Case Study - How Bids & Algorithms made AdWords Profitable for Moby Memory

Situation

Moby Memory is an e-commerce store that sells memory cards and other electronics. Margins are very low, and we're up against big boys like Amazon and Ebay. A true David vs. Goliath story. Luckily, it takes more than a fat wallet to win at AdWords.

This is an ecommerce store that makes many sales every day at low transaction value and at low margin. The one thing we have in our favor is data. Many sales = lots of conversion data to feed our decision making and our algorithms.


Adapting Base Bids to Sale Values

A very important campaign for Moby (and for e-commerce stores in general) is the shopping campaign (previously known as PLAs). Bids are an important factor that impact volume via impression share (how often your ads show) and also ad position (how prominent your ad is).

In a store like Moby Memory, products sell for prices anywhere from £5 to £500. If we consider a margin of 10%, that means that the contribution (revenue * margin) ranges from £0.50 to £50. Assuming a conversion rate of 5%, that means we can break even with £0.025 CPC on one extreme of the spectrum and £2.50 CPC in another. What I'm trying to demonstrate here is that bids should vary drastically depending on which product you're listing.

If you just starting out with setting bids for a campaign, you can use the following basic formula to determine starting max CPCs:

Item Sale Price * Margin * Expected Conversion Rate = Starting Max CPC

For example, if you're selling a tablet for £200, and your margin is 10%, and your expected conversion rate is 5% (not uncommon in ecommerce in this industry - your industry might have higher margins and lower average conversion rates), you could set your max bid to £1.00. 5% conversion rate means 1 out of 20 clicks turn into a sale, so we need 20 clicks at £1 = £20 to make a sale = 10% of the £200 tablet which is our break even point.

Because your max CPC won't actually be your effective CPC (it will be lower), you could probably raise your bids to £1.30 and hit your intial target of £1.00 effective CPC.


Using "Conv. Value / Cost" to Adjust Bids

But with conversion data via e-commerce tracking, we can do something far better. We can look at how much revenue an item actually brought in compared to it's cost. This is much better because in reality, people sometimes end up buying another product, or multiple products.

Inside AdWords this is called Conv. value / cost. This is a metric that you need to go and enable in customize columns:

alt

To work out whether a segment is profitable you need to have a break-even point in mind. Here's how you find that:

Break Even "Conversion Value / Cost" = 1 / Margin

For example, if our margin is 5%, our break even conversion value / cost is 1/.05 = £20. This means that anything over £20 cv/c is profitable and anything under is unprofitable.

From here it's easy - increase bids with cv/c's over £20 and decrease bids with cv/c's under £20. You could of course go further with Excel and have variable margins per SKU and so on, but this is already a good start. If you want to take it a bit further you can look at your range of cv/c's and make tiers. For example, if we have cv/c's varying from £5 (very unprofitable) to £50 (very profitable) we could make the following changes:

Tier CV/C Range Proposed Bid Modification
Tier 1 £0 - £10 -30% Bids
Tier 2 £10 - £18 -15% Bids
Tier 3 £18 - £23 No Bid Modification
Tier 4 £23 - £30 +15% Bids
Tier 5 £30 - £50 +30% Bids

We could repeat an analysis like this every month or so to keep adjusting our bids - how often you're able to adjust is dependant on how much conversion data you get.


Base Bids and Bid Modifier Algorithms

The use of our algorithms, optimizing for relative performance of time of week and geographic regions, combined with Google's ECPC (enhanced cost per click) led to drastically positive results for Moby. Here's a closeup where the impact is very visible.

alt

At first glance, overall CPC decreased while volume increased.

Here's a closer look - clicks up 40%, cost pretty much stable. That's like getting 40% more clicks free!

alt

But also, we're getting clicks that Altus and Google thinks is likely to convert into a sale. And here's the impact: Revenue up 95% and ROI up 72%.

alt

alt

Revenue from Non-AdWords traffic was actually down -13.27% so this is unlikely to be a market trend or any other external factor.

The account has maintained strong ROI since.


Increasing Conversion Rate and CLV

Of course bids are only one battle. To win the war we need more. With Moby, we used Dynamic Remarketing (where you display products that a visitor has viewed but hasn't purchase) to strengthen conversion rate. We also split tested product pages and the checkout process which combined increased their overall conversion rate by close to 40%.

The next step is to increase their CLV to be disproportionate relative to their industry eg. via partner deals and revenue share agreements (for example people who buy SD cards have a good chance of being interested in one of our other client's products - Photography Education! Expert Photography Case Study). All of a sudden, an SD card purchase of £0.50 contribution (revenue * margin) is worth more a lot more than £0.50 if there's an X% chance that we get a £Y commission on a parter's cross sell after that purchase. And all of a sudden, we can afford to bid more than our competition...