Sep 212010

Users Don't Always Buy What They Seek

We find that, more often than one might think, people initially searching for one manufacturer brand by name end up actually buying a different brand from the advertiser's site.

In paid search it's important to measure the value of traffic as precisely as possible. This means not just granular bidding -- to the ad level, when statistically possible -- it means using the best measure of value on a conversion event. In retail, margin data provides the most accurate view of an order's value, but assuming that the margin rates implied by the keyword are the margin rates on the order isn't always a great idea.

Not only do people change their mind based on selections, promotions, etc; they also tack on items from other manufacturers, potentially changing the margin %.

We did a down-and-dirty analysis of this phenomena for 5 retail clients in different product categories.


We grabbed a random sample of twenty orders from each advertiser following a user search that expressed a specific manufacturer brand preference.

We looked at all the items on those orders to determine:

  • The Keyword Match %: How often does the order include any items from the manufacturer searched?
  • % of Items on Matched Orders: What fraction of the items on the orders that do match are from the named manufacturer?
  • % of Sales $ Matching: All in, what fraction of the total sales dollars come from the manufacturer brand named in the search phrase?


Pretty clearly, this can be a meaningful effect for a number of advertisers, and it's also clear that the variance from advertiser to advertiser is significant.

In some cases, this phenomenon is less a function of add-ons and user's changing their minds than the nature of the category. Someone searching for "Ford engine parts" is very likely to buy a belt, filter or hose made by someone other than Ford.

Nevertheless, those who assume average order values or margin percentages based on the keyword may need to validate that assumption.


Whenever possible, the actual margin on the order is the best way to go. We find a back feed of order-level margin is ideal. In addition to getting the margin right at a granular level, it allows for frauds and cancels to be knocked-out and discounts to be incorporated.

Obviously, for those who have very consistent margins, return rates, and lifetime values across all categories and manufacturers, this isn't a big deal. But, for most folks, it's well worth looking under the hood.


7 Responses to "Users Don't Always Buy What They Seek"
I've been struggling with a similar question. My dad offers a dental seminars combined to travel, and we get people looking for 'product a in location x' as a phrase match for 'product a'. Yet my dad doesn't offer location x. And it's clear from the lander that it's not in location x. I initially added 'location x' as a negative keyword, but we've seen conversions on it, and now I'm doubting myself again. On the one hand, I fear lead quality will drop. On the other it's plausible people aren't 100% set on location x. Thoughts?
Hi Gab, Yes, I think you may be right. Much may depend on the proximity of X to a valid location -- folks may be willing to travel a bit -- but it may be worth doing a quick assessment to see whether any of those leads end up converting. People are notoriously bad at reading details, particularly when signing up for info is free. Could be these are "false wins."
Kenny says:
I was just trying to clarify this phenomenon to a client the other day, and in their case it happens even across categories. Just because your sales on x category of product are low, getting more aggressive on keyword terms related to category x isn't necessarily going to solve it automatically. There's an assumption that there's a nice, clean 1-to-1 relationship there, but it's far from it. Many people come in on a keyword related to category x (high ticket item), but you can look at the average sale and tell that often times consumers from category x keywords are buying things from category y (much lower ticket item). Way, way more complex of situation than saying "Green widget sales are down...boost the green widget terms!" Not to mention what that could do in terms of hosing A/S ratios.
Great point, Kenny. I'm actually planning a follow up post on the cross-category phenomena. One interesting caveat is worth mentioning: be careful with cookie windows, here. We find that folks with LONG cookie windows will see a much greater mis-match between the click and the order when the time lag between the two is long. Raises questions about whether those orders are really related to the paid search click. We advise clients to shorten the window when we see a great disparity in the mismatch percentages between the first few days or hours and the orders that come X days after the click.
Kenny says:
Good point, George. In this case, almost every order is achieved on the initial site visit.
George, Great things to know. Amazing to see how much order plays a part in search. Makes sense to me.


Check out what others are saying...
[...] Users Don’t Always Buy What They Seek, Rimm Kaufman [...]

Leave A Comment