Attribution Part 7 - Marketing Value Analysis
Back in Part 3, I mentioned a quote which could have kicked off this whole attribution thing.
“Half of my marketing is wasted. I just don’t know which half.”
Even though there is no evidence that this was ever said by either John Wannamaker or Lord Leverhulme, it hits on the challenge faced by every marketer. The way this quote is worded however has sent us down the wrong path. By focussing on what is wasted, marketers have spent the last 100 years on the back foot justifying their spend rather than celebrating their value creation. Our justifications can easily be undone with a “yes, but have you considered…” So, we add that into the model only to have another “yes, but…” conversation at the next meeting.
The attribution models we have looked at so far were built for accountants. in addition to the weaknesses we mentioned with each model, they are all geared at measuring the past. We need a tool that will help us answer the big strategic questions like “what marketing budget do we need to hit our target” and tactical questions like “what is the best way to use this extra marketing budget we just found.” The question should be answering is not, can you find out where the marketing budget is being wasted? The question we should answer is what marketing activity is driving the most value to the business?
Marketing Value Analysis is a cohort analysis that looks at the value of different groups of customers. The customer groups are defined by the channels with which they interact over a fixed period of time. To explain this, let’s go through an example.
If you only used email and Instagram in your marketing strategy, you would have the following groups:
| Group # | ||
| 1 | ||
| 2 | X | |
| 3 | X | |
| 4 | X | X |
Groups 2 & 3 are those interacting with email and Instagram only. Group 4 is for those people interacting with both. Group 1 would be exposed to your brand advertising but have interacted with neither email nor Instagram. You can ignore this group or use them as your purchase control group.
The value delivered is the total sales for the members of each group over the set period of time. The numbers in each group and the value each group delivers, are in the table below. From these two data points we can calculate the average value for the members of each group.
| Group # | # of People | Total Sales | Average Value | ||
| 2 | X | 10,000 | £ 60,000 | £ 6.00 | |
| 3 | X | 10,000 | £ 50,000 | £ 5.00 | |
| 4 | X | X | 20,000 | £ 200,000 | £ 10.00 |
| Total | 40,000 | £ 310,000 | £ 7.75 |
Just this little bit of analysis gives us a some very useful information:
The average value not only tells you what the member of the group is worth. It also gives you the maximum you should spend to acquire new prospects into that channel. In this example, you should spend no more than £5 to acquire an Instagram follower if you want to break even in one period.
Email is the best single channel but the two together is 66% better than email only and 100% better than Instagram only.
Most importantly, you can use this data to calculate what you need to do to hit next year’s target. If we were set a revenue target of £350k, we just need to divide that by the average value of £7.75 and we see that we will need to increase our overall reach across these two channels to 45,162 people.
| Group # | # of People | Total Sales | Average Value | ||
| 2 | X | 10,000 | £ 60,000 | £ 6.00 | |
| 3 | X | 10,000 | £ 50,000 | £ 5.00 | |
| 4 | X | X | 20,000 | £ 200,000 | £ 10.00 |
| Total | 40,000 | £ 310,000 | £ 7.75 | ||
| Target | 45,162 | £ 350,000 | £ 7.75 |
The big caveat here is that you can find 5,162 more people that behave like the 40,000 you already have – same conversion rate, average spend, etc. If you cannot or do not think you can then we need to look at where we can maximise value by digging into the purchasers a little more. By adding the number of purchasers in each group and dividing that into the value we get the average value of a purchaser.
| Group # | # of People | Total Sales | Average Value | Purchasers | Avg Purchaser Value | ||
| 2 | X | 10,000 | £60,000 | £6 | 2,500 | £24 | |
| 3 | X | 10,000 | £50,000 | £5 | 2,500 | £20 | |
| 4 | X | X | 20,000 | £200,000 | £10 | 5,000 | £40 |
| Total | 40,000 | £310,000 | £7 | 10,000 | £31 |
If you can only choose one channel, choose email. It delivers £4 more than Instagram per purchaser.
You should put some of your budget towards converting people from either channel to both. How much of your total budget will be based on how many people you can add that fit the existing profile.
The difference between the purchaser value and the average value tells you the maximum you should discount for each group. If we look specifically at Group 2, the average value of a purchaser is £24 versus an overall average value of £6. We should not offer more than £18 in purchase incentives to convert a non-purchaser to a purchaser.
Remember all of this is over a standard period of time. That period is either your typical business cycle or your budgeting cycle. You want to make sure the time period is long enough to give you good data but short enough that long term changes in behaviour get lost.
This approach can also be used to measure the actual impact of external events. By splitting your group activity at a point in time and comparing the before and after, you will know the actual impact of that event. For example, you could look at the behaviours before and after the referendum and measure the impact of the political uncertainty created by that vote.
The challenge with all attribution models is that they can be easy and lead to bad decisions or more accurate which gives you better (not necessarily good) decision making but the models are harder to build, explain and use. However, they neither help you determine what to do going forward nor do they handle externalities. Marketing Value Analysis is relatively easy to build, easy to explain, handles external events and can be used as a planning tool in addition to reporting.