Last Click Is Flattering Your Best Channel. How to Tell.
One channel in your report has an excellent cost per conversion. It has had it for months. Everybody knows it is the thing that works.
Under last-click attribution, the channel that gets the credit is the channel that happened to be there at the end, and the end is the cheapest place in the funnel to be. That is not a flaw anyone introduced deliberately; it is what the convention does.
This page is about telling the difference between a channel that is creating demand and one that is closing it.
Why last click flatters the closer
Somebody hears about you somewhere, thinks about it, searches your name, and buys.
The search on your name gets all the credit. Everything that produced the intention gets none. If the intention was produced by something you also paid for, you are now reading a report in which the expensive thing looks bad and the cheap thing looks excellent, and the cheap thing is at least partly harvesting the expensive thing's work.
The channels that benefit most from this are the ones that tend to be last: brand search, direct visits, retargeting, and email to people who were already customers.
The channels that lose most are the ones that tend to be first: broad paid search on non-brand terms, paid social, display, video, and anything doing discovery.
The three checks
None requires a tool and none produces a percentage.
1. The brand versus non-brand split
The most useful five minutes in this whole subject and it is startling how often it has never been done.
Separate your paid search results into terms containing your brand name and terms that do not.
If most of your conversions and most of your efficiency are in the brand half, you are looking at a channel that is largely closing demand rather than creating it. That does not mean the spend is wasted; it means it is not doing what the report implies.
If your brand half is small and your non-brand half is producing conversions, your paid search is genuinely acquiring.
Whether to bid on your own brand at all is a separate and contested question, and our article on bidding on your own brand name sets out both sides without picking one for you.
2. Look at the paths, not the totals
If your analytics shows conversion paths, look at what appears earlier in them rather than at what appears last.
A channel that appears frequently as a first or middle touch and rarely as a last touch is doing work that last-click cannot see. That is the specific finding this check produces, and it does not require you to switch model, only to look.
3. Turn something off and watch the other thing
The only check on this page that produces evidence rather than an inference, and the only one with real risk attached.
If a discovery channel is genuinely creating demand, pausing it should show up later in the channels that close. Brand searches fall. Direct visits fall. Total conversions fall by more than the paused channel was credited with.
If nothing else moves, the channel was not creating anything you were subsequently harvesting.
Four rules if you do this, because it is a spend decision and it is the one place on this page where getting it wrong costs money.
Change one thing. If you pause two channels at once you learn nothing about either.
Give it long enough for the delay. Demand created this week converts over following weeks, and a two-day test measures nothing.
Watch the total, not the paused channel. The paused channel's numbers going to zero is not the finding.
And know what a bad outcome costs you before you start. Our article on when to pause a campaign covers thinking that through properly, and it applies here with more force because you are pausing something deliberately rather than because it was failing.
What this does not mean
Three overcorrections, all of which are common.
It does not mean brand search is worthless. Someone searching your name is a person you may lose to a competitor bidding on it, and defending that is a legitimate reason to spend. It means the credit it receives overstates the demand it created.
It does not mean you should switch attribution model. A different rule applied to the same incomplete data produces a different set of assumptions, and our article on attribution models covers why for most accounts the model is not the binding constraint.
And it does not mean your discovery channels are all secretly working. Some of them genuinely are not. The point is that last click cannot tell you either way, so a channel you cut on last-click evidence was cut on evidence that could not answer the question.
The number this page will not give you
You will find pages quoting a percentage for how much brand search is incremental, or how much of a channel's credit is really borrowed.
No such general figure exists. Published incrementality results differ enormously between businesses, categories and brand strengths, and a number derived from somebody else's account is a number about somebody else's account.
This site publishes none. The three checks above are the version that produces an answer about yours.
What to actually do with this
Run check 1 today. It is free, it takes five minutes, and for a lot of accounts it changes how the whole report reads.
Run check 2 if your reporting supports it.
Run check 3 only when a real decision depends on it, one channel at a time, with the cost of being wrong understood in advance.
And report the split rather than the total. A report that shows brand and non-brand separately is a report where nobody can accidentally conclude that paid search is producing demand it is closing, which is the specific misunderstanding that gets budgets moved in the wrong direction. Our article on what wasted spend actually looks like in an account covers the neighboring version of this problem, where the number that looks efficient is efficient at the wrong thing.