How to Measure PPC Cannibalization

In this article I’m going to advocate that you turn off your PPC entirely.  Seriously.

Attribution Modeling Shmattribution Modeling

You can speculate on the true lift from Adwords and other paid search all you want, but there is only one true way to figure out Adwords net effect.  I’m not talking about last click, I’m talking about NET effect.  This includes first click, last click, assists, view-thru, and any other possible intangibles.

There’s only one way to figure that out and that is to pause all your paid search.

Simply put:

Performance with Paid Search

– Expected Performance without Paid Search

=Net Lift

Prerequisites

1. PPC is not the only source of traffic. If you know that PPC is 90% of your traffic, it’s safe to say that this experiment is unnecessary.

2. You can afford to take a hit for a few days.  You will have to pause for long enough to get good data.

3. You have enough volume for statistically significant data.  I wouldn’t bother if your business is doing less than $5,000,000 in annual revenue.  Work on increasing your revenue instead.

4. If your data tends to bounce around week over week, then the results of this experiment will be inconclusive and not worth the trouble.  You need to be fairly certain of your “Expected Performance Without Paid Search.”

 

 

1. Select a time period to use as a baseline.  This should be a very stable, almost static time period when you are not running any special promotions, changing the website around, building links, etc.  Watch out for seasonality as well.  You want as predictable of a performance as possible.

 

2. Record all the relevant metrics for this time period.  Traffic, sales, sign ups, calls etc.

 

3. Pause all PPC for the same time period.  If you stayed paused Monday – Sunday, make sure your PPC paused time period is also Monday – Sunday.  Record all relevant metrics.

 

4. For each of your metrics, subtract one total from the other.  Divide the amount of money you would have spent by the difference in each metric.  Now you have an accurate assessment of your cost/visit, cost/sale, cost/lead.  Now, take the difference in sales, and divide it by the amount of money you would have spent for the given time period.

 

If we narrowly define ROI as Money In/Money Out, then this is your true ROI.

 

How long should your time period be?

 

It is tricky to assess the length of the time period.  On the one hand, the longer your observational period, the better the data.  On the other hand, the more revenue you forego, the more revenue you forego.  Select the minimum length of time where your data is relatively consistent.  One week should be fine for most businesses receiving a few hundred visits a week.

 

Example

In the beginning of April, a Top 200 Internet Retailer asked me to determine the true ROI of their PPC efforts (about $10,000,000 in spend annually).  I looked through their data and confirmed that about 40% of their revenue was coming through Adwords and Bing based on a few different attribution models.  They were sending about 4% of their annual revenue on Advertising.  What is their ROI?

 

40%

4%

1000% ROI

 

This company dominates in SEO.  They rank extremely well organically in both Google and Bing for many of their top search terms.  Their conversion path is complicated and involves multiple touches from a variety of sources.  I suggested the unthinkable: let’s pause PPC entirely and see what changes.  You can create as many attribution models as you’d like, but nothing will beat a controlled experiment.  Guess what, with their PPC entirely turned off, sales fell by only 25%.

25%

4%

625%

 

The real ROI was actually 625%.  Let’s look at this in terms of advertising margin.

We will define advertising margin as (revenue-cost)/revenue.  This is the percentage of the revenue that is left after advertising cost.

 

1000% ROI translates to (40%-4%)/40%=90%

625% ROI translates to ((25%-4%)/25%=84%

 

Since this company is such a large internet retailer, their actual profit margins are slim.  It just so happened that after all expenses, their profitability was about 15% excluding advertising expenses.  This means that they went from making about 5% per transaction to losing 1%!

For a company doing over $100,000,000 in sales annually, pausing PPC for a week is a $2,000,000 hit to their top line revenue.  However, the lessons learned were well worth it, and we instantly began lowering budgets for campaigns where this cannibalization was the highest.  The company will get its money back for this experiment within 6 months.

Speak Your Mind

*