For a business looking to hire a pay per click (PPC) agency, cost-per-sale (CPS) performance models are great. The business pays the agency a set price for each sale, so fees are entirely based on the agency’s performance.
From a client’s point of view, this is great. There is little risk – agency fees are only payable once sales come in. Guaranteed profit!
From an agency’s point of view, it’s also great. Each extra sale is extra revenue, so an agency which is confident of its abilities to deliver value from paid search is rewarded heavily (and fairly) for their efforts. Performance-related pay creates an incentive for agencies to invest their best resources and expertise into making PPC campaigns a success for their client.
Researching cheaper and high-converting long-tail keywords, restructuring ad groups to improve relevancy and regularly carrying out landing page testing to increase conversion rate become all the more worthwhile when there’s a monetary incentive. If an agency only gets paid when they deliver sales, it is worth their time and effort to deliver sales.
Sounds too good to be true. Client risk is minimal. Agencies which perform are rewarded. Agencies which don’t perform…well they are forced to perform if they are to stay in business.
So you’ve decided you want to give performance pricing a go. But how exactly would a performance deal work? And how should you go about creating one for your PPC agency?
To see how such performance deals would affect your bottom line, and the bottom line of the PPC agency you’ve hired, we’ll use a similar method to that used in Economics of PPC Pricing: Why the Markup Model is Flawed, and look at the cost and revenue structures for both client and agency.
Let’s suppose Shiny Shoes Corp launched a new range of shiny red golf shoes, which they sold online for $15 (OK, they’re very cheap golf shoes). Sales were stagnant, so in an effort to increase sales of shiny red golf shoes, Shiny Shoes Corp decided to pay a PPC agency $10 for every pair of shiny red golf shoes they sold through PPC. If the PPC agency only sold 10 pairs a month, the client would pay the agency $100 (10 x $10), regardless of whether the agency spent $50 or $5000 on clicks. If the PPC agency were to generate 1,000 sales, the client would pay the agency $10,000 (1,000 x $10) – again, regardless of whether the agency spent $50 or $5000 on clicks.
How many sales will the PPC agency deliver? How much profit will Shiny Shoes Corp make? How much profit will the agency make? Will the agency be incentivised to deliver extra sales? Is $10 per sale a fair amount, or should it be $12, or perhaps $8? All burning questions we can consider with some simple economics.
Look at the green line ‘MC (client)’ in the figure below. MC stands for manginal cost, and shows what happens to the costs of Shiny Shoe Company as sales volume increases by extra units. It’s a flat line, which makes sense – the client is paying the agency $10 per sale regardless of sales volume.
Since Shiny Shoes Corp pays the agency $10 per sale, we can also add in the agency’smarginal revenue (red line), which again is flat. At every sales volume, whether it’s 20 pairs or 2,000 pairs, the client’s cost per sale is $10 and the agency’s revenue per sale is also $10.
We also know that Shiny Shoes Corp sells the shiny red shoes online for $15, so we can add in the client’s marginal revenue. Again, it’s a flat line, since Shiny Shoes Corp receives $15 for each sale, regardless of how many pairs they sell.
Now that we have the cost and revenue structures for Shiny Shoes Corp, we can see what profit they’re making. Shiny Shoes Corp’s profit is the difference between their cost and revenue, which is represented by the green shaded area. The higher the sales volume, the higher their profit. Since Shiny Shoes Corp makes $5 profit for every sale ($15 they recieve from the customer online minus $10 they pay the PPC agency), they would want as many sales as possible!
How many sales will the agency will deliver depends on the agency’s profit, so let’s add in the agency’s marginal costs (red line). The agency’s MC is an upward sloping line, since click costs increase progressively as volume is increased.
Why is it upward-sloping?
Well, the first few sales will come from cheap clicks from brand terms, such as ‘Shiny Shoes Corp’, as well as highly-relevant long-tail keywords, such as ‘red shiny golf shoes for sale’. Not only are long-tail keywords generally cheaper, but conversion rates for these highly-relevant keywords will also be high, meaning fewer clicks (and therefore fewer click costs) will be needed for each sale. The agency can probably bring in the first few sales for only $1-$2 each.
Once the PPC agency has saturated brand terms and highly-converting long-tail keywords, they are forced to show ads for more competitive, more generic and more expensive keywords, which might not convert as well. The higher cost per click prices and lower conversion rates for less-relevant keywords such as ‘golf shoes’, mean sales might cost the agency $8.
Each extra sale costs the agency progressively more, hence the upward-sloping marginal cost line. Diminishing marginal returns set in.
Now we have all our cost and revenue lines for the agency, we can see how many pairs of shoes will be delivered by the PPC agency, and how much profit the client and agency will make.
After a while, when all profitable keyword opportunities are exhausted, when all that is left are expensive and poorly-converting keywords such as ‘shoes’ and ‘birthday gift ideas’, the the agency will start to make a loss from any extra sales they deliver, where the agency’s Marginal Cost goes above the agency’s Marginal Revenue . At this point, the agency will reduce bids, pause keywords and cut back on sales volume until they stop making a marginal loss.
The agency’s maximum profit is where their Marginal Cost equals their Marginal Revenue, at 5,000 pairs of shiny red golf shoes. Any less than 5,000 pairs (say 4,000 pairs), and the revenue for the agency delivering extra pairs of shoes would be more than their cost of delivering those extra pairs (their Marginal Revenue would be greater than their Marginal Cost), so it would be profitable for the agency to increase volume. Any more than 5,00 pairs (say 6,000 pairs), and the agency would make a loss from these extra pairs (their Marginal Revenue would be less than their Marginal Cost), so they would be better off by reducing volume.
Maximum agency profit is therefore where the agency’s MC = MR, at 5,000 pairs of shoes.
At this point of maximum agency profit (5,000 pairs of shoes), we can see the profit for both client and agency. Looks healthy, right?
At 5,000 pair of shoes, there is a wastage, an inefficiency – what economists call a loss of utility. There is extra profit to be made (dark grey area), but since the agency doesn’t want to make a loss from helping the client achieve this extra profit, the extra profit goes unrealised. This makes model inefficient. There is extra profit to be made, but the model does not sufficiently extract this extra profit the market is offering.
The cost-per-sale model starts to breaks down.
What’s more, with a $10 cost-per-sale model, the fee the agency receives is a poor reflection of their efforts.
A good PPC pricing model should be an accurate reflection of the agency’s work and expertise, which is not the case with this $10 cost-per-sale model. The first few sales will be from showing ads for brand searches such as ‘Shiny Shoes Corp’, which require little agency effort and will cost the agency next to nothing. Agency profit will be high.
But as sales volume increases, and the agency is forced to invest considerable time and effort to research new keyword opportunities, agency profit will be relatively low. This and low effort/high reward and high effort/low reward problem might cause the agency to cut back on volume even further, thinking these these extra sales and tiny profit (which are highly valuable for the client) are just not worth the extra effort.
So, to recap, Shiny Shoes Corp’s decision to pay a PPC agency $10 for every sale is inefficient for two reasons:
Okay, so a set cost-per-sale model doesn’t work.
But what about a structured cost-per-sale model, such as below? Shiny Shoes Corp pays the agency only $7 for the ‘easy’ sales (brand terms), and $13 for the more difficult sales. With two different pricings, the fee the agency receives is a better reflection of the value they are adding (great!), the loss of utility is reduced (great!) and the agency now has a monetary incentive to increase sales volume to 6,000 (great!).
Awesome! It’s solved both of the problems above! Right?
There is still a loss of utility (the client can still make extra profit if it received these extra sales for $14 or $14.50). And although the agency is more fairly rewarded for their efforts, the agency’s red Marginal Revenue line still does not follow their Marginal Cost line closely. They are not being paid proportionately for their efforts.
How about a multi-step (or progressive) cost-per-sale model, such as below? Shiny Shoes Corp would pay $6 for the first 1,000 sales, $8 for the next 1,000, and so on.
With multiple pricings, loss of utility is tiny, and the agency is fairly rewarded at all spend levels (notice how the agency’s red MR lien closely follows the agency’s red MC line.) What’s more, sales have increased to 7,000! Great!
Such a well-structured model would be great, and as close to the Holy Grail of PPC pricing as one might hope to get. It’s fair, creates incentives, maximises sales volume and maximises total profit (client and agency).
But just as the Holy Grail is hard to find, so is such as pricing model.
A multi-step (or progressive) cost-per-sale model is great in theory, but in practice such a model might be prohibitively difficult to construct. For a progressive cost-per-sale model to work efficiently, it requires an in-depth understanding of revenues at multiple sales volumes, knowledge of brand strength, seasonality, offline marketing influences, not to mention tracking inefficiencies and bias which might exaggerate or under-report sales volume.
One scenario might be where brand strength and click costs are under-estimated by the client, such as below. Fees set too high, and the agency is over-compensated for their services. The client would make little profit, and would be better off with a percentage of spend or management fee pricing deal.
Another scenario might be where click costs and brand strength are over-estimated by the client, such as below. Fees are set too low, so the agency makes a loss from delivering any sales, and the client will struggle to find an agency wanting to manage their PPC for more than a few months. As they regularly jump from one optimistic agency to another, long-term stability goes out the window and are replaced by extra costs, effort and confusion.
A more realistic scenario might be the one below, where in an effort by Shiny Shoes Corp to closely match click costs to fees, offers a slightly under-priced cost-per-sale structure. At some volume levels, delivering extra sales would make the extra agency profit, but and at other volume levels any extra sales would be at a loss. If the agency was at point B, for example, they would be better off either increasing volume to point C, or reducing volume to point A.
Very inefficient. Sales are vastly under-delivered and nobody wins.
This difficulty in creating a cost-per-sale model and setting it at the correct level leaves the client (and agency) with a difficult decision.
A correctly set cost-per-sale deal would be great for the client, as it minimises their risk (they only pay when sales come in), as well as encouraging the agency to perform. It would also be great for the agency, who are fairly rewarded for investing their time, effort and resources into doing what they do best.
But an incorrectly set cost-per-sale model (such as the three above) could be disastrous. If the agency is making a loss, they will refuse to deliver any sales, which is a waste of time for both client and agency and a massive loss of potential sales. If the cost-per-sale fee is set too high, the client will make only a small profit a would be considerably better off with a set management fee or percentage of spend (markup) deal.
So, if you are a business looking to minimise your risk from PPC, as well as incentivise your PPC agency to perform, should you consider a cost-per-sale performance deal?
Firstly, be wary of offering a set fee for every sale. You will pay heavily for the agency to simply harvest the value of your brand, and will find that sales are restricted way below your most efficient point.
Secondly, be careful of offering a multi-step progressive model, unless you have accurate sales and revenue data allowing for seasonal fluctuations, offline marketing efforts and special offers, and are confident (and preferably experienced) regarding likely PPC click costs at a range of different spend levels. If guesswork makes up a large part of your research, a progressive cost-per-sale model might be more trouble than its worth.
A happy medium, especially if you a business new to PPC or cost-per-sale models, might be a two-step cost-per-sale model, which separates brand and non-brand sales. It’s not ideal, but limits the brand / non-brand problem and is relatively risk-free (you only pay for sales so are guaranteed to not make a loss from PPC). It does, however, require a decent knowledge of brand searches and likely PPC click costs, so perhaps only consider this two-step cost-per-sale model once you have collected at least a few month’s worth of brand and non-brand cost and sales data.
And remember, a performance deal does not have to be based on sales. Leads, enquiries, email sign-ups downloads or anything which is a desired outcome for your business can potentially be incorporated into a cost-per-action performance deal. Just make sure you have the performance data to back it up.
Are you a fan of the cost-per-action model? Have you made it work for client and agency? Or does it cause more problems than it’s worth? Share your thoughts in the comments section below.
Next: profit sharing. Does it work? What are its limitations? Coming soon.