Keyword intelligence: why Google Traffic Estimator is dead wrong

If you’re an entrepreneur thinking about entering a new market niche, you’re probably also wondering how much it will cost to get people onto your site. If you want to size the market and figure out how much AdWords would cost you might turn to Google AdWords’ Traffic Estimator for metrics.

You’re guaranteed to get the wrong answer.

For several years I helped financial services companies build out their online marketing strategies and the first question a new client would ask me is, “How much traffic will we get and how much will it cost?” I used to use Traffic Estimator to give an answer but I eventually realized how wrong the estimates are.

To demonstrate this point, I’ll pick a set of keywords where I have data from real sites with big budgets and compare their results to what Traffic Estimator projects.

For a set of keywords related to “auto insurance,” here is what Traffic Estimator predicts.

Summary (per month)
Total Estimated Clicks
2370 - 3000
Average Estimated CPC
$6.13 - $7.26
Total Estimated Cost
$28,024 - $47,343
Estimated ad position
1-1.5

For those same keywords in February 2011 (a typical month), here was the true daily result.

Summary (per month)
Total Clicks
20,130
Average CPC
$4.29
Total Cost
$86,580
Ad position
3.2

Now the sceptic in me says, well, maybe the quality of the ad was better than average so the campaign got more clicks. Ignoring the fact that the campaign in question got 7x more clicks than the estimate, let’s look at Estimator’s projection for Impressions, not clicks. There, Google should be spot-on since the ad itself would make no difference.

Traffic Estimator’s estimated Impressions:
1,807,523
Actual Impressions:
267,399

I’ve observed these order of magnitude mis-estimates from Traffic Estimator for a wide variety of keywords.

Instead of relying on Traffic Estimator’s predictions, when you need to price and size a new market:

If you have some dollars to waste, write a decent ad and pay for enough click throughs to also get a handle on CPC. As an untrusted advertiser, your initial CPC will have a risk premium baked in so you can safely estimate a long-term steady state CPC that will be 25-50% lower.

 
6
Kudos
 
6
Kudos

Now read this

How to choose the optimal domain name using Adwords split testing

Choosing a new domain is usually a gut-driven decision. Maybe you’ll spend a few hours getting domain suggestions, check out a few Sedo auctions, build a shortlist, ask some coworkers for opinions, then pick one and hope for the best. It... Continue →