Making the most of your AdWords budget using negative keywords.

During an undergraduate economics class, I learned of a story about putting armor on WW2 planes returning home from war.  Many planes were getting shot down, and a man named Abraham Wald attempted to tackle the problem.  Wald's first inclination was simple; inspect the planes returning home, and document the location of the bullet holes.  His diagram looked like the one to the right.

From this point, the course of action may seem obvious, but may not actually be that intuitive.  Many people would probably suggest that you should put armor in the places where the planes were shot.  However, these bullet-ridden planes were actually the successful ones!  Remember, the pilots made it back alive.

Wald's theory was to put armor where there were no bullet holes.  Why?  Because the locations on his diagram showing no damage, corresponded to the planes that didn't make it back.  So...if a plane could return home after being shot in the wing, why reinforce the wing?  However, you may notice there are no bullet holes in the cockpit.  That's because there were no surviving examples of planes that had been shot in the cockpit.  Best solution; reinforce the cockpit!

This same approach can be applied to nearly any AdWords campaign, though I wonder how many actually try.  I've typically found that when someone is running a campaign with any success whatsoever, they usually want to grow that campaign, and rightfully so.  Sometimes this happens through adding new keywords, duplicating the campaign for a similar product or landing page, raising the bids on existing keywords, or raising the entire campaign budget.  While these methods are likely to increase overall sales or conversion volume, they will probably have negligible or negative effects on customer acquisition costs.  So rather than looking to expand positive trends first, it's best to eliminate negative trends, then expand. After all, why would anyone want to expand inefficient spending?  We'll examine a real world AdWords client and their data to understand how how this might be accomplished in AdWords.  

First a disclaimer, and a primer.  Disclaimer: If you do not have sufficient data (I'd say at least three digits worth of clicks for a particular campaign), you are taking a huge risk making future predictions from that minimal data.  If your ad has only been clicked 20 times, but hasn't converted yet, you don't have enough data to suggest that ad/keyword/combo is a loser.  If we have 1000 clicks on an ad and it hasn't converted, it is a loser.  Primer: The data we'll be using is a client who provides e-commerce software as a hosted service.  It is a technical market that is highly competitive, and keyword bids/clicks are very expensive .

Keywords like "shopping cart" are used, but obvious negatives like "metal" and  "wheels" are used to distinguish between between people searching for shopping cart software, and people searching for actual shopping carts.  However, after time, creative searchers will reward you with false positives and many more negative keywords to exclude.  So let's take a look

The first thing you'll need to do is go to Keywords > Details > All.  This will show you what your customers are ACTUALLY typing into Google!

Now that we can see what our customers typed when they found us, let's setup a filter to see who wasn't interested once they got there..  Filter > Create Filter > Conversions = 0.  From here you can sort your data by cost, or clicks, or whatever you're interested in improving. Just make sure you have enough data!  Zero conversion doesn't always mean zero effectiveness.  Since my client advertises against their competitor's brand names, we've decided to include an additional filter to show only queries that contained the word "vs."  OK, so now we have a list of terms that had clicks, but were unsuccessful.  While our total cost per lead and conversion rate was acceptable, the cost (over the period of a month) for these loser "vs." terms was $487!

Since we want customers to who are comparing our client to their competitors, we might hope to see queries like "client vs. competitor."  But remember, this is the loser batch, and thankfully, my client's name isn't anywhere to be seen.  All of these negative trends are mainly "competitor vs. competitor"  So what does this all mean to the client?  In short, if a potential customer is already comparing two of my client's competitors, the client's product doesn't stand a chance - It's just too late in the sales cycle.  Plus, we were paying to bid against both competitor names, possibly driving up the bid for these "vs." queries.

So...what should I do?  Add "vs." as a negative keyword?  No.  This would eliminate all searches for "client vs. competitor."  These are people who are already considering the software, so I need to make sure we reach them - maybe even increasing my bid on "client vs. competitor"  The real solution is to exclude every competitor brand against every other competitor brand.

A simpler example was done using a filter and setting the search term to "how" or "what" and turned up multiple queries like "how to customize url on 3dcart" or "how to install zen cart."  What we thought were potential customers researching software, were actually clueless competitor users clicking on my client's ads in an attempt to use them for technical support!  In the end, we excluded all who, what, when, where, and why queries.  After all, if we're looking for tech savvy users who can understand the value of the software, we need to be skeptical of those to talk to search engines like they're humans.

So there you have it.  By simply running a (conversions=0) filter on your ACTUAL Google queries, you can see what all the losers have in common.  These are negative trends, and should be eliminated in a manner that fits your goals.  I should also mention that Google has a conversion optimizer that SHOULD accomplish the same effect.  However, I doubt that Google has the financial incentive to, or understands your company as well as you.  Your end result may even be fewer clicks at a lower cost per click.  So asking Google to do it for you, is like asking a car salesman to find you the best deal.

Next time I write about AdWords, I'll discuss Gremlin rule #3, don't feed them after midnight.