I’ve read other threads on webmasterworld about this topic, but I did not find the answers I was looking for. I’d greatly appreciate any insight and reviews you might have about SEM agencies.
Here’s what I’m looking for:
(1) A profit maximizing algorithm (or at least something smarter than account level ROI targeting). ROI and profit maximizing is not the same thing. For example, I could minimum bid and watch traffic trickle in while I bank $30 of profit at 900% ROI. I’d rather bid higher and increase my profit at the expense of a lower ROI. But at what ROI point does each keyword capture the most profit for me? If I set bids to an ROI target of say 50% across my account I will certainly under perform. For some of my low margin adgroups, I will underbid and get no traffic. For my high margin adgroups I will overbid and waste spend. I need a system that tests adgroups and keywords at different levels to find where they perform best.
Profit max = marginal revenue - marginal cost.
[en.wikipedia.org...]
Also see the flash video at MediaBoost.com for an excellent description of profit maximizing.
(2) Integration with at least Adwords, YSM, & Adcenter. I advertise across the major platforms. Also, the system must be able to bid discriminate between content and search match.
(3) Weight bidding based on recent data. My revenue for conversions changes cyclically. The bidding system must account for what I tell it a conversion is worth to me now, not it’s historic worth.
--- MY RESEARCH ---
Here’s what I’ve found through my research. MediaBoost seems to be the only company that overtly claims a profit maximizing bidding algorithm. Unfortunately, they only work for Adwords. If anyone has a review of MediaBoost I’d be interested to hear how they perform (Booster, I’ve seen you posting out there and I invite you to chime in!).
Efficient Frontier uses a portfolio-based bidding algorithm. Their system is premised around the idea of maximizing for X (Revenue, Signups, Orders) given a constraint Y(budget). Efrontier then manages your budget across your portfolio of keywords. I’m not interested in maximizing revenue, signups, or orders and I don’t have a set budget that I need to manage.
I have also talked to the sales people at DART. Their big claim is that because the search engines share actual bid info with them, they are able to make better decisions based on real data, not estimates. I can see real data creating a competitive advantage, but DART only applied simple rules like “bid to ROI” or “bid to CPA”. Also, I did not care for their interface or syndication process to different ad platforms.
Now I’m starting to research firms like Omniture, ChannelAdvisor, Inceptor, and Bidhero. Anyone tried these agencies? Any other suggestions or links to trusted reviews? I know I might not find my wish list I outlined above, but I can’t see shelling out big bucks to have these agencies “bid to ROI.” I can do that.
I know this was a long post.
Thanks!
The bid management tools also need to collect historical data about how various keywords have performed over time in different positions, different max bids in order to accurately forecast what the likely consequences of any bid changes are likely to be.
I've found that the longer the bid management tool is running & collecting data, the better the results will be over time. Whilst results can usually be improved to a certain degree fairly quickly, the best results will take time.
It makes sense to have at least 7 days of data so to iron out variations due to the day of the week. Also you have to consider how data is distributed across your keyword set. As Zuckerman rightly says, the more data the better. That's why at Efficient Frontier (where I work) we take in as much historical data as possible from an advertiser before we launch.
I’m okay with making decisions based on limited data -- it’s just reality. This low confidence interval problem is likely something that isn’t as big a deal in many other industries. If you have a specific service or widget you’re selling, you are focused on that and you buy clicks and build up data. I deal in much shorter time frames and don’t particularly care if an adgroup fails, I’ll just move on. These agencies seem to want to cover their behind by not giving the option to take over bidding responsibilities until they have high confidence intervals.
I guess from an agency perspective, they can't commit to improving the results when data is difficult to gather - I think all SEM comes down to is statistical data & probability (as I write I'm working through clicks/conversion/ROAS figures in excel!). On the agency side, it's difficult to share the responsibility of bid management with client(s) - when things go well, great, when things go bad, it's always the agency that's to blame, which is not necessarily always the case.
I was very impressed with Marin, both by the people working there and by the software I saw. Our sales contact, Andrew, seemed to know his product and the industry well. He did a good job of listening to and addressing our needs.
Marin seems to “get it” from the perspective of search marketers. Andrew stated that people who work at Marin are encouraged to run their own search marketing campaigns (affiliate based, I’m guessing) on the side. It shows in the software design. Many of the tools we saw seem to solve practical problems that we often encounter.
With Marin you can assign bidding strategies to separate folders. I’d need to use the software to better understand this folder system, but it sounds similar to SearchForce’s sub-portfolio system that I described earlier. Each folder requires a minimum sample size of 100 visitors and 5 conversions before the assigned bidding rule kicks in (nice, we’re talking about relevant statistics info). I asked if we could alter the sample size requirements for our system, and I was told that they would get back to us. Our rep seemed concerned about whether lowering the sample size would be a good thing for us.
Have Andrew call us back tomorrow and we’ll talk.
Another negative is that Marin only has integration with the big three, but they are a newcomer with plans to expand. I see the fact that they are not giants as a positive because they seem more flexible and don’t have so much corporate bureaucratic garbage. Also, the only reason I didn’t recommend that CernyM check out Marin Software is because they start their pricing at a percentage of $50K in spend a month.
Tools tend to be 2-8% of spend, with an industry average I'd put at 5-6%, which SEM agencies who have their own proprietary tools tend to charge 7-12% of spend, with an industry average, IMO, of 8-10%.
Per-click pricing is usually only asked for by traditional agencies whose billing model with clients requires it, and is, IMHO, a horrible way to pay for SEM tools. That said, however, firms that *are* willing to charge on a per-click basis tend to charge 2-4%.
Hope that helps.
a)it brings with it the built-in assumption that keyword management is a commodity like banner ad serving, when in reality it's much more than that (or should be).
b)it assumes that the way agencies' clients have become used to paying for ad technology is the way they should pay for it going forward, which is bad for everyone. Bad for the advertiser because they mistakenly work under the assumption that keyword mgmt is like stock trading fees, when in reality keyword management should be looked at akin to sound investment *advice and mgmt*.
a) I think there is a big differnce between commodity pricing (CPM) where you are just spraying the world (albeit with more targeted and stickier messages) and hoping some people click, and some of it sticks, as opposed to the search pricing (CPC) where a tool/service is paid more for providing more, and only paid for what it provides.
This most espesially makes more sense to me then a % of media spend which has no corelation to the work put in or the results that come out. As if a click on a 20$ key word took more processing power from a system to trafic, optomize, or report on then a click on a .20 word.
B. I believe that the idea that a Search marketing tool should be "advice and mgmt" falls for me under the assumption that BID MANAGEMENT is SEARCH MARKETING, wholy and completly.
When evaluating at a tool it there are 4 key aspects I take into cosideration.(1) Of course 1 is bid management if the tool is not going to get the job done, I suppose it doesn't matter how it gets it done. (2) Secondly taken into acount are Work flow effeciency, this comes into ad trafficing, campaign management, bid strategy structuring/restructuring and reporting integration. (3) Then you have to ask about acuracy, what is the Data is that your optomizing off of and reporting on, how is it attained, are there built in discrepency tolerances to your tool (do you optomize on 80% acurate data, and therefore only optomize at 80% at best). (4) Lastly, what sort of insight does a tool provide? Can you look at a granular level, can you look at it vs the big picture, across campaigns, across engines?
Another over arching question to this, is what sort of resources are being put into your tool to keep up with the engines. Quality scores, features, and even API integration points, change constanly (near weekly) not only on one major engine but across all 3. Can this be done to a piece of software on your computer? How often are you getting releases, how much lag time is there? Even using a digital agency how many $$$s and engineers do they have working on their in-house tool?
I think its intresting that you list Banner Ad serving as a commodity, and it deffinatly outlines one frame of thought. In this frame of thought it makes sense that a Search Marketing Management tool is just providing "keyword management" or "Bid Management". However if you look at a tool as more then the cost of doing business, but instead look at it as something that affects not only the bottom line with its cost, but the top line with its contibuting revenue margins, then the language changes from total cost, to total value then I think you can have a real discussion.
Sorry if that was a bit ranty... its early.
[edited by: Sylvanrage at 1:56 pm (utc) on Feb. 20, 2008]
In an odd twist of fate, I ended up joining Omniture as VP Search Sales about 4 weeks ago.
To answer your question, I think the person you spoke with at Omniture simply didn't know enough about search to answer the question correctly. We're building a bigger pre-sales team & one with more SEM industry knowledge, but I would've absolutely answered your question by saying that it's a function of the volume of data & not time. Specifically, if you're going to use the system to optimize to a transactional business goal, you'd need 2-3 transactions for a given keyword in order for the system to be able to optimize.
Another answer, though, is to say that defining and optimizing (initially) to leading indicator metrics can let you have more data with to optimize. For example, if you know that for every person who consumates a transaction (of profit $X), 10 people take a precursor action (say, filling a shopping cart), then you can assign a value to filling a shopping cart of 1/10 * profit $X. That gives you more data to work with for more keywords.
Happy to talk about SEM optimization strategies in general and Omniture SearchCenter specfically as time goes on.
-Shorebreak
I have two questions..
1)Pardon my paranoia, but are the tools mentioned in this thread stand alone software products or do the vendors end up getting access to your business information (keywords, bid levels, websites, etc).. I am very concerned about privacy in this regard. Especially when one of the companies apparently encourages its employees to run their own campains as a side business.
2.) What is a realistic expectation for performance improvement? Some of the sites say 100%+ others have testimonials for 40%.. Is that sales fluff or is that kind of improvement really possible?
2. performance improvement will vary - best to try a vendor with a free trial that will also enable full bidding so you can see for yourself; these tools are not magic so if you are already bidding to within a decimal of perfect the tools will only do the same but much faster.
Hope this helps.
