The whole point of this book came about after Markus had been pestering Steve for years, demanding respect for the scientific method and hypotheses and urging him to create measurable, repeatable laws for online advertising and metrics. Now we have one. It's simple, but damn effective.
The model is not like a law of science, it's more like driving a car. You put resources in (like petrol, water, oil) and you get the ability to move between point A & B. Of course to get between point A & B you have to follow the rules of the road like signals and traffic lights. So think of it like this, the petrol is your money, the car is your company or brand, the direction is what you want to do and the measurements you take are the signals and traffic lights.
A traffic light has three lights: a green one, a yellow one and a red one. In our model, the green one represents the campaigns that are performing within a set of key performance indicators (KPIs). Yellow ones are performing worse than optimal, but are still within the tolerance and although they require attention and changes, they are redeemable. Red ones are beyond the point of no return and should be stopped immediately.
The model: Green means go, yellow means slow down and look, red means stop.
There are enough books telling you how to set up a campaign, so we won't bother with that kind of approach. Using our method if you're a small player, you can save tens of thousands of dollars per year, but the big spenders could cash in millions.
So, the method: We'll give you plenty of examples, but we'll open it up first a bit. This is done in three stages: First we go through the history of what has been done, then we create the traffic lights for our system by understanding our operating environment and the third phase is nothing but execution and moderation.
Before you do anything, you have to define your goals and set measurable objectives. These have to be concrete like more leads, more purchases or more cost savings. The second thing is to go through your history and benchmark it across each customer lifecycle. You take all your campaigns for a single year (to take seasonal change into account) and benchmark your KPIs.
The big thing is to benchmark each lifecycle separately. They have different metrics and they (should) perform differently. This exercise is actually a good tool to help with lifecycle recognition. People are there to do something purposeful. That's why when trended over large numbers and over time they behave quite rationally.
At this point you should have a big pile of numbers sorted by each lifecycle phase.
Once you have the data gathered per lifecycle and an average performance within those numbers, you can start creating your own set of traffic lights. Congratulations. The first thing is to find all three groups: average (yellow), above average (green) and below average (red). The way we do this, is by using excel's percentile function - a great time saving function. If you're not a math wizard or otherwise a mere mortal, you can use our simple rule of 25/50/25, where you want to find the top and bottom 25%. The rest is your average performance. Why 25/50/25, you ask? We've tested hundreds of applications of this model and the worst 25% and best 25% are the best opportunities to optimise. So once you have a range of numbers (let's say they're in column a in your excel spreadsheet) and you want to find the bottom 25% of your range of numbers. The excel formula in the cell you choose would look like this;
The thing to remember here is do you want the bottom limit to be high or low? For instance with cost values you want the lowest range of numbers to be a green light and the high numbers to be your red light.
Richard has done 14 e-mail marketing campaigns for a particular lifecycle in the previous year. Most of his campaigns had a different average cost per customer acquisition and he wants to find out which ones were good and which were not. The campaign CPA's were $1, $2, $2, $4, $4, $7, $9, $12, $15, $15, $15, $18, $20, $25.
The model: Keep the top 25% and kill the bottom 25%. Learn from them all. Repeat.
We set the percentile function away and for the green lights we use 25% because we want a low number to the green lights. This means if he got a customer for $4 or less he's really performing well. Then we run the percentile function for the range of numbers in the top 75%. It means if he pays $15 or more he should stop the campaigns as they're too expensive.
So at this moment we have our data sorted and benchmarked, and we know the operational limits we are looking for. That means we can start putting things into action. Richard now knows that his best campaigns should have a CPA lower than $4 and he should stop everything that's above $15. Now he has clear numbers to look out for and he can do the same exercise for every lifecycle in his marketing palette. We can also set up the traffic lights for future campaigns, so we can react early. This allows us to kill the ones that suck and improve those that don't.
The lifecycle starts with a prospect and ends in a re-purchase from an old customer.
But how much is that customer worth to us? By just calculating our campaigns doesn't give us any real insight on whether we're doing the best job ever or just sucking like everyone else. Surely you know the value of a new client? We have to go backwards a bit. Richard runs a plumbing company and he has calculated (no he hasn't, he's a plumber) that a contact left in his web sites contact form is worth $25 to him. How did he calculate this? There are many ways, but the simplest is to figure out the average business earned from a client ($250 in Richard's case) and the number of contact forms sent together with the chance of getting the job (For every ten contacts sent through the form brings him one new client worth $250 on average.) That means Richard has a 10% success rate. But how many get to the form? How many people who click the advertisement contact Richard? Just divide the number of clicks by the number of forms sent. In Richard's case every 1 person out of a hundred also send the contact request.
Now we are approaching the core of the model and stage 3.
We could just look at the traffic lights and say that at least we are taking care of that. But the point is that the traffic lights automatically change as you get better at what you're doing. The process of optimising (stopping the red light campaigns from happening and improving the yellow light ones) means that your red light flag always shifts in the direction you want it to go. Your green lights become even better performers and the yellow lights shift in the same direction too. The other BIG thing is that you can't live in the CPC or CPA world, you need to live in the profit world.
You can't live in the CPC or CPA world, you need to live in the profit world.
If every person who sends a contact form is worth $25 to Richard, every person who lands on the site is worth $0.25, since 1% of them convert. That means cost per click under that sum is good business and the rest is wasting money. In his case, that would mean 3 out of 14 campaigns might have been a waste of money ($2.50 being the cost per acquisition that is profitable when 10% of forms processed become customers). So it means the limits need to drastically change. Just because we're optimising the average performance with the traffic lights doesn't mean average is good enough to be profitable. This applies even more when you scale up. A lot more.
Imagine Richard was the dude who's running Virgin, Richard Branson. His company would send out one campaign every day to different groups, which all amazingly had the exact same performance levels as the plumber Richard. It costs $4000 to create a campaign and $0.10 to send an e-mail and there are 100,000 recipients for each e-mail. If 3 out of every 14 of his campaigns were optimised so they fell into a specification the cost saving would be huge. You see where we are going with this?
Our method doesn't have a fixed performance that is good enough. We believe in continuous improvement. When you've done the math the first time around, you can start all over again and get rid of the worst 25%, improve the 50% and keep the top 25%. Rinse and repeat for great success. Instead of saying that your X dollar CPC is good enough, you're always trying to do better.
Eventually in theory you will drive to the plateau where you can't improve any of your KPIs without increasing your direct or indirect costs to the level where it isn't worth trying to improve any more. We've yet to see that happen in our combined 30 years of doing this, but if you have that problem it's a very positive problem to have.
If you liked this chapter, please recommend it to others.
"Data, data everywhere and yet all decisions from the gut!" That just about encapsulates why our marketing strategies are faith based, why our websites are barely functional ("the CEO loves purple!"), and why we are not making the types of profits we deserve. I love this book because Steve and Markus provide specific advice on how to unsuck our lives! Buy. Don't suck. Win.
Digital Marketing Evangelist - Google
Author - Web Analytics 2.0
In your face and a Must Read for beginner and expert analysts alike.
Founder - eMetrics Summit
Author - Social Media Metrics
Chairman - Digital Analytics Association
We have a single goal together, to make our customers a billion euros in profit. This won't happen in one year, it might take five years, but we will not stop until we have generated a billion of provable profit for our customers.
We love meeting and talking to new people! Get in touch with us. Come have a pint with us! We are available for projects, especially in warm places with great fishing opportunities and/or great beer.