exhausted the search inventory on your top converting keywords and now you’re spending on broader terms that cost more and convert less.

Even though the aggregate CPA on the entire account is under or near the goal, maybe the time has come to rethink that extra budget and how it should be spent?

As my good friend, Kenny Rogers, once sang:

You Gotta Know When To Hold ‘Em
Know When to Fold ‘Em
Know When to Walk Away
Know When To Run
– “The Gambler” written by Don Schlitz (1978)

A lot of my shift in thinking comes from my work with Conversion Attribution. As we move away as an industry from “the last ad click” methodology and start measuring all of the media interactions that influence users to perform our converting actions, it’s become clear that other engagement strategies should be considered within the media mix. In a recent campaign for one of my clients, 39% of all search conversions had a previous banner view in the mix. It’s very possible that a portion of those conversions would never had happened unless the banner drove a user to search.

On this example account, maybe the $40,000 of “inefficient” budget be better spent on:

• Content Targeting
• 2nd tier Search Engines
• Vertical Search Engines
• SEO
• Placement Targeting
• Banners
• Email campaigns
• Social Media
• Mobile marketing
• Video ads

I’m not saying that SEM pros should immediately do a Point of Diminishing Returns study on their accounts and stop buying on any keywords that don’t immediately hit CPA goals. However, I think it’s important to at least have that dialogue. General/Broad terms are important to advertisers as they reach users on the front end of the buying cycle and get them into the users’ consideration sets. So, more paid search may be the answer. It is double opt-in and users searching on these terms certainly have some interest to you. And reaching these users while they search has shown to be a very powerful tool.

But what if I told you that taking that budget and putting it into a top-notch social media campaign would ultimately drive users to double the search volume in Google on your top converting words?

The only answer here, as it always will be, is to test.