The Commerce360 Blog

This third post in the series isn’t about what’s wrong with ROAS as much as how ROAS is misperceived. As the only real measure of performance provided by the search networks and web analytics packages it’s easy to get the impression that ROAS is actionable information. But like a lot of other analytic metrics, it’s really only useful as the starting point for questions and investigations.

Is a 500% return-on-ad-spend good? Is a 159% ROAS bad? As discussed previously, that depends upon the net profit margin of the goods/services sold.

If ROAS is below a certain threshold for a particular keyword, should you delete or pause it? If ROAS is above a threshold is everything peachy-keen? The answer to both is probably ‘no’ at least until the current bid, text-ad, landing page, and offer pricing are considered, modified or tested.

It’s easy to fall into the trap of thinking that individual keywords are either ‘good or bad’. But the truth is that in paid search you’re measuring a system – the keyword, search query, text-ad, position, bid, landing page, conversion funnel, checkout process, product selection, and pricing offers and terms. While it’s obviously not possible to infinitely test every possibility for each of these variables, keeping them all constant and rendering judgment on the keyword clearly isn’t fair or reasonable either.

From a workflow perspective neither the search networks nor web analytics packages are of any help in measuring return (on ad spend or investment) as you perform a more complex series of tests of the other variables that impact the success of a keyword. The fact that all currently available tools fail to support are more complete and realistic view of the search environment clearly has a lot to do with the overly simplistic way most campaigns are still managed and reported on and evaluated.

Take one simple variable, text-ad creatives. Wouldn’t it be great if the revenue or ROAS graph for a keyword marked the points in time where different creative tests where run and showed the impact of those tests (v1 vs v2 vs V3)? Wouldn’t that be better than reviewing results and toggling to the text-ad report and then continually switching date ranges to check when ads were changed and how results varied? Or setting different tracking codes and comparing the tests to each other without being able to directly graph their progress?

ROAS is just one example of the conspiracy of false simplicity in the paid search world today. It’s helping the engines make a lot of money, and costing advertisers a bundle.

Posted by Craig Danuloff at 11:02 PM | Permalink | Comments (0)

For a long time, all we had were clicks. Paid search marketers bought keywords and ‘traffic’. The engines reported on click counts and costs. But they couldn’t tell you anything about sales and revenue.

Web analytics packages, on the other hand, identified paid search traffic and associate the engine and the keyword with the revenue. But they didn’t have cost-per-keyword data. So for some time we were left to eyeball (or painfully match up in spreadsheets) the interplay between cost and revenue.

Several years ago the two began to meet. Google and Yahoo added ‘conversion tracking’, and added API’s that enabled cost data to be pushed into web analytics packages. Suddenly, if you could properly tag your site (often no simple feat) and get the API’s to work properly (often an intermittent miracle) it became possible to see how many orders and how much revenue each $dollar of paid search (or each explicit keyword, adgroup, etc.) was generated. We entered the era of ROAS.

Compared with having no meaningful relationship between cost and revenues, this was a great advance. But as mentioned in the first post in this set, return-on-ad-spend (ROAS) is a very limited and superficial way to measure results and success.

The problem is that ROAS ignores the cost of goods or services (plus many other fixed marketing costs) and gives the impression that every dollar returned above those spent for the advertising is a positive result. A 400% ROAS, for example, sounds terrific.

But if your margins are 25%, to take one simple example, you need a 400% ROAS just to break even. The chart below gives further examples of the ROAS required to hit certain ROI depending on your average margin.

Most businesses don’t have consistent margins across their product lines or service offerings, so the simplicity of translating ROAS into ROI in your head isn’t practical. And with promotions impacting online pricing - coupons, discounts, and special offers – actual margin calculations can vary sale by sale. To further complicate things, ROAS is reported on a keyword basis, but revenue is often generated from a basket of products that themselves have different margins.

The deeper you look, the less relevant and less useful ROAS becomes.

Posted by Craig Danuloff at 9:58 PM | Permalink | Comments (6)

Paid search advertising is incredibly simple to purchase. But it’s nearly impossible – and in many ways literally impossible – to understand and analyze what you’ve purchased or how effectively that money has been spent.

Why? Because the fine folks who make it so easy to buy and pay simply do not share the information you need (and I’d argue deserve) in order to understand what you bought. And to a lesser extent because the after-market analytics packages take what they’re given and pass it on to you without adding any value beyond rich formatting.

What about ROAS you ask? Ah yes. Return On Ad Spend. The one aspect of your results that is easy to calculate and even supported by the search networks. ROAS measures whether you are ‘making’ more money than you are spending. It’s fine as far as it goes, providing an initial indication of the health of your overall program and perhaps individual campaigns.

But ROAS has a number of shortcomings and even flaws, and as a metric should really be killed once we finally get Page Views into the ground.

What’s wrong with ROAS? Here’s a partial list:

• As a measure of gross revenue it ignores the tiny little reality of COGS. In other words, you can have a large positive ROAS and still be losing money on every sale. A metric only an ad network could love.
• As provided it only tells you how a blended set of search terms, positions, text ads and landing pages are performing and gives no visibility or account for the prior or subsequent actions of site visitors. As a result, it’s impossible to have an intelligent surgical reaction, and very likely to drive actions that produce quite unintended consequences.
• With its basis in the advertising spend, the easiest way to drive up ROAS is to cut certain spending. Mathematically that often doesn’t yield the highest revenue or profit. It’s like driving a car and optimizing for miles-per-gallon without regard for your destination.

In the next three posts I’ll take a deeper look at each of these flaws and talk about things you can do to overcome them.

Posted by Craig Danuloff at 5:55 AM | Permalink | Comments (2)