I would like to start a discussion on best practices regarding account optimization.
The way we do it most of the time is to make multiple ads per adgroup and use all match types for keywords. Over time we evaluate both the performance of the ads as well as the individual keywords. We typically throw keywords out or add more negative words, in case of broad matches, if they do not bring any conversions after 200-400 clicks (this of course depends on the product margins)
For ads we evaluate (after at least 400 clicks) their effectivity; looking at the combination of CTR and conversions. I typically have 3 ads up per adgroup and once in a while kick the worst ad out and replace it with a new ad. For clients with advanced tracking (shops) we compare the revenue per click to determine ad effectivity.
Like to know how you guys and girls go about this!
Tracking both the keyword ROI/CTR and the ad ROI/CTR can lead to interesting insites. As said recently around here, the highest CTR ad isn't always the most profitable. However, sometimes keeping a high CTR ad around to lower the costs of the KW can also be useful with ad optimizing turned off.
There are others that we do it on a timeframe - especially for accounts with low conversions or low search volume(such as selling expensive services, certain products, etc.).
Often this is from a week to a month, depending on available spend. The reason to do it weekly is that there is a tight budget consideration, and AdWords needs to prove itself. In these cases, there are times we delay the launch of AdWords until the prime week of a month depending on the product.
(i.e. Often 'accounting software' sells best in the last week of a month as new companies are trying to figure out their sales and how to track it. While luxury items often sell best in the first week of a month due to people having a new paycheck and rent not due until the end of the month).
Monthly is kinda the same as the above, but we're able to track sales throughout the month to determine if there are sales patterns. There are certain products where the bids aren't adjusted by competition, but by day/week of the month due to past preforming sales patterns.
While I like to track every keyword individually for ROI, sometimes the niche keywords just don't get enough impressions for this to happen. In these cases, I'll combine tightly related keywords (usually these are already seperated in their own AdGroup), and consider them one keyword for optimization/reporting purposes.
For ecommerce sites, I like to run a few 'sales campaigns' every once in a while to jumpstart CTR, or to test new some new ideas.
As for me, generally, we look at CTR first and then conversion rate second. I figure that if we get the clickthroughs, then the product will sell itself or the web design where it displays.
I am the webmaster of a specific store selling a large selection of one particular product. Our Adwords campaign just started.
When it comes to targeting, we have one ad campaign for our household market and one for our commercial market. Within those campaigns we have multiple ad groups for categories and specific brands. Since our field is brand driven, most of my ad groups are brand specific with less ad groups for specific categories.
We have only been doing this for less than a week. So, I have no good strategies by way of reporting, but I am looking at the CTR of individual keywords. If the CTR is low, the keyword gets thrown out. If the CTR is HIGH and the conversion rate is low, then I have to do something to increase sales on the web page itself.
This is your thread - and quick question.
Do you want to limit this thread to all the optimization types - or do you also want posts included that discuss pros/cons of how others optimize campaigns?
Any optimization related discussion is in place here i would say. Question to you is how to handle time related optimizations with limited volume?
I'm no statistical wizard but do know that a certain amount of clicks and conversions are required to make a good estimate on future conversions. I always find it hard to optimize on little data like that (and often get more broadmatched keywords to get volume up, for most of my customers budget is not they major concern)
Question to you is how to handle time related optimizations with limited volume?
Almost all are on a case by case basis after discussion with the company in question as to what they can afford, what their goals are, what their conversion rate is for non-ppc products, and other info.
Unfortunately, I can't name a magic formula for this time frame running and volume. What usually happens is after the trial period, I look at the volume, conversions, CTR, etc - and then do an analysis of the account.
Sometimes, the analysis will include the fact that over the trial period, the account did not get enough hits to justify considering these numbers significent, and recomend extending the trial.
As all products/sites are different, almost all the stratedgies are slightly different as well.
Rather than try and insert a human in the middle of the billions of calculations that must be done hourly across the 1000-750,000 keyword campaigns we manage, we use a portfolio trading algorithm that we've adapted from Wall Street to optimize keyword bidding automatically. Its advantage over the rules-based approach others are using is that we can make decisions on individual keywords not just in their own context, but in the context of the *overall* keyword portfolio. Additionally using a portfolio algorithm approach lets us manage to the client's business goal, not an arbitrary position or CPA-based goal; ultimately clients have profit & volume goals that cannot be met by optimizing to a simple CPA goal.
Obviously, this only works where there's a measurable transaction of some sort, and where the client's business goal is clearly defined, relatively stable, and based on those measurable goals (revenue, CPA, margin, daily net profit, etc). Likewise, it only works where there's enough volume (of impressions, clicks and transactions) to build a statistically valid model. The way we get around low-volume keyword data is to cluster data from multiple keywords that share some similar characteristics.
Don't know who owns it but it is nifty to get a clear decision on a consistant basis. Wouldn't wait for hundreds of click though.
[edited by: eWhisper at 1:59 pm (utc) on Jan. 28, 2005]
[edit reason] Please Don't Drop URLs [/edit]
we use a portfolio trading algorithm.... make decisions ... in the context of the *overall* keyword portfolio.
Based on something such as CAPM?
I'd be very interested to hear more about this.
The way we get around low-volume keyword data is to cluster data from multiple keywords that share some similar characteristics.
The problem with this is that very similar keywords (e.g. plurals) can have major differences in conversions.
I like the approach of Shorebreak although non of my clients qualify for this kind of heavy approach. I'm thinking about automating some of the optimizing process using excel. I'm making pivot tables from time to time to look at the correlation between ad copy and keyword effectiveness.
Did anyone do any work in this area?
If a campaign is going to go bad, it will do so very early on. When we put new campaigns we tend to place them in "intensive care" and refine the campaign based on our real time data, and not waiting for Google's data, as normally by the time the system updates, the damage if there is any, will be done.
As for tricks :
1. Dynamic titles
2. Phrase and exact match only, no broad, very few negatives
3. Break your keywords out into smaller ad groups, rather than having a catch-all ad group
4. Track per keyword
5. Use individual pricing per keyword rather than a sheepdip pricing model
6. Be very aware of who you are bidding against and spend some time understanding their strategy/tactics
7. Disable content matching but keep syndication
8. Disable automatic ad optimization in the first instance
There's plenty more things we do, and I'm not saying that all of the above will either be best practise or will work for you, but they have certainly worked for us in every instance.
Also many Adsense publisher sites are of dubioius quality.
