Analysis of Data from 1 Million Visits

[TimmyMagic]

Our AdWords campaigns need some work. They used to be profitable but now it's hard to break even. I know this is in part due to neglect and taking my eye off the ball. I want to re-focus and build our campaigns again from scratch, so I'm hoping for a bit of advice from you fine fellows.

I have used Google Analytics for tracking and have AdWords campaign data from August 2006, with just over a million visits. It seems silly having all this data and not really doing much with it.

There is so much data but it's hard to find a way of getting the information I need into a report that includes revenue for each keyword and the costs. For starters I'd like to find a way of building a good list of negative keywords. But I'm sure there's much more that can be done.

What would you do with such data to help rebuild your campaigns?

[buckworks]

good list of negative keywords

That would be a good place to start.

Working to strengthen your negative keyword lists would be some of the most profitable time you can spend, especially in the early stages, and the research would stand you in good stead even if you decided to start over with some of your campaigns.

TIP: Sometimes you'll want negative keywords that apply to specific ad groups or campaigns, but some keywords will be unwanted in multiple campaigns, or even your whole account. It can be a time saver to make shared library lists for terms that you'd want to apply to more than one campaign.

You will often want different negative keywords for search ads versus the display network, so make sure you have separate campaigns for each, with their own negative keyword lists.

[TimmyMagic]

Thanks Buck.

I already have a list of active campaign level negatives. I'm certain there are many more to be found... it's just a case of finding them. I've set up a custom report in GA for match type queries and it's throwing up some interesting stuff. It's going to take some time going through them all.

I paused our entire display network campaigns yesterday. These have always been unprofitable. There are some adgroups that look okay and can probably be continued, but I was alarmed at the overall costs and pathetic ROI... so paused the whole thing.

[lgn1]

We turned our adwords campaign around (in a very competitive field), but it involved planning, organization, and using sound statistical methods to make decisions, not gut feeling.

I would start with one of those for dummies books (yes they have one for adwords) for organizational and marketing skills, and then either take an introductory statistics course at a local univeristy, or hire someone to do the math, if this is not your cup of tea.

[buckworks]

display network campaigns ... always been unprofitable

Something I've found that helps on the display network is to block sites that have anything to do with games.

There are a gazillion sites out there with cheesy cartoon games that trigger our ads because they happen to mention the right words but they're not the right audience at all IMHO.

Worst of all, they don't get reliably blocked when we add "Games" to our list of undesired topic areas. I have to hunt them down one by one. It's an ongoing battle.

[TimmyMagic]

We turned our adwords campaign around (in a very competitive field), but it involved planning, organization, and using sound statistical methods to make decisions, not gut feeling.

That's the thing... I don't intend to use gut feeling. I want to use the important data we have in GA to help turn things around, not guesswork. It's just a bit overwhelming.

I would start with one of those for dummies books (yes they have one for adwords) for organizational and marketing skills, and then either take an introductory statistics course at a local univeristy, or hire someone to do the math, if this is not your cup of tea.

I bought an eBook about AdWords quite a few years ago. It was great. That was probably back in 2006 / 2007, and while many of the principles remain, AdWords has moved on a bit since then. I'm sure I could do with a refresher and I've spent a lot of time these past couple of days reading up on things.

As for hiring someone... I've done that, although not for just statisical analysis, but to run the entire campaigns. In July 2010 I outsourced the management of our AdWords campaigns and I took a back seat because I was finding it hard to devote time to it. The person was well qualified (or so it seemed) and I was hoping he could push things on a bit. We parted ways about 6 months ago because I was unhappy with the performance of the campaigns.

I just did a comparision. I looked at some data from when he took over to the present day (I've not changed much since we parted), which covers 31 months. I then compared this to the same data range before July 2010 (so back to November 2008). Here are some stats:

Metric (before / after)
Clicks = 140k / 81k
CPC = £0.43 / £0.44
RPC = £0.74 / £0.62
ROI = 73.02% / 41.86%
Margin = 42.20% / 29.51%
PVV = £0.76 / £0.59

These figures exclude the display network (only started that in July 2010).

This got me thinking... why did I hire this guy? I remember thinking at the time that the campaigns were slipping, so I also looked at the 6 month period leading up to July 2010 to see if that was the case. The stats show it was almost on par (and in some cases better) than the full 31 month period prior to outsourcing.

I know this isn't totally accurate and doesn't account for other variables. There was less competition a few years ago, that's for sure.

I still have the data for these old campaigns. Maybe I should just rebuild based on those.

[instand1]

1) best way to find negative keywords, imho:
In AdWords under the index tab "Keyword-Details"
Select all keywords for a long period, say a year or so. Then remove all obvious junk and those keywords that have a low conversion rate.
2) Display network: you can select only those placements with a high conversion rate. Automatic placements can be switched off completely after an initial "trial and error" period.
(Can be done only seperately for each ad group.)
3) Keep search and diplay campaigns separate: Gives much better overview. Use AdWords Editor to duplicate and separate the existing campaigns, if they are mixed.

[RhinoFish]

instand1, good advice.

keep in mind that the logs only record kw search data for places you've gotten clicks - the worst stuff that you want to neg out, even over long haul time frame datasets, still might not show up.

so in addition to pulling the reports instand1 mentioned, find alternative ways to eliminate imps that aren't relevant as well.

[TimmyMagic]

1) best way to find negative keywords, imho:
In AdWords under the index tab "Keyword-Details"
Select all keywords for a long period, say a year or so. Then remove all obvious junk and those keywords that have a low conversion rate.

Thanks. I've been using the 'match search term in analytics, which has been very useful. But being able to see what search terms are displayed (rather than clicked) is even better. Excellent tip.