Determining How Much To Bid CPC on AdWords

Too many marketers are not sufficiently data-driven. They act as if they were engaging in the old world, where things were far less tangible due to lack of transparency and data. This is no longer an excuse. It doesn’t mean that you should always let cost-efficiency dictate your budget allocation – for some brands, it is absolutely necessary to get the visibility by bidding for unprofitable keywords (regardless of your attribution model). However, placing the right CPC bid is something most PPC books neglect and something that Google, Facebook, etc. has little interest in helping you with (they’ll be happy to suggest increased bids though).

Branding aside, many marketers are running their marketing on a tight budget, where acquisition cost in the form of an order (even re-purchase if you are sophisticated), an app install or other hard KPIs define the success or failure of the company – at least in the short term.

In such cases, I am surprised that I still meet marketers who look at irrelevant metrics such as whether or not an ad is displayed in a particular position on the SERP (front page vs. the second page or top 3 vs. top 8), or if the target audience of a display ad seems like to work or not. In the cynical world of performance marketing for the purpose of user acquisition, none of those things matter. What matters is the acquisition cost.

It doesn’t have to be complicated. Excel is your friend. For example, if you are an e-commerce company and know that you can afford a CPA of a certain amount, then it’s easy to find out how much you can afford to pay per click. Your cost per visitor should never be more than your gross profit per user. Depending on your business, you may even set the target lower. In that case, it’s straightforward. First, a couple of definitions for the calculations:

• Cost per visitor = CPC / (1 – dropout rate)
  I’ve gotten some questions on the dropout rate. The dropout rate is the percentage of people who click an ad but for various reasons don’t end up on your website. This is a particularly frequent scenario on certain platforms, e.g. Facebook
• Gross profit per visitor = Gross profit per order * Conversion rate

Now, the actual calculation. Remember that the highest cost per visitor is in the scenario where the cost of a visitor equates to the gross profit that the visitor brings.

• Cost per visitor = Gross profit per visitor
• CPC / (1 – dropout rate) = Gross profit per order * Conversion rate

As the gross profit per order in the breakeven scenario equates to the highest cost per order (CPO), the highest CPC you can afford to pay in a breakeven scenario is, therefore:

• Max CPC bid / (1 – dropout rate) = Gross profit per order * Conversion rate
• Max CPC bid = Max CPO * expected conversion rate * (1 – dropout rate)

In many cases, your dropout rate is 0%, which makes the conclusion even simpler. With significant data volumes, you will be able to predict your conversion rate on a micro level such as a particular keyword on Google or a particular banner on a given web property. The more granular you can be, the better.

You will obviously not know your expected conversion rate for a brand new marketing campaign, so you will have to make an educated guess and take the hit if it is too high.

Remember, this calculation gives you a ceiling to your spending, not necessarily a level to aim for. It should help you avoid stupid mistakes such as overpaying in order to stay on the first page of a search engine or hit a certain audience on Facebook. You may have other reasons to bid lower than the calculated Max CPC above. Such reasons include e.g. lack of cash/financing (in case your Max CPA is based on some deferred payment, e.g. customer lifetime value) or if the conditions of the auction indicate that you can get the same or roughly the same traffic at a much lower price (i.e. optimize your profit).

If you found this post insightful or useful, I suggest you read my post on controlling your spending with your bidding instead of budgets.

– Stefan