THE RKGBLOG

Search Behavior from the RKG Labs

My monthly paid search column for SEL in case you missed it:

A couple of months ago I made some observations about search behavior that I witnessed first hand. Namely, that people search differently from each other and that these differences might explain conversion rates and traffic volume on different types of keywords.

My wife shops online differently than I do. When she’s looking for a blouse she goes to Landsend.com, Talbots.com, Chadwicks.com, etc. She doesn’t type urls into the address bar, she searches for the brands by name. She prefers the natural links to the sponsored links. I tend to search for the product or service by name.

In general, she searches differently than I do. We raced each other to determine the rest of the famous quote “To be or not to be? That is the question…” I found it on the first click by searching for “hamlet ‘to be or not to be’ soliloquy”; it took her 4 or 5 tries on different links that came up when she searched for “Shakespeare Quotes”.

The question that naturally came to mind was: “Is this why conversion rates on general terms are so low, and why the conversion rates of much more specific terms is so much higher?” We and everyone else who has studied the data have rejected the paid search buying cycle. Indeed the fact that the volume of sales on the general terms is so high indicates plenty of purchase intent.

The conversion rates are low not because of lack of purchase intent, but because the more general the search, the less clear the user’s intent and the less chance there is that the advertiser carries what they seek.

This is a fine theory, and I went on to posit that there were three searcher profiles: Trademark Tracy, who prefers to search by store names; Yellow Page Paul, who searches for the category of product or type of store rather than the product itself; and Modern Mary who searches for exactly what she wants using longer than average phrases.

However, I didn’t have time to do the research to support those theories.

Today, we’ll present some research that suggests we might be onto something.

METHODOLOGY:

I took a random sample of retail clients and grabbed every instance in which:

  • there were exactly two paid search visits through our ads, and
  • the ads were for two different clients

By insisting on ads from different clients we remove any biases there may be for people to shop for this particular brand or this particular product category differently than they shop for other categories. Here we’re trying to determine whether people who come through a brand ad for me are more likely to do other searches by navigational searches as well.

We classified search phrases as being either: a “Brand” search for our client’s trademark; a “Head” Keyword search defined in this case as generating more than 500 visits in a 90 day period; or a “Tail” Keyword generating fewer than 500 clicks in 90 days.

We then studied the relative frequencies of first search to second search transitions.

To re-iterate: this is NOT about conversion funnels or buying cycles, these are clicks on ads from two different clients selling different stuff.

FINDINGS:

The 65,000 instances studied involved 130,000 ad clicks (2 * 65K).

The majority of the ad clicks were on middle to low traffic “Tail” keywords for this particular sample.

When we study the breakdown based on what was first touched the results are fairly compelling:

Note that people who used a brand ad to visit one of our clients are almost twice as likely to visit the other client through their brand ad as well.

When the first touch is a “Head” keyword we see a 50% greater likelihood of their next visit to an RKG client being a head them than one would expect, and much less than average likelihood of them using a tail term next.

Finally, those whose first visit to an RKG client was through a low to mid-traffic tail term are more likely than average to visit the second RKG client through a similarly specific keyword.

CONCLUSION:

Conversion rates may be more a function of the difficulty users have in clearly expressing what they’re looking for than any absence of buying intent. While this cursory pass at the data is less convincing than it could be, it does suggest we may be onto something here.

Love to hear what others think!

Comments
8 Responses to “Search Behavior from the RKG Labs”
  1. This certainly makes it clearer why Google put so much effort into localizing & massively improving Google Suggest recently. If Google can help enough users better express what they’re looking for via G Suggest, it’ll help advertisers and thus drive back up otherwise moribund CPC’s.

    Interestingly, [now-defunct] UK SEM agency Latitude put out data (July ’09) on the impact of Google Suggest on search volumes:

    http://www.latitudegroup.com/blog/we-do-want-what-google-suggest/

  2. Thanks for your comments, Chris,

    Search suggest is helpful, but only goes so far. When Yellow Page Paul types in “furniture” when in fact he’s looking for a three-piece leather sectional sofa, search suggest doesn’t really force him to be more specific, it just adds a few words like “Store”, “stores near [location]” and the names of companies that start with “Furniture”. Helpful, but not the drill down options needed to really target his inquiry.

    Studies showing that query lengths are increasing suggest that either people are slowly learning to be more specific, or that younger generations that grew up with search are becoming a larger chunk of the searching population.

    In either case, the money best spent might be on user education. Google and Bing could run ads preaching the merits of typing exactly what you want, showing examples and how it helps folks find what they want quicker. Not sure how that would work, but it’s a thought…

    G

  3. Terry Whalen says:

    Hi George,

    I enjoyed your post – it’s interesting data. If we were to assume that these findings are representative of purchase intent relative to the head and tail generally (across many different categories), how might one act on this data?

    In other words, is this academic, or might it lead to a modified way of running (or testing) SEM campaigns? I’m trying to figure out how to connect the dots here, and I’d love some help.

    Thanks,

    Terry

  4. Hey Terry, thanks for your comment.

    To be frank, I’m not sure what “action items” flow from this analysis. The notion might be more for the engines: they should figure out who knows how to express their intent and who doesn’t and offer more guidance to those who historically have a hard time finding what they seek. “Are you interested in X, Y, or Z?”

    I actually spoke with Hal Varian at Google today about this research. He said they hadn’t looked at the phenomena but would.

  5. Hi George,

    Interest piece of research. I find the psychology of searcher behaviour fascinating, so it’s great to see some data on their habits.

    I agree that behaviour can vary massively between searchers, and that the genericness of a user’s phrase does not necessarily imply lack of purchase intent. Someone may really want to buy a camera today, I mean REALLY want to, but may search for something generic like ‘cameras’ due to lack of search engine experience.

    You could argue that tehe technically-minded and internet-savvy people out there tend to understand search engines better, so tend to be more likely to make more long-tail and highly-specific searches. This in itself could be a form of pre-qualification for potential customers of technical products.

    If you were an SLR camera retailer, for example, your target audience be likely be these technical, savvy searchers, who use well though-out and calculated search phrases. Ensuring your keywords and ads cater for these technical searchers, and less so for generic searchers, could be a profitable strategy.

    A bingo website, however, may not be so lucky (a previous bingo client I worked on had over 80% of paid search traffic from the term ‘bingo’ alone).

    It would also be interesting to see if any differences in searcher habits exist between males and females, as I imagine that could have implications for retailers of gender-specific products.

    As Chris pointed out in his comment, Google suggest is likely to have an affect on searcher habits (as shown in the Latitude article), but you’re right that it can only take you so far.

    I guess it leaves the humble search marketer in a difficult position: how do you efficiently target and cater for less-informed searchers (for search of a better stereotype), while at the same time filtering out the time-wasters who search for the same keywords?

    Doing so will surely be incredibly powerful in helping to increase ROI from paid search, but until advertisers can target searchers based on their self-selection settings (such as Google’s recently added ‘more shopping sites’ option), I don’t think there’s a short-term solution.

    Cheers,
    Alan

  6. Jim Jansen says:

    George,

    Great post with good empirical data, as usual.

    Concerning your comparison of you and your wife, it may be a gender thing. Women and men (on average but with high variance) search differently. Depending on the underlying intent, each search strategy may/may not be the most effective. In a non-fact finding search, your wife’s strategy may be better.

    I would think that this issue of intent would underlying interpret the percentages that you present. The search terms are expressions of some need/desire.

    Best,
    Jim

  7. Thanks for your comments, Alan.

    Jim, I agree, particularly for apparel, just heading directly to the stores you know you like is a perfectly rational way to shop, likely more efficient than hoping the search engines will guess the styles you like. It also explains why navigational searches have significantly higher conversion rates — but not tremendously high nevertheless. The fact that they’re a previous customer of yours doesn’t mean you get 100% share of wallet. They’re your competitor’s customers, too.

Trackbacks
Check out what others are saying...
  1. [...] This post was mentioned on Twitter by Alan Rimm-Kaufman, Metatags, Negative Keywords, Long Tail, Meta Keywords and others. Meta Keywords said: Search Behavior from the RKG Labs: My monthly paid search column for SEL in case you missed it: A couple of months… http://bit.ly/9sTxRc [...]