PPC Averages can Hide Incremental Nightmares

In case you missed it, this was posted on SEL yesterday:

How much can you afford to spend on marketing? This simple little question hides some vexing issues that are worth exploring during rocky economic times.

Suppose you budget \$200 to buy a DVD Player. When you arrive at the store you find the model you want is actually on sale for \$100! You arrive home, your 5 year-old hears the good news, grabs the left over \$100 and feeds it into the paper shredder. You are: a) indifferent because you planned to spend \$200 anyway; or b) irate because your son just wasted \$100?

Is this happening in your PPC program?

Let’s take a hypothetical retailer with 50% gross margin on all products and 10% variable costs tied to credit cards, pick and pack, cardboard, commissions, etc. They’re willing to spend the rest (40% of Sales) on marketing to capture as many orders as possible at break even and then make money on lifetime value. Their targets and goals aren’t really the point; the point here isn’t about setting targets, it’s finding out what goes into hitting the target.

Let’s say their PPC Search program (excluding their brand terms) spends \$160K per month and generates \$434K in sales for a 37% Cost to Sales ratio. On the surface, it appears that this program hits on all cylinders and achieves the desired metrics.

However, it’s important to look at both the averages and the incremental efficiencies to really determine if the program is doing what makes the most sense for the company writ large.

WARNING: THE FOLLOWING CONTAINS GRAPHS…

The mechanics are different, but the math is very similar to catalog circulation. If we look at the paid search spend in incremental chunks, like mailing segments, you’d obviously buy the most efficient advertising first. The law of diminishing marginal returns would then show that each successive advertising chunk would be somewhat less efficient than the last. So, for our hypothetical retailer, the curve might look like this:

Graphically, plotting the \$10K chunks of ad costs on the horizontal and the resulting sales and Net Margin on the vertical you see a classical representation of diminishing returns.

Let’s look at these same numbers a few different ways: first, let’s see what happens when we plot just the “incremental sales” rather than the total sales. In other words, for each \$10K increment in spend, how much did we generate in sales?

The first \$10K generated \$100K in sales, but that last \$10K in spend (bringing the total from \$150K to \$160K only generated \$3K in incremental sales). Plotting this as a function of efficiency and measuring the Cost to Sales (A/S) ratio for each increment yields:

Here, we can see that while the average efficiency increases from 10% to 37% as the spend increases, all of the spend after \$80K has come at worse than 50% A/S with the last \$40K coming at more than 100% cost to sales ratio. That last slug is tantamount to buying your own merchandise with marketing budget to push the top line!

Perhaps the best way to look at this is as a function of Marketing Income (Net Margin – Ad Cost).

The top line shows total Marketing Income, which is maximized when the Advertising Spend is \$60K. If PPC Advertising is to be a cash generator, this is the point where it makes sense to stop. However, there are many other goals to be addressed, and as with catalog circulation, one must be careful to avoid the death-spiral of collapsing marketing budgets.

Whether to view the program in aggregate or by increments is an important consideration, and the right answer depends not only on your firm’s tolerances, but on the shape of this curve. How smoothly does the efficiency degrade? For some of our clients we’ve found the efficiency curve to increase steadily to a point then shoot upwards. The shape of the curve depends on your vertical, the competitive landscape at the time, and other factors.

Determining the shape of the curve is not trivial. Experimenting with different efficiency targets to assess the ROI of the last increment and the next increment is the best approach. If you’re currently aiming at 30%, try 25%, and 35%. Recognize that the lag between clicks and orders can make any pull back in bidding look profitable, and any increase look inefficient; you’ll need to study the effect of the change on “same session” sales, or let the test periods run long enough to wash out the latency. Remember also that the tracked value is not the whole picture. If search is responsible for a big chunk of your company’s web sales keep an eye on overall ratio of marketing expense to sales to make sure any pull back isn’t costing you more top line than you think.

As we study our businesses to try to wring the last inefficiencies out don’t forget to look for efficiencies in places that already appear efficient on the surface. How much can you afford to spend on marketing? This simple little question hides some vexing issues that are worth exploring during rocky economic times.

Technorati Tags: ,

16 Responses to “PPC Averages can Hide Incremental Nightmares”
1. Ankur Mody says:

Hi Alan,

This is like advanced PPC from a statistician point of view. I feel you are posing a question and answering at the same time to all the PPC marketers that are involved in good spending over a period of time.

2. Hi Ankur,

We think these are important questions, but we don’t think the answer is the same for every company. Not every company will use PPC as a profit center, many will view this as an acquisition channel and are willing to lose some money on the first sale to capture lifetime value. Others use some component of PPC for branding.

There are many different, intelligent ways to build a business and we don’t pretend to be expert on all of them. We do think it’s important to keep an eye on the data as you’re doing it.

George

3. Very useful. I hope folks take the time to read and understand this.

I played around with the numbers on my own because I asked myself the question “why wouldn’t you do PPC using LTV?” (I’ve never worked anyplace, for what it’s worth, that didn’t use LTV when making marketing investment decisions.)

When you make some assumptions on the investment buckets–such as the multiple of sales generated over the lifetime of the top few buckets vs the lower buckets, and the incremental A/S required to generate that additional sales)–the curves become much more pronounced. The aggregate marketing income slopes up much more quickly, the maximum profitability is achieved a couple of buckets earlier, and the curve slopes down much more quickly once the maximum value is reached.

That just confirms to me that it’s even MORE important to try to do this (bucketize your investments) in troubled and all times.

Our CFOs want to believe the customers will keep returning without additional marketing investment. However, we know the truth and a few facts help you stay out of the “cut” marketing death spiral you’ll fall into without some solid facts to bring to Finance.

4. Very interesting and well written article. Thanks. Actually, I believe almost everyone in marketing understands that spending more does not necessarily bring you proportional value, however, you have presented a very good practical example on how to get this “idea” some statistical background.

Thanks again

5. Interesting idea, Mark, but I’m not sure I agree with your analysis.

I thought of this the following way. Let’s say the AOV is \$100 and each customer ads an incremental \$10 in value over the next 12 months (that is margin \$ less the cost to advertise to them — catalogs, search ads, whatever). The initial bins add many more new customers generating much more incremental value, but the extra value doesn’t cost you more. Your model assumes that you’d advertise deep into the hole on repeat purchases as well, which isn’t necessarily true.

Using my assumptions, the curve is much steeper at the front, but the break even point shifts \$20K to the right as a result of the incremental value driven.

I know a number of retailers have also found that LTV is dropping like a rock particularly in today’s economy. Some are taking the approach that LTV is no longer a given and as such each program, and perhaps each increment of each program needs to be cash positive.

Tough times!

6. Thanks for taking the time to look into my model and commenting back, George.

I re-checked my model to make sure there were no flaws (and found I’d dropped some values in the lower buckets–that had no impact on my assertions–so many thanks for helping me catch an error!) and found that I still believe the curve either maximizes for profit at around bucket 6. You’d implement incremental marketing efforts (at lower than acquisition A/S and higher ROI) in buckets 1-4 and then hold off after that.

I think the (slight, actually) difference between our points of view is in post-acquisition value and methodology differences (I didn’t model AOV in each bucket).

I assume the \$100K in sales in the top bucket is worth another \$80K over remaining life and that it slopes off pretty quickly to an incremental 5% for the bottom bucket–if you can be bothered to get it profitably, which I assume you can’t.

You indicate an incremental \$10 for each customer over the next 12 months on top of the AOV of \$100. Basically, just different assumptions.

I thought of the model from a continuity point of view, where I’ve seen the very best bucket drive another 80% (even a lot more) on top of initial acquisition sale for continuity products.

We both agree on LTV dropping. My recent work with some continuities has seen it manifest as higher churn rates, meaning you get fewer ARPU-generating months from the customer over the next X months.

I conceptually like the concept of making each effort stand on it’s own. However, I think removing LTV from the equation for most marketers is short-sighted. In the old direct mail world, from whence I came, that would mean going back to sweepstakes offers! Sweeps were great for a short ROI and cash flow kick and you didn’t have to worry about that pesky LTV–there wasn’t any. But when times got better and you wanted to resell them, darned if it wasn’t a pain in the neck!

Keep up the good work!

7. Thank you for your insights, Mark! It’s good to get different perspectives.

Check out what others are saying...
1. [...] PPC Averages can Hide Incremental Nightmares, Rimm Kaufman [...]

2. [...] PPC Averages can Hide Incremental Nightmares, Rimm Kaufman [...]

3. [...] you waste money on PPC, and so does George Michie from the Rimm Kaufman Group. He reminds us that average PPC ROI can hide diminishing returns at higher ad spend levels. High level stuff. Of course, to complicate things, an astute commenter pointed out that life-time [...]

4. [...] you like taking a dive into the uber-statistical, there’s a great post at the Rimm-Kaufman blog how spending more money on PPC doesn’t always bring proportionate value (diminishing returns, [...]

5. [...] you like taking a dive into the uber-statistical, there’s a great post at the Rimm-Kaufman blog how spending more money on PPC doesn’t always bring proportionate value (diminishing returns, [...]

6. [...] PPC Averages can Hide Incremental Nightmares [...]

7. [...] written fairly extensively on the topic of diminishing marginal returns in PPC, in catalog circulation…in most things in [...]

8. [...] Are the managers aiming at the right targets? [...]

9. [...] as their Search Engine Marketing (SEM) agency. In fact, RKG has also blogged on this topic – see RKG Blog for their view.  Here are some of our early [...]