Oct 282008

The PPC Buying Cycle: Buyer Beware

This was posted at SearchEngineLand yesterday in case you missed it:

How often have you heard that Keyword level performance data can be misleading? That PPC managers need to consider the phases of the buying cycle when evaluating terms? That specific keywords tend to steal conversions from the more general keywords that started the customer's consideration?

It's pretty obvious why the engines trumpet this story: it makes them money. By convincing advertisers that they should spend money on general search terms regardless of the observed efficiency advertisers are encouraged to spend without the moorings associated with ROI goals. Is it surprising that Google's quarterly profits were strong in spite of the fact that most advertisers are struggling?

Google and Compete Inc. presented a study of this subject that managed to answer none of the salient questions.

We presented a study almost 3 years ago at SES that challenged the notion of the buying cycle, but decided it was time to revisit the topic.

First, we grabbed data from a number of retail clients representing different verticals and different business models (pure plays, catalogers, brick and mortar retail, etc). We then sought to answer the questions:

  1. How often are multiple ads touched prior to an order?
  2. Does the interaction happen the way they say it does?
  3. Do these data suggest that crediting the first touch or the last touch or some smeared credit allocation would be closer to the truth? and finally,
  4. Do the different types of retail businesses witness different behaviors?

First, let's talk about cookie windows. The longer window of time you allow the greater this effect appears to be. But for a retailer, unless your AOV is huge, you need to place some "make sense" limits on how long to give credit to an ad. Most of our clients have windows of 14 to 30 days with some shorter and some longer depending on what the data suggests. Out of respect for the argument that there is this long consideration cycle, I went ahead and looked at a 45 day window for these clients.

Then, recognize that we're only really interested in orders with more than one different non-brand keyword involved. We define "brand" as the retailer's trademark and variants exclusively, hence "Sony Walkman" is a non-brand phrase for Best Buy, but a brand phrase for Sony.com. For well known retailers there is a 4% - 8% impact of non-brand clicks being followed by brand clicks as customers remember that they found the perfect ring at Zales. We count these as non-brand orders, and uninteresting for this purpose unless there was more than one non-brand keyword involved.

With those parameters understood, we found that interesting cases of multiple non-brand keyword touches occurred in between 10% and 15% of the orders for the vast majority of retailers. A few higher, a few lower. That's not zero, but it's not earth-shattering either.

More interesting: when we spent some time studying the impacted orders we found that in only about 35% of the cases did the story play out as advertised, with general searches being followed by more specific searches; in about 15% of the cases the other keywords were simply slight variations on the initial search (eg: "VCR sales" and "Buy VCR", or "Nikon lens" and "Nikon lenses") and in 50% of the cases the initial keyword had no relationship whatsoever to the last keyword prior to the order (eg "loveseat slipcovers" and "gold earrings").

The more we extend the cookie window the greater the propensity for the keywords to be unrelated, indicating that, in fact these were separate shopping events, and that person looking for loveseat slipcovers, bought from someone else :-( Our research on what the person actually purchased was anecdotal, we just tested 15 or 20 samples and found that in each case the order related to the last keyword, not the first.

Those sites that appeal to hobby enthusiasts see more multi-touch interactions than general retailers. Indeed, the impacts of AOV, business type and vertical were all quite interesting.

To our thinking this confirms several tenets of Search Marketing strategy:

  • The engines are not evil, but neither should they be expected to look out for your interests;
  • Cookie Windows matter and should be carefully considered.
  • Within Paid Search, "last click" credit schemes seem to be a better proxy for the truth than either "first click" or "shared credit"
  • If Keyword level performance data suggests a keyword is underperforming, it probably is.

Researching your own search data is a valuable exercise. It may help you determine whether the Buying Cycle should impact your keyword efficiency targets, or whether, for you, it's much ado about nothing.

See for yourself!


15 Responses to "The PPC Buying Cycle: Buyer Beware"
Brandon says:
Are you in favor of Max CPC (manual) bidding in Adwords as opposed to Preferred Cost Bidding (average cpc)?
Jim Novo says:
George, tracks with what I have seen. I wonder what this says for the supposed influence of Display on Search performance? If there is no real "buying cycle" to speak of, then how can Display influence Search in a meaningful way? Seems to me it's more likely that people who are Searching on a topic end up on web sites where Display is pointing, and because there is no cross-cookie between Display and Search, one could come up with "view through" as a concept which leads to Display influencing Search. I wonder if Google-Click will let us know what the real story is on this? Umm...perhaps not.
Hi Brandon, yes we advocate setting a max bid based on your economics. In every case that you give Google the ability to make decisions for you -- like preferred cost bidding -- you're guaranteed to waste money. Say you can afford to spend $1 per click on an ad, but due to the cost landscape at the time you only pay $0.75 for the traffic -- do you really want Google to decide to overspend for the next X clicks to bring the average back to $1? I don't think so. Jim, I'm actually arm deep in research on multichannel activity and it shows more interaction between channels than we've seen within Search. Indeed, the research suggests that how those allocation schemes are set up REALLY matters. More to follow to the extent that NDAs allow :-)
Jon Clark says:
I've actually been involved in a somewhat heated battle with our IT team. They have decided (without consulting with me prior) to change the cookie tracking from 30 days to 10! I believe this can't possibly beneficial for tracking purposes but also skews analyzing historical data for comparison purposes. Bsaed on your research, would you agree a 10 day cookie is too minimal?
Hi Jon, thanks for your question! I have a post coming out later this week on how to set cookie windows. 10 days is probably too short if your AOV is more than $150 or so and your products are "considered purchases", but may be too long if your AOV is really low and your products are more "impulse buys." My post, which is half way written, will outline how we find the "right" window for a particular client. One thing we've done and I'll advocate in that piece is to set "permanent" cookies and handle the cookie window through reporting algorithms. That way if you change the window to 7 days or ten days and decide that was a mistake, you can actually recast history and see what the results would look like under different scenarios.
Matt Trimmer says:
Hi George, Just found your site and think it's really excellent. Never seen such a site that attempts to answer questions that others don't even dare ask! We are a Google Analytics Authorized Consultant and one thing that has been interesting me lately is this blog post at one of our fellow GAACs. My question is one to really stimulate what system, in your opinion, is providing the best picture of conversion? GA with "last click" or Adwords CT with "first click".
Matt Trimmer says:
Hi George, Just found your site and think it's really excellent. Never seen such a site that attempts to answer questions that others don't even dare ask! We are a Google Analytics Authorized Consultant and one thing that has been interesting me lately is this blog post at one of our fellow GAACs: http://www.roirevolution.com/blog/2007/06/adwords_conversion_tracker_or_google_analytics_whi.html My question is one to really stimulate discussion about what system, in your opinion, is providing the best picture of conversion? GA with "last click" or Adwords CT with "first click".
Hi Matt, thanks for your kind words. At this point I'm taking the rather odd-sounding position that within a channel, credit should go to the last ad, not the first, but then between channels folks should credit the first touch within a reasonable cookie window rather than the last. I'm going to have to defend that notion in a full blown post, and I'm not entirely sure it's defensible :-). Historically Direct Marketers have found crediting the first touch to be wise, and I certainly am of that school of thought, particularly with affiliates "stealing" credit at the shopping cart. From an evolutionary perspective, I think we're heading towards smart ways to share credit, rather than all or nothing allocations. The tools for smeared allocation are pretty limited at this point, but I think it's the right direction.
George, THANK YOU! My sentiments exactly. I sat in on the same presentation that they co opted with Compete.com and I walked away feeling used. They went round and round about the phases of the buying process and how non-branded search is extremely important to driving the lead/sale. Yet for over 2 years a major brand of ours continues to struggle on non-branded traffic. I wrote up a good summary of their talking points that can be found here http://www.breaklinemarketing.com/blog/?p=11 It was laden with unsubstantiated facts and theories on non-branded's influence on getting the buyer in early and closing later with the branded terms.


Check out what others are saying...
[...] The PPC Buying Cycle: Buyer Beware, Rimm Kaufman [...]
[...] The PPC Buying Cycle: Buyer Beware [...]
[...] We’ve been trying to debunk this myth for years, but it has strong [...]
[...] PPC buying cycle is one of those notions we’ve tried to put in proper perspective. The buying cycle argument [...]
[...] probably isn’t the right way to think about the phenomena. As we’ve noted previously, most PPC buyers touch only one ad before buying. The fact that they used only that one engine is perhaps less compelling evidence of [...]
[...] #1: In retail, the data just doesn’t support the explanation. I’m told it’s more prevalent in other verticals, but in retail this only explains a [...]

Leave A Comment