A certain member keeps going on about "portfolio theory" without stating methodology or details. And I know that more than one member has disagreed with this "theory" when it has been mentioned in the past.
So...
What is it? How exactly does it work? Why would any of us want to do it? How would it benefit the average PPC advertiser?
When it comes to keywords there are still probabilities involved but you have more control over whether your keywords convert than you do over how a stock moves.
We prefer to focus on the keyword level but if you had a large enough data set over time, a "portfolio management" strategy might prove to be better.
As in the off-line world there is still a black hole (though a smaller one) between online advertising and sales or business growth the way most companies track results.
For us the idea of portfolio strategy is to use blended results.
By that I mean that if you establish that you can make a profit at 17 cents a click, you'd figure that by bidding only 17 cents across the board would be the best way of handling that right? Wrong.
We work on the basis that some of our keywords will come in below that figure, so that gives us scope to go above that figure on keywords that we know to convert.
Blended we will be 17 cents solid.
In a finance portfolio you would have some investments that give high yield, others would be dividend based, some would be growth, some low risk, some no risk.
Bidding on keywords works in a similar way. As most people talk about buying cycles, it's important to know where in the cycle each keyword might fit, and bid appropriately.
Smart pricing has been introduced by Google to lower prices where the keyword is a browse rather than a buy.
As a portfolio manager, it's our job to get the best results based on the risk/reward profile of each client, we just don't use software to manage the process. I'm not saying that it doesn't exist, but there is still a lot of unknown factors where PPC is concerned and we feel you still need to humanise elements to make it work.
ASSUMPTIONS
1)Assume, for the sake of argument, that you have two keywords ('plasma TV' and 'flat screen TV') and that because you've been tracking cost and revenue diligently(both historically and near real-time), you're in a position to know what each of the two bid positions for each keyword will cost and produce.
2)Assume your business goal is maximum revenue at or below a $2 CPA:
Keyword #1 - 'plasma TV'
Position #1 CPA = $2.00; produces 15 transactions
Position #2 CPA = $1.00; produces 10 transactions
Keyword #2 - 'flat screen TV'
Position #1 CPA = $2.50; produces 20 transactions
Position #2 CPA = $2.00; produces 10 transactions
[obviously this is as simple a scenario as you can get, as is the goal for this portfolio of keywords; but it's to make a point, nothing more]
What you'll do if you're
a)using a rules-based bid management tool (like the ones taking out big booths at SES, PubCon and Ad:Tech)
b)using an internal spreadsheet that periodically marries cost and revenue data to kick out keyword-by-keyword bid changes
is take position #1 on the first keyword and position #2 on the second keyword, netting you 25 transactions at a $2 CPA. ***That is the opposite of the optimal choice for this portfolio of keywords.*** The optimal choice is position #2 on the first keyword and position #1 on the second keyword. Why? Because it more closely meets the business goal for the portfolio of keywords = *maximize* transactions at or below a $2 CPA.
Obviously a human could reach the same conclusion when the portfolio is 2 keywords, but what about when the portfolio is 1000 keywords, 10,000 keywords or 100,000 keywords? And what about when (as is the case more often that not) the competitive environment for the keyword portfolio is constantly changing, hour by hour, and when conversion rates, ASP and margin is changing over time for each keyword?
If you have 1000 keywords running on 3 SE's and there are, for example, 5 bid positions of meaningful volume, that means you have 3 x [5 to the 1000th power] potential portfolio choices, only one of which is optimal relative to your business goal. It's in situations like that that a portfolio-based bidding strategy is overwhelmingly advantageous.
WRT how to implement this, the only way I'm aware is to use a portfolio trading application of the same type used on Wall Street by the quantitative, or 'black box' hedge funds.
Some of the rules-based bid management vendors, and many agencies looking to snare a client, say they apply a portfolio approach, but what that usually means is
a)they think the client is dumb and won't know the difference
b)they periodically look at the top-performing keywords and try to average things out so they allocate profit from one keyword to on another less profitable one. In this case, though, the averaging is anything but data-driven, and certainly doesn't bother to look at the actual, immense number of possible scenarios that exist, uniquely, at that and every moment in time.
Granted, there are a couple types of businesses where this approach doesn't work:
a)sales where historical data has no value for predicting future performance, sporting event tickets, for example, or new movie-related keywords
b)situations where no historical, keyword-level data exists (and that describes 50%+ of the PPC market right now, especially the low end)
c)companies who haven't taken the time to clearly define business goals and values for the different types of actions users take on their site.
Interestingly enough, some of the largest PPC advertisers try to approximate this approach by employing banks of financial analysts to do all the calculations need to employ a portfolio approach. That's not ideal either, but it explains more than one ecommerce site headed towards IPO.
Keyword #1 - 'plasma TV'
Position #1 CPA = $2.00; produces 15 transactions
=> $1500-$30 = $1470 profit
Position #2 CPA = $1.00; produces 10 transactions
=> $1000-$10 = $990 profit
Keyword #2 - 'flat screen TV'
Position #1 CPA = $2.50; produces 20 transactions
=> $2000-$50 = $1950 profit
Position #2 CPA = $2.00; produces 10 transactions
=> $1000-$20 = $980 profit
Surely taking position 1 from kw1 & position 1 from kw2 is what any half-decent PPC manager would choose?
Using your portfolio theory, although your average CPA is lower, you get fewer sales & less profit.
And if you scale it up, with greater profit margins & thousands of keywords, you can see why I don't get it - it's just not as profitable!
I can see how a lower CPA can be beneficial, but surely the only metric that really matters is profit?
In the scenario I created, and within the business goal I set, I think it's evident that a portfolio approach does a better job of achieving the business goal = maximizing revenue within a $2 CPA constraint.
I could just as easily, though, build examples showing the advantages of a portfolio approach where the business goal is:
1)Maximizing gross margin within a max cost/order constraint
2)Maximizing gross margin within a fixed budget
3)Maximizing profit
If you'd like to see those examples it'd be much easier if I just stickied you a couple slides.
-Shorebreak
Assuming that you only ever need to manage 20% of your keywords, what would be the minimum number of keywords required before someone could consider a portfolio methodology?
And how would it work with AdWords (where position is not determined solely by CPC)?
We've been looking at external factors such as weather, sporting events being on TV, seasons, and a whole host of other reasons that affect our algo.
We've found that if we emulate the job of a trader, then our process becomes a lot more simple to follow.
This in turn allows us to look at the bigger picture while still determining short terms opportunities for gain. There is even occurances when we have bought keywords that we know the we are going to make a losss on, because we want the user to make us money on the third visit etc.
The real beauty here is that our most profitable work has come from gut feeling rather than anything the computer spits at us! And our portfolio has certain elements that just keep on performing, and haven't needed to be monitored since day one.
Is this what you meant by a Portfolio, or have I got it wrong again! :-)
Assuming that you only ever need to manage 20% of your keywords, what would be the minimum number of keywords required before someone could consider a portfolio methodology?
Unless I am missing something, this would depend on the number of keywords. In one campaign I have about 1000 keywords of which 200 are watched closely (I didn't choose 20% - it is a coincidence) while the others are either new or can be kept for a little longer or are ready for disposal. Of the 200, I separate major stats by time periods, e.g. old web site vs new site, so that I am comparing apples with apples.
I sort first by Revenue, mark the terms with a colour; then sort by some other columns, each time marking the terms; then I finally sort the terms in alphabetical order, as this groups the same term beside its other match types, e.g. "red widget" might be coloured but its broad term might not. This process identifies the high CPA terms and the very low CPA terms and if the overall CPA is good, then we're doing well with a portfolio strategy. If I had a lower budget or some arbitrarily low number of keywords, I'd take much longer to find this PPC equilibrium.
As is to be expected, the best terms for one SE are not just as hot for another but would still be in the top 50. Periodically I check such anomalies to see if I can do something such as copying a good ad to another SE within the constraints of the medium. Sometimes I try a new 3rd tier engine with penny clicks to try wild theories or terms that are too expensive on the majors. Sometimes a major such as Overture has an overlooked secret such as Content Match where there are cheap terms still available.
So yes, someone has just borrowed a buzzword from the finance industry but it's all about spreading risks.
