Jan 272011

The Performance-Based Pricing Fallacy: Part 1

"You won't pay a penny unless we produce results!"

Sounds compelling, right? A no-risk proposition, and the vendor's incentive is perfectly aligned with your marketing goals.

Or is it?

Simple math will demonstrate that, in fact, the incentives are not perfectly aligned and that performance based pricing -- whether CPA or revenue sharing -- guarantees overpaying your vendor and under-promoting your business.


Acme is happy to drive new business through marketing as long as they keep the ratio of cost to sales to 25%. They contract with a vendor who will deliver customers to them for a 25% share of the revenue, and the vendor takes all the risk by paying for the media.

This seems like a terrific deal for Acme, guaranteeing efficiency and a chunk of new business, but because of the law of diminishing marginal returns, it turns out to be an inefficient mechanism for pursuing that marketing channel -- whatever channel that may be.

The Law of Diminishing Marginal Returns

The first marketing dollar spent will generate greater return on investment than the last dollar spent, which in turn will generate better returns than the next dollar spent. This is true whenever marketing dollars are spent wisely because at each level the smart marketer buys the most efficient traffic available. By definition then, the ROI has to decline as the spend increases. If it does not, the marketing team doesn't know what it's doing.

To be clear, we're not talking about chronology here, we're talking about incremental spend levels. An advertiser who generates $100K in revenue on $25K in media one month is likely to be able to do that again next month. However, that same advertiser will find that spending $50K on on the same type of media in a month will not generate $200K, but rather something less than that.

So, let's look at some hypothetical data for a marketing channel purchased efficiently as the spend level scales:

This graph shows advertising spend purchased in $5K increments (the small tick marks), with each successive level of spend being slightly less efficient than the last.

From this same data set we can show the incremental revenue generated by each additional $5K in ad spend

This is the crux of the problem with revenue sharing and cost per action models.

The Revenue Sharing Model

This advertiser agreed to pay the vendor 25% of revenue generated as commission, and it's up to the vendor to buy the media.

If we take a look at the incremental sales graph from above, and add in both the incremental commission paid to the vendor and the incremental out of pocket cost in media to the vendor we see something interesting:

While the incremental commission paid to the vendor decreases, the vendor's out of pocket expenses do not. Let's look a bit closer:

The point at which the incremental commission gained drops equal to and below the incremental cost to the vendor is the point at which it becomes irrational for the vendor to spend more money. [Note: a good case could be made that this is the right media spend level for the advertiser, too, but we'll come back to that point]

Looking at this data in a slightly different way. We can calculate the vendor's take-home revenue -- commission paid to the vendor minus the media expense -- for each level of media spend and get a very pretty graph:

Commission minus Media Costs = Real Revenue

What this tells us is that a rational vendor that buys media efficiently will spend ~$105K in media for this advertiser, because that's what maximizes the vendor's revenue.

If we look back to the commission versus sales graph and add in some drop lines we'll see:

The vendor spends $105K in media, which generates just over $866K in sales for the advertiser.

Here's the rub: the advertiser has agreed to pay 25% of sales, so they pay the vendor a touch under $217K, less than half of which went to pay for media!!!

The Percent of Ad Spend Model

If instead, this advertiser decided to pay their vendor based on a percentage of ad spend (using 12.5% as a fair number) let's see how the vendor's revenue compares.

Since the vendor doesn't get the huge windfall from picking the lowest hanging fruit, the % of spend model means the vendor is just as happy to spend the next $5K on media as they were to spend the last $5K.

Note also how much less of the out of pocket expense goes to the vendor's coffers vs media.

Charting this against advertiser sales shows:

In the rev share model, $217K in commission generated $866K in sales. Under the percentage of ad spend model the advertiser spends $190K on media, pays the vendor $24K in fees, and for a total spend of less than $214K the advertiser would get $1,121K in sales.

Performance-Based Pricing Fights Back

The rev share vendor deserves to make more money because the vendor has assumed the risk and fronted the media expense. The advertiser has to pay for that security and that loan.

Sure, I grant that. However that doesn't mean advertisers should pay those premiums when an alternative pricing model will get them significantly better results for the same total spend. There may be circumstances under which no vendors will operate on a % of spend model, but for paid search, display advertising, re-targeting, comparison shopping engines, and more there most certainly are!

The overall cost lines are converging, and in fact if you run the numbers you find that at a media spend of just under $295K the advertiser would end up paying a total of $330K for media and vendor fees under either pricing model and generate $1.32 Million in sales at a 25% overall cost to sales ratio. See, the model doesn't matter!"

True, but the problem is the advertiser will never be able to push the rev share vendor to spend that aggressively, because the vendor loses money every time they spend a dollar beyond $105K. Their incentive is to spend less aggressively than the advertiser might be willing to spend.

Look, the reason the rev share vendor loses money beyond $105K in media spend is that that is where the marginal media cost to sales ratio reaches 25%. Every dollar spent on media beyond that is less efficient than the advertiser's stated target efficiency.

Good point, but they could pay a percentage of spend vendor a shade over $13K to spend $105K, instead of paying a rev share vendor $112K to do the same, and as we showed before, if they're happy spending $217K total, they can get a great deal more media and sales by hiring a vendor on a percentage of spend basis.

You've cooked the numbers to make rev share look bad and % of spend look good.

Not so, this all falls naturally from diminishing marginal returns. A rev share or CPA model makes the advertiser pay the same amount for orders or dollars regardless of spend level. The low-hanging fruit is the easiest to get cheaply, and the cost of acquiring a sale always goes up from there, hence the incentive to leave the rest of the fruit hanging.

The raw data is here, feel free to play around with it.

If the advertiser pays the media, and comps the agency on a rev share, that would fix the problem, wouldn't it?

Not completely. As the incremental sales revenue drops the incremental commission to the vendor drops creating increasingly less incentive for them to do productive work. That may or may not be a big deal. Certainly even under a fixed fee model, the vendor has an incentive to keep their clients happy so as long as the fee makes sense, they'll keep working. But performance-based pricing is attractive for its supposed alignment of incentives, and the point is that the real financial incentive created is to grab the lowest hanging fruit and leave the rest.

However, the problems don't end here. The larger problems with any type of "performance based pricing" will be discussed in Part 2 of this series.


18 Responses to "The Performance-Based Pricing Fallacy: Part 1"
George, Thanks for a really good, well-thought-out, thought-provoking article. It's of particular interest to me since my company is compensated for PPC campaign management on a pay-for-performance model. In a typical client engagement, we are continually finding new sources of traffic, helping improve the conversion rates of the client's site, and improving profitability using a variety of tactics that includes ad testing to improve CTR and hence quality score. I'd be interested in your take on how the graphs and figures change when those factors re variable rather than static. Thanks, David
Jim Novo says:
Brilliant. Now if we could only get people to care about optimizing for profit rather than sales...
David, thanks for your comment. I certainly do not mean to poke your fine team in the eye. As we both know smart agencies focus on their long term interests in keeping clients happy more than their immediate term financial interests. If your clients pay their media expenses directly and your p-f-p commissions are therefore lower, the disincentive to spend evaporates. And, for firms managing thousands of tiny accounts, Brad Geddes pointed out that p-f-p or fixed fee pricing is likely the only model available to those advertisers, because no firm will work on a percentage of spend -- the numbers are too small. In fact, we can't manage Facebook ads on a percentage of spend for our clients for the same reason. No question, good firms will work hard for their clients regardless; my point is that p-f-p pricing is often pitched as a golden recipe for incenting performance, and it simply isn't that.
Thanks Jim, and in part 2 of this series... :-)
George, No poke perceived! I thought your comments were fact- and data-based and indisuptable. and I agree that "...p-f-p pricing is often pitched as a golden recipe for incenting performance, and it simply isn't that." FYI, all of our clients pay media expenses directly to the media owner. Looking forward to Part 2! Cheers, David
Tomas says:
Hi George I just want to add something to David's comment (comment 1): I've seen many cases where the sales curve is very different based on the incentive model used by the advertiser. Although I mostly agree with you, I do think there are more variables required to make your case more accurately. In the end it comes down to return on effort, which is much easier to achieve on a % of spend model and most people won't put the effort in if they're not required to.
Hi Tomas, Yes, this is an "idealized" model assuming everyone will wring the same performance from the opportunities on the table. That's clearly not the case in the paid search space, nor in many others. A great agency with a bad pricing model will likely run circles around a poor agency paid under a better incentive structure. I wanted to use the "ideal" to highlight the point that even in theory the performance-based models don't create the incentives people believe they do.
Eddie says:
I can't wait for part two... Spot on thus far.
Fernando Constantino says:
Hi, i love this blog. I think it's the best to deal with technical aspects of ppc. Unfortunately, as has been suggested, this article disregards what should be the main goal of every search marketer, profit maximization. Setting a fixed cpa target is a horrible starting point, so the results are worthless in my opinion. With a profit maximization goal, paying the agency a percentage of the spend generates a total missalignment between advertiser's and agency's goals. This is just my (an advertiser) opinion. Keep up your great blog!
Thanks Fernando, stay tuned for part 2 :-) Share of profits (or marketing income) is better, but that has problems as well.
Tim says:
George... I'm looking forward to part two as well... I know first hand the share of profits line.
It's out, Tim. Enjoy.
Terry Whalen says:
George, I wish I had been able to reference this a while back when talking to a large dating site (that starts with an "e"). My slide didn't do it - you illustrate and explain this well!


Check out what others are saying...
[...] This post was mentioned on Twitter by Rimm-Kaufman Group, Search Marketing and Long Tail Tool Maven, search tool madman. search tool madman said: The Performance-Based Pricing Fallacy: Part 1: “You won’t pay a penny unless we produce results!” Sounds compel... http://bit.ly/gtgNyl [...]
[...] The Performance-Based Pricing Fallacy: Part 1, Rimm Kaufman [...]
[...] In Part 1 we demonstrated that the Law of Diminishing Marginal Returns dictates that revenue sharing/commission and CPA models do not create the positive incentives they are widely thought to create. In fact, they create disincentives to push the limits. [...]
[...] Performance Based Pricing Fallacy Part 1 – Performance Based Pricing Fallacy Part 2 [...]
[...] be a no risk proposition for the advertiser. But is it? Most performance based pricing models have hidden pitfalls that make the reality less advantageous than it appears on the [...]

Leave A Comment