Choosing a Pay-Per-Click (PPC) pricing model which works efficiently for both client and agency is a difficult process. A good pricing model should be simple, should create incentives for the agency to perform and should be a fair measure of the work and expertise involved.

One common model that many agencies use is the ‘markup’ model (also commonly known as the ‘percentage of spend’ model). If the agreed markup is 10%, and the client spends $30,000 on clicks, the client pays $33,000, of which the agency receives $3,000.

Nice and simple.

But does it create incentives for the agency to maximise profit for the client? Does it fairly reflect the work and expertise involved at all spend levels?


Conflict of Interest

In short, the percentage of spend model is a highly inefficient pricing model for paid search management, and should be avoided. As pointed out by George Michie in his recent post on SEM Pricing Models, since the agency receives a commission on every dollar spent, there is an incentive for the agency to spend as much as possible, which can be far in excess of the point of diminishing marginal returns.

To find out exactly what George means by diminishing marginal returns, and how it creates a conflict of interest for client and agency and renders the markup model pretty much useless, let’s consider the cost and revenue structure of the client.

Look at the line CPC (marginal) in the diagram below (Cost Per Click). It shows what happens to the Cost of each subsequent Click as the volume of clicks increase. It is upward-sloping, so each extra click costs progressively more than each previous click. Makes sense – since the first few clicks are usually very cheap, and raising bids and showing for more expensive keywords is generally needed to increase click volume.

Now have a look at the line RPC (marginal). It stands for Revenue Per Click, and shows how much Revenue is generated from each subsequent Click. It is downward-sloping, which again makes sense, since each additional click is likely to be less relevant and have a lower conversion rate than each previous click (this is known as diminishing marginal returns). A rational advertiser would always go for the low hanging fruit first (the most relevant keywords), which will naturally convert better than the high hanging fruit (less relevant generic keywords).

(If this all sounds very confusing, you may like to check out Wikipedia’s articles on marginal cost and diminishing marginal returns first).

So we have an upward-sloping marginal CPC line and a download-sloping marginal RPC line.

Now, assume the client has no other costs other than paid search click costs. If the client spends $1,000 on clicks and generates $1,500 in revenue, the client makes $500 in profit (this is of course very unrealistic – but just bear with me for the sake of argument).

Look at where the two lines cross. This is the level (2,000 clicks) where the client will make the most profit.

Upward Sloping Marginal Cost (MC) Curve


Well suppose click volume increased to 2,100 clicks. The 2,100th click is now costing $0.85, but is only bringing in $0.55 of revenue! The client is losing money on these extra 100 clicks, so reducing click volume would increase the client’s total profit.

Downward Sloping Marginal Revenue (MR) Curve

What about reducing click volume?

Well, at 1,900 clicks, the 1,900th click is costing only $0.65 but generating $0.85 of revenue. The client is making $0.20 profit from their 1,900th click, so why not increase click volume further, and make $0.19, $0.18 and $0.17 profit from additional clicks? It makes sense to increase click volume until no additional profit is being made – until the cost of an extra click equals the revenue of that click.

Profit Maximization Where MC=MRIt is where the marginal CPC and marginal RPC lines meet that no additional profit would be made, and it is at this point which makes the client the most profit. It is this level of click volume that the client should aim to target with their PPC activity.

Economics Maximum Client Profit

Now let’s complicate things a little. The graphs above were only concerned with marginal costs and marginal revenues – costs and revenue at the margin. They show what happens to the next click or the previous click, but they don’t show what happens on average – to the average cost per click or the average revenue per click. Average CPC and average RPC lines are therefore needed for this purpose.

Have a look at the red CPC (average) line and see if it makes sense. Like the green CPC (marginal) line, it’s upward-sloping, but flatter. Why?

Think about it for a second.

If you just spent $1,000 on 2,000 clicks, each click costed you $0.50 each on average, right? If you then decided to go crazy and purchase a few extra clicks on some expensive keywords for a hefty $4.00 each, what will happen to your average CPC price? It will increase, but not by very much, maybe to $0.41? Adding some expensive clicks will pull up the average, but only by a relatively small amount. Hence the flatness of the average Cost Per Click line.

The same is true with the average Revenue Per Click line. If a few extra keywords bring in only a small amount of revenue, it will pull down your average revenue, but only slightly, hence it’s flatness.

Average Cost Curves Are FlatterStill following?

Great. So we now have 4 lines which represent the cost and revenue structure of a client:

  1. Marginal CPC – shows how much each extra click costs
  2. Marginal RPC – shows how much each extra clicks generates in revenue
  3. Average CPC – shows how much clicks cost on average
  4. Average RPC – shows how much clicks generate in revenue on average

These 4 lines are all that’s needed to assess the client’s profitability.

Now remember how we decided that 2,000 clicks was the most profitable click volume for the client? Let’s see exactly how much profit the client is making from 2,000 clicks.

Well, at 2,000 clicks, the average cost per click (CPC average) is $0.30. The client is spending $600 on clicks ($0.30 x 2,000).

At 2,000 clicks, the average revenue per click (RPC average) is $1.50. The client is generating $3,000 in revenue ($1.50 x 2,000).

PPC Pricing Markup Model

Minus one from the other ($3,000 – $600) and we have a healthy client profit of $2,400.


Profit from Percentage of Spend PPC Model

Now this is where the inefficiency with the percentage of spend (markup) model comes in. Since the agency is paid a commission on every click, the agency will always want to increase click volume and spend as much as possible. As we’ll see from the following examples, this is often in excess of the point of maximum client profit.

Suppose the agency increased click volume to 3,000. The average Cost Per Click (CPC) increases from £0.30 to $0.45, and the average Revenue Per Click (RPC) falls from $1.50 to $1.20. The client is still making a healthy profit of $2,250 (revenue of $3,600 ($1.20 x 3,000) minus cost of $1,350 ($0.45 x 3,000)), although their profit of $2,250 is less the previous level of $2,400.

The agency, however, has increased their profit, since they now receive a cut of a bigger spend ($1,350 instead of $600). Assuming the markup is 10%, the agency’s profit has increased from $60 to $135, at the expense of the client’s profit.

But here’s the thing. The client is unlikely to complain – the agency is making them $2,250 of profit, after all! How is the client to know that they could be making $2,400 of profit, should the agency decide to reduce click volume? The client is none the wiser and would most likely praise the agency for their ‘efficient’ work in making them such as tidy profit of $2,250!

Why Marking up Click Costs is Inefficient

But why stop there?

The agency makes a cut of every click spent, so why not increase click volume further? Why not increase it to – wait for it – 4,000 clicks!

Here, at 4,000 clicks, click spend will be $3,000 ($0.75 x 4,000) so the agency’s profit will be $300 (10% of $3,000) – much better than the measly $60 or $135 from the previous click volumes of 2,000 and 3,000.

But look at what’s happened to the client’s profit at this new click volume of 4,000. Their costs are now $3,000 ($0.75 x 4,000) and so is their revenue! The client is making no profit at all! Any more spend, and the client’s average revenue will fall below their average cost, and they will make a loss. If the client is losing money, they will most likely leave the agency, or at least apply massive pressure on the agency to increase performance (reduce click volume), so any click volume in excess of 4,000 is not sustainable in the long-run.

AdWords agency has incentive to spend as much as possible

So it’s in the agency’s interest to maximise their commission by getting the click volume as close to the point of 4,000 click where the red lines cross (where average costs equals average revenue) – but without going over, so as to keep the client happy (the client will still be making some profit).

Confict of interest between SEM advertising agency and client

What happens is a negotiation of pushing and pulling between the client and agency until a compromise is found – say 3,000 clicks. Exactly how close to the point of maximum client profit or the point of maximum agency profit is settled upon depends on the relative bargaining strengths of client and agency, access to cost and revenue information and countless external influences to name just a few factors at play.

Paid search markup model is inefficient

What is clear though, is that whatever click volume is reached as a compromise, it will not be efficient. It is impossible to maximise the profit of both client and agency using a percentage of spend model. At every click volume, there will always be a way to increase agency or client profit by adjusting click volume.

With a percentage of spend model, there is no working together of client and agency, no common goal, no shared vision. It’s a constant pushing and pulling and conflict of interest. Time and effort is wastefully diverted into politics in an attempt to move click volume closer to one party’s optimum, not to mention the reluctance of each party to be open and transparent and share useful information with each other. Doesn’t sound like the foundations of a successful and lasting business relationship to me. Perhaps it’s why some agency churn rates are so high?

Of course, no PPC pricing model is perfect – every method will have its weaknesses. The key is to find one that works best for your goals and objectives as a business, and one which appropriately compensates the agency for their efforts in helping you develop your paid search marketing strategy. But in terms of aligning the agency’s monetary motivations with that of your business, and creating incentives encouraging them to maximise your profit from paid search, the percentage of spend model fails miserably.

My advice: use the percentage of spend model with caution.

Are you a fan of the percentage of spend model? Have you made it work for client and agency? Or does it create more problems than it’s worth? Share your thoughts in the comments section below.

Next: cost-per-sale performance models – rewarding agencies based on how they perform. Do they work? Are they efficient? Economics of PPC Pricing: Why Performance Deals Often Fail

Alan Mitchell

Alan Mitchell is a Google AdWords PPC specialist, based in Melbourne, Australia, with a proven track record at improving return on investment (ROI) from Google AdWords. Find out how Alan can help your business.

  • Pingback: Twitted by ppc_advice()

  • Pingback: PPC Prices and Marginal Utility theory « Bett-zi()

  • George Michie

    Alan, this is great stuff.

    I’d make three minor points based on our experiences at RKG:

    First, there’s a distinction to be made between “observed revenue” and “actual revenue”. Because of cookie breakage (dropping cookies, which is rare, and shopping and buying on different machines, which is much more common), spillover to the call center, javascript blindness for those who rely on it, and the like, we often don’t see all the revenue driven. Driving to observed profit maximization may not maximize actual profits.

    Second, immediate profit maximization isn’t always the best plan. Lifetime value considerations and even some “brand-building” thoughts make many of our clients believe that breakeven (observed, or even factoring in that tracking drop off) is the right target to maximize their long-term benefit. Short-term profit maximization can lead to a business death spiral over the long-term.

    Third, we rarely find ourselves in the position of determining what the advertisers target should be. Sometimes they ask us to help them think through all this stuff, but oftentimes they know the metric they want us to hit, and we’re evaluated by how large we can grow the program within that target.

    All that said, I think any pricing model can be gamed by agencies who put their short-term revenues first, and their clients interests second. We think the shrewd agency recognizes that pursuing the clients interests first and last is what builds a profitable business in the long run. Client churn is costly and requires huge marketing investments to continually replace the alienated. Not only does “doing right by your clients” allow you to sleep better at night, it turns out to be a good business decision as well.

    Thanks for the call out! I love any blog post that uses graphs smartly and references marginal analysis!

  • George Michie

    Reading my comment just now I’m appalled by several apostrophe catastrophes! We’re on vacation over here!

  • Nomansland

    Great post! You are absolutely right. There is a definite conflict of interest between the two parties. Completely unethical practice to continue, you are right on about not basing the foundation of a relationship on a mutual conflict of interest, that road will be treacherous for all concerned.

    I look forward to reading your posts.

  • Alan Mitchell

    Thanks for your comments George and Nomansland.

    @ George

    You’re right – observed, or reported, revenue is generally a pessimistic measure of actual paid search revenue, which can make profit maximizing decisions extremely difficult.

    Interesting point too about short and long-terms goals. Guess my analysis was based more on maximizing short-term profit, which as you point out, is often unrealistic for many businesses.

    I think if the article had more of a long-term focus, we would see the point of ‘compromise’ closer to that of ‘maximum client profit’, as keeping existing clients happy will always be more profitable for the agency (and client) in the long-run.

    Guess that’s the problem with economic modeling – over-simplify and it becomes unrealistic, over-complicate and you lose clarity. Glad you enjoyed it though!

  • Digital Lookout

    Excellent post Alan. As well as George’s comments I would also throw in sales volumes as a complication to this model. In my experience many clients are facing an ongoing battle between profitability and sales volumes. Some also have volume targets they must hit to keep the business afloat, or incremental profits at different volume levels, all making this sort of analysis 10 times more complicated!

  • Alan Mitchell

    @ Rob

    Thanks for you comment!

    You’re right – sometime maximizing volume becomes a larger priority for clients than maximizing profit, at least in the short-run, which can make analysis all the more complicated.

    But I think in the long-run, assuming the client understands their cost and revenue structure accurately, and assuming it’s possible to accurately attribute all the revenue generated from paid search at every touch point (which are big assumptions), there will undoubtedly be a optimum point which would make the client the most profit, which will inevitably be at a lower click volume than the optimum point which will make the agency the most profit.

    But you are right – adding growth as a business objective (rather than just profit maximization), as well as short and long-term motivations does complicate the model considerably.

  • Rob McCance


    Heck of an analysis.

    Very thorough and well researched and I don’t mean to belittle it when I say that I can’t believe anyone pays to have someone else run a PPC Campaign for them.

    I know this is mypotic and there are many business people with legit businesses that don’t know much of anything about the internet, or PPC, or web sites, or SEO or any of it.

    I once had a Broker that paid huge bucks to a PPC reseller who did not much more than come up with all the main primary KWs and recommended he pay $5 a click.

    Probably they were on the % of spend model. LOL!

    Good work on this though. RM

  • Tom Jones

    What a post – brilliant analysis and observations of the agency/client dynamic. Great work, Al.

    I am completely with you in that the % spend model , when applied incorrectly (and often it is) is divisive and leads to the problems you’ve described.

    There are instances, however when the model can work very well indeed – that is, when a client’s marketing goals are achivable within the threshold limits of the marginal CPC/RPC. In such cases both sides are equally incentivised to maxmise performance. The difficulty is, of course, calculating this tipping point – which can usually only arrived by through an open and clear communication between both parties. This is the problem many agencies/clients make – avoiding this step in the hope of ‘gaming’ the commercials for short-term profit maximisation. There then also needs to be a pre-determined negotiation point at which commercials are renegotiated, ensuring that interests are realigned.

    Another consideration is that the %spend model does not work well when spend levels are so low as to not adequately reward an agency for working at developing a campaign.

    Often this position can result in a impasse where both parties, influenced by short-termism, are unwilling to invest the time or money needed to that the required effort to the detriment of fulfilling the long-term potential of a campaign.

    Keep the good stuff coming, Al!


  • Justin Hayward

    Good insight on the markup model Al. The only thing i would add is to consider many agencies, us included are working on percentage spend models but these are tied to additional metrics including cost per sale or ROI. So while the agency does make more with a higher spend this is only within the restrictions of the requested client efficiency including fee.

    This works particularly well, and i think is a fair model as you mentioned in your post it sometimes does not accurately reflect the level of work required to achieve targets but over a longer working relationship period it evens out.

    The other main option of charging a flat management fee per month is clearly suicide for the client; no incentive for the agency to do any better than hit a specific agreed CPS/ROI target and sit on that with minimal workload as there’s no incentive to do any better. The only way this would work is to add an override into the mix for performance over and above what is expected. THis can still lead to grey areas however.

    Lastly, consider that the markup % of spend model is the same as most other media out there (TV, DM, Radio etc) and is something that can be easily understood by every client, budgeted for and fit into existing client finance models and payment terms.

  • Alan Mitchell

    Thanks for your comments Rob, Tom and Justin!

    @ Tom

    You’re right – the model does become more practical and beneficial to both parties when targets are regularly set and long-term goals take priority over short-termism, although as you point out, this does require a open dialogue between client and agency and a clear understanding of the client’s long-term strategy.

    Nice comment about low spend levels – there is definitely a minimum amount of effort required to build and maintain a successful PPC campaign, so a minimum spend will be needed to make it worthwhile for the agency.

    @ Justin

    Again, you’re right – performance metrics such as cost per sale and ROI definitely improve the percentage of spend model, since they encourage the agency to maximise the client’s sales subject to a profitable cost structure.

    Any performance-related pricing model, however, requires a good understanding of marginal costs and revenue at different spend levels (in order to efficiently set targets), which is easier said than done – especially if the client is relatively new to online marketing and conversion performance data is scarce.

    Interesting point too about the value of simplicity. Creating an intricate pricing model can be extremely time consuming, expensive and distracting for all concerned, so it may be in the client’s (and agency’s) best interests to go a simpler model, even if it means some loss in potential profit.

  • Richard

    Hi Alan,

    I thought that this was an excellent demonstration of economic theory which has been apllied to PPC expertly.

    I would like to take this a stage further and conduct your analysis with one of my clients.

    My question to you is:

    How do you accurately calculate the marginal CPC when you work out the figures for real as I am struggling.


  • Alan Mitchell

    @ Richard,

    You’re right – calculating marginal CPCs can be tricky. One option would be to adjust your keyword bids methodically, and take a note of click volumes. If enough data is collected for different CPC levels, you could have a fairly good understanding of the click volume at say $0.65, the click volume at $0.95 and the click volume at $1.15, and hence work out your marginal CPCs.

    This obviously assumes that there are no external fluctuations in click costs (which of course there are – think seasonality, days of week, times of day, competition), so to improve accuracy you might want to take advantage of bid adjustments using Google’s ad scheduling. Set different hours in the day (and week) to as many different bid levels as you can (20%, 50%, 85%, 160% etc), then run an hourly report and plot your cost per click prices for each hour against your click volumes for that hour. This should reduce external bias and help improve the accuracy of your marginal CPC calculations considerably.

    Another (quicker but less reliable) option is to use Google’s Bid Simulator to get an idea of how click volumes change as cost per click prices change, although I would tend to be very cautious using such data as its typically normalised across a large number of different advertisers.

    Hope this helps!