Pay-per-Click Observer

I know, everyone says that. I mean it a little differently.

One of my clients is switching from one analytics package to another. The process of doing so requires that we retag every one of the many thousands of keywords that we maintain for them. This would be impossible by hand, of course, and so we utilize technology– usually Ruby in our case– to do the grunt work.

Unfortunately, we’re a little fringe in our application, and so there’s really no such thing as a support process for people like us. You do searches, call this person, they talk to that guy, and he emails this other woman. Hopefully, we patch up the thing to work enough to accomplish our goal. It’s not going to get shrink-wrapped. It just needs to work.

So, today, I found myself banging my head against the Google AdWords API. It’s great. I love it. It doesn’t always work the way you think it will. Case in point, updating keywords with a new destinationUrl.

Turns out that if I included an ampersand in that URL, my code would, properly, encode it into ‘& amp;’. Then, when I called updateCriteria, which updates the keywords, with that my keywords would all get updated with that ‘& amp;’ in the URL– not reconstituted into a ‘&’. Google said “you don’t have to encode that” when I asked about it.

OK, so I patch soap4r to just ignore that entity. It submits the request with the ‘&’ in its original form. Error– you have to encode those. Actually, that’s the right answer in this case.

Anyway, I decide to explain this situation to the client. Empathetically, she listens and asks me the simple question “Those are so common; don’t you use those in other campaigns?”

D’oh! I do. How come they work, and this doesn’t? It’s because in those cases, I tend to drop everything and replace it to avoid synchronization problems. In this case, I was trying to “update” the keywords in place. Sure enough, replacing them works; updating them doesn’t.

So, I learned two things: one is some geeky tidbit about the AdWords API and the other is a valuable lesson about listening to my clients. Instead of pretending that everything was fine, I explained exactly the problem we were facing. My client used her fresh, but informed, perspective and broke the dam.

The older I get, the more I learn that, if you really want to do a good job, you need to able to admit that you don’t know everything. I’m not saying you need to admit that you don’t know anything. There’s a big difference.

PPC bid management was much easier when I started out.

First, I didn’t know what I was doing. That always helps to make things seem easier.

Second, you could actually see your position based on your bid. That’s a pretty dumb thing to fixate on, but if you review the first thing, you’ll see why I thought it wasn’t. You’d just announce “I want to be #2 for these words!” and you’d know just what to bid. I wrote all kinds of code to go to the Overture (now Yahoo!) site (the only game in town for a while), pull all our bids, and adjust them $.01 over our competition for our chosen position. I was so smart (see previous paragraph.)

These days, I’ve been at it long enough to know that generally applied bidding “strategies” are usually counter-productive. Bid management must arise from an understanding of the goals of the particular campaign and a regular looping back with analytics to continually tweak– first for large improvements and later for maintenance and small improvements. Sometimes, conversions are key regardless of bid. Sometimes, the cheapest click is the best click. Usually, it’s something in the middle.

Foolhardy experiments should be done periodically to challenge current assumptions and try out new things. If this ongoing tweaking stops, spend gets out of whack, CPC creeps up, opportunities are missed, and the campaign starts to, well, rot.

What’s more, the Google and Yahoo! have both moved to a system wherein you can’t tell ahead of time what your position will be. Your position is determined by something called a “quality score” which reflects your bid and your click-through-rate. However, it’s opaque to you, the advertiser. It could involve the phase of the moon or some internal strategy to “mix-up” results. We’ll probably never know exactly.

So, that means, you just can’t know what your bid has to be to hold a given position– though Google will allow you to opt in to a position preference, but you then don’t know what your bid will do. That just sounds like such a bad idea to me in about three different ways.

So now we know that nice, simple bidding strategies don’t really work anymore (or ever). In part 2, I’m going to talk about why you can’t necessarily just increase your bids to get more clicks and why there’s no such thing as the “optimal” bid. Prepare to have your mind blown. Groovy, baby. Yeah.

Synchronicity…

I just got off a call with a company looking for some help in determining the appropriateness of PPC for their purposes. They want me to help them design and manage an experiment to see if we can revisit the subject after an in-house experiment left them with a really nasty taste in their mouths.

I made note of a couple metric-related issues that were core to our conversation that I wanted to mention given my previous article.

Specifically, the client is currently using affiliate marketing. Interesting thing about affiliate marketing: it’s perfect when the cost-per-acquisition is determined by something that doesn’t cost anything– a trial, let’s say. Affiliate marketing aligns the interests of that particular seedy underworld into a well-oiled machine for generating clicks and sign-ups. The only problem is that all those people who were inspired to get you to pay for a “lead” are usually less inspired to give you any “revenue.”

Suddenly, the $100 that a paying customer should generate for you becomes $2.50 when averaged out over the population of leads. That, in turn, drives the reasonable CPC down, which drives you to lower and lower quality traffic. Ah, metrics.

Another complicating factor in this case is that there’s a download associated with this client’s service. Viewers see the PPC ad, click on it, and bail before they ever get close to opening their wallets.

Well, what if we told them that the download was required? What might happen? The prospect would see the ad, realize that they’re on a dial-up, and just skip the whole thing. Click-through-rates plummet! Oh no!

On the other hand, maybe what we’re doing is setting expectations appropriately. Maybe we’re “pre-qualifying” our prospects. I used to take every call for potential work. Every email got a call back. Now, I probably return 1/10. Does that make me unresponsive? I don’t think so. I think it means I waste less of my time and less of the prospect’s time.

I don’t do $200 engagements. I don’t build web pages. I don’t market ringtones.

The best thing I can do is ignore those requests. If I have time and they seem genuine, I’ll let the non-prospect know, but if I even did that every time, I’d be sending “no thanks” emails every night instead of playing with my kids.

The same thing applies in PPC. Let the user know what they’re going to get. Don’t try to trick them into visiting you. If you know that there’s something about your offering that is getting in the way of the flow, point it out or change it.

Ebay used to seem to buy ads for every term under the sun. Even I once clicked on an ad that told me that I could “Find Pope John Paul II” there.

I checked. He wasn’t there.