THE RKGBLOG

Google’s New Bid-For-Calls

How much is a click worth? It depends on the business, the keyword, the season, the engine, the time of day and a host of other factors. We’ve built an amazing platform for anticipating that value.

How much is a call worth, instead?

That’s the new question we and others need to tackle as Google rolls out its new bids for calls feature to desktop machines in AdWords. As Greg Sterling explains, participants must sign up for Google’s call tracking, and Google chooses the phone numbers. Bidding will only be available at the AdGroup level and there are click volume and call volume minimum thresholds that must be met for the system to work. Call-Through Rates will be a factor in Quality Score for participating advertisers.

The questions raised here are fascinating!

  • Will the calls be incremental, or would callers have clicked the ad absent the phone number?
  • Will the people who call be closer to a purchase decision or further away from one, since they’re seeking a human interaction?
  • Will the conversion rate be higher for the people who call than it would have been for those same people had they clicked instead? If so, by how much?
  • Will the quality of the calls generated by searchers who haven’t even reached your site be as good as the average call quality to your call center?

This last question strikes me as particularly important. The cost of a visitor to your website is pretty close to zero. You might pay for the the click, but having one more visitor on your website costs you nothing. However, a call to the call center costs you real money. You have to pay for a sales person to talk for some period of time. If the quality of these calls is low, you have a problem. If the volume of these search ad calls is material there also may be opportunity costs associated with having sales people tied up talking to bozos, while more qualified calls wait in the queue and sometimes stop waiting.

This feature is certainly worth testing, but I’m skeptical of this one. Click to call didn’t work well for many firms. Getting calls from folks who haven’t even been to your website to know what you have strikes me as likely to frustrate phone sales reps and be a money loser all-in. The calls from folks who are already on your website are one thing, but I have my doubts about the quality of calls from folks off of a SERP.

I’m particularly dubious given the high traffic requirements. If you can only use this on relatively high-traffic head keywords that are often generating the lowest quality, least targeted traffic it compounds the problem. If it ends up only making sense for Brand searches, what’s the point? Presumably your phone number is on your website as well, and wouldn’t you rather have them look around first, before they call?

Love to hear the reaction of others to this!

George

Technorati Tags: , ,

Comments
6 Responses to “Google’s New Bid-For-Calls”
  1. Tom says:

    A test we currently conducting with the bid for call is this, we have keywords with very high click costs ($9.00 +), adding the google phone number at only $1.00 per call in the short term seems a much better way to attempt to convert the customer at a much lower cost, if the calls are not junk and our call center can convert at the current levels. What I’m also waiting to see is what happens to ctr’s and quality scores if the ads are not being clicked as often.

  2. Really interesting, Tom. Let us know how that test turns out!

  3. Hi George,

    Any idea how this will affect Ad Rank? Could I actually bid my way up the page by increasing my Cost per Call bid on an Adgroup where I know I will receive very few calls, and effectively get more clicks (due to higher position), without increasing my CPC?

    I guess another way of saying this is: what is the new formula for calculating CPC? It used to be the AdRank of the guy below you, divided by your Quality Score + .01 (i.e. what you would’ve had to bid to get his position + .01). Now that Adrank is composed of 4 variables (CPC, clickQS, CPCall, callQS), how in the world would the cost be calculated?

    Thanks – really appreciate the sophistication and usefulness of the posts on this site – it’s clear you and your company are confident enough in your abilities and value that you don’t feel the need to always hide behind the wall of “proprietary” and get that the value you receive from an open dialogue often outweighs any increased risk.

  4. Hi Scott, thanks for your comments. We don’t worry too much about our competitors in the agency world. Few of the big players seem interested in pushing the envelope of performance.

    The bottom line is the combination of those 4 signals (Text QS + CPC and Call QS + Cost per call) will relate very tightly to Google’s revenue per impression for that ad, which is the core of Ad Rank. Just as with CPC a high bid for calls won’t get you a better ranking if the call-through rate is low.

    Can’t say for sure how they’re boiling it into Ad Rank, I suspect its weighted based on the fraction of those who call versus click. Which matters more will depend on which users engage with more.

    Interesting topic though! Let us know what your experiences suggest!

  5. Thanks, George – appreciate your reply. You guys definitely seem to stand out from the other players in this space, and I really enjoy reading your entries/articles. I’m pursuing getting more details from our Google contact, so I’ll be sure to post any more details I get here.

    It’s too bad Ad Rank is more complex now – what was so great about the old system was that if I was bidding $10 and saw my CPC was $5, I knew that, if I wanted to reduce risk, I could lower my bid to $5 and maintain the same volume of clicks (assuming marketplace was stable). Now, given that my CPC no longer reflects Ad Rank of guy below + .01, I can no longer use that technique.

    Am I thinking right here?

    Also, I get what you are saying about call-through rate, and agree that would impact any effect of a higher CPCall bid, but I guess the reason my question REALLY matters to me is that now, I’m not sure how to change my models that predict the effect of a simple CPC bid increase on volume of clicks.

    I typically model this as a simple one IV problem, with bid X quality score for the keyword as the independent variable (based on Google’s claim that they have equal effects on Ad Rank), predicting clicks as the dependent (I’ve played around with separately modeling the effect on position first and then modeling the effect of position on clicks, but I’ve yet to find a compelling theoretical reason to separate out these models (and I suspect it just introduces more error variance).)

    So, I’m wondering how should I model this now, given there are 2 additional variables that could effect Ad Rank. I can’t see my way around having to include them (even if I am primarily interested in clicks and calls are secondary), but adding two more independent variables will certainly make modeling the bid-to-click relationship much more challenging.

    This, and the other problem above, make me see bid-for-call as an unwelcome development, but perhaps I can be convinced otherwise eventually.

  6. Scott, I see your dilemma. To my thinking the first question is “Do you want calls to your call center off of SERPs when they haven’t been to your site?” If the answer is “no” (and I think it will be for the vast majority of advertisers) then you can punt on click to call and more or less ignore it. If it works for you then you may need to deal with it.

    One challenge with the assumptions you’re making is that there are really two stages to the auction: 1) qualifying for an auction; 2) competing within the auction. The notion that you can bid $5.01 in your example and get the same traffic volume generally isn’t the case unless the ad is on exact match and google.com only. What you find is that at the higher bid you qualify for a larger number of auctions on less tightly related searches and on less desirable syndication network partners. Might be lousy quality traffic, but you’ll definitely end up with more of it.

    I’d fall back on Bid Simulator data rather than the type of modeling you’re doing. Have you played with that at all? I’m not sure how it deals with the pay per call option, but we find it to be the best predictor of the incremental cost/traffic ratio.

    We have posts on Bid Simulator on the blog if useful.

    Best of luck!