A4U Expo Live Blogging: The Digital Mix and Point of Diminishing Return w/ David McDermott
David is going to talk about statistical analysis and attribution. He is the performance director for 7 Things Media.
What is Diminishing Return?
David talks about when does profit disproportionately falls in relation to spend. Is it still worthwhile investing in. He was recently asked by his client about this and faced with the dilema of cutting their budget on affiliate marketing.
Multi Attribution Modelling?
David shows a slide from Google Analytics and Multi Channel Funnels. The different touch points people can now review is becoming more important.
Why is it important?
We need to weigh up the value of the whole funnel and the differences in reach and performance. Every change affects the output of a campaign. Its something David looks at closely. He goes on to talk about the recession and why people are looking more closely at where they invest their budget.
As a result planning models needed to be more concise. David looks at micro and macro level models and activities. This gives him a more clear picture of what provides value and how to optimise his spend.
No Definitive Solutions
There are no plugins or tools to work this out. It involves mining data and using lots of complex mathematics. Different modelling might provide a more optimal result and is something David is looking into.
What Can We Use to Measure It?
Data + Maths + Technology
This gives us the information we need to make correct decisions. Historical data is key however and the more data you have the better.
The Monte Carlo Simulation
This is where continual simulations of random scenarios are run until a pattern emerges. This process often outputs a train line like pattern from which you can pull data from.
How it Works: CPA v Spend
David shows a graph of the simulation results which gives you an idea of the lowest CPA you would get from a particular channel. From the line you can develop from the graph you can also see the point of diminishing returns. This is where the level of sales no longer grows or limited growth is achieved after a certain spend.
Plot Optimal Spend Ratios
From this graph you can then put together a spend ratio graph.
Limitations of Model
Large level of data required. If you don’t have the data you can’t compile this.
** David goes on to provide examples on how the model is used against real life clients **
– Search generally provides the lowest CPA when CPA increases
– Channels quickly change efficiencies at specific points of spend levels
Difficult session to blog without slides but i will add in a link to the slides once available.