As we discussed in Chapter 3, automation has changed all of this. We should no longer assume that bottom-funnel search terms are more valuable than top of funnel search terms. However, this traditional approach is still useful for many accounts.

Behavioral targeting has less to do with the medium and, instead, relies on customer behavior to determine the timing, placement, and audience of an ad. This placement favors the who.

The NASA engineers refer to this analogy as the H-Metaphor, which can be described as such: the humans chart the course, then rely on machines to manage the real-time calibrations.

We refer to a confident algorithm as one that has run enough tests in the learning phase to accumulate statistically significant confidence intervals regarding its hypothesis. However, just because a machine is confident does not mean the machine is accurate.

You can also choose to pay for conversions instead of clicks on Smart Display campaigns that use Target CPA bidding.

It is likely that ad recall on YouTube skips is so powerful because, in order for a user to skip the ad, they have to actively engage with

In every case, allowing machine learning algorithms to process as much data as possible improves liquidity.

At this point in the game, your competitors have the same basic knowledge as you do.

A client will leave if they feel that the value your agency provides has declined over time. Campaign performance might be strong—phenomenal even—but if your perceived value has declined, the client will choose to terminate the relationship in search of a more proactive agency, or decide to bring the efforts in-house.