Searching Beyond the Paid

Thursday, September 13, 2007

The Challenge of Estimating Paid Search Volume

There's a good thread at Search Engine Watch Forums that's been going for a month now about estimating search traffic, specifically paid search traffic. So far, the consensus is that it's really difficult - some would say even impossible - to do this accurately.

I've been struggling with this issue ever since we started in paid search over 5 years ago. Early on, it wasn't a big deal. I could just say "well, it's a new program with no history, so we really don't know what's going to happen" and that was good enough. Besides, the programs were small enough that they didn't hit the management radar screen too hard.

Now, however, paid search makes up about 50% of our total acquisition volume. It's definitely on the radar screen. And I've got a problem - albeit a good problem. Our fiscal year began July 1. Google volume has exceeded my forecasts by over 50% for the first 2 months of the fiscal year. And management wants to know why, and whether this will continue.

I know, I know - many of you would just say "take the money and run!" and not worry another second about it. But, realistically, forecasting is important - it affects how we run our business. Company priorities, income, expenses, cash flow - they all depend on reasonably accurate forecasts.

And therein lies the dilemma: How does one estimate search traffic with any amount of accuracy?

Do you look at your own history? We've been a Google advertiser for over 5 years, yet clearly history isn't an accurate predictor. I started with last year's results when I forecast this year, and it's not even close.

Do you use traffic estimation tools? Well, such tools might give a more accurate picture of the current search landscape - and then again, as pointed out in the SEW thread, they might be really, really inaccurate. Furthermore, how do you use these tools to forecast impressions, clicks, and conversions on 8,000 keywords? We use Google, Yaho9o, MSN, and Ask. 8,000 keywords on 4 engines is 32,000 different estimates. Unless you have an API with one of the tools, there's no way to go through and manually estimate traffic at this level of volume.

I've toyed with the idea of forecasting by gut feel - you know, look at current trends, close my eyes, and pick a number. I've stayed away from this before, because of course management wants to know how I arrived at my forecast numbers. But now that I'm being asked why we're up so much, maybe it's time to see how much of a PPC guru I really am!

How do you forecast paid search?

Labels: ,

0 Comments:

Post a Comment

<< Home