How can marketers maximise sales conversions?
At any point in time, companies may have hundreds, thousands or even millions of prospects who are interacting with their marketing but whose current journeys have not yet resulted in sale.
In order to convert as many prospects as possible, marketers need to know what steps to take and how much to invest in them. This information needs to be immediately available so that the right actions can be taken at the right time.
How do you determine the most cost-effective actions to take?
Marketers have a wide range of potential tools that can be applied to assist conversions. These include email, direct mail, SMS, phone calls, PPC, display and so on.
All of these activities have a cost. Sometimes that cost doesn’t justify the end result—and sometimes there is no need to do anything.
The overall objective is clear: optimise spend to maximise the number of sales. To do this you need to understand the impact that your actions have at an individual level. In other words, how much would each action improve the chance of a successful outcome.
What is individual level data?
Creating an individual level view of how prospects engage with marketing has become easier with the use of intelligent digital data collection, connection and management. More and more companies are now collecting all touchpoints in the customer journey, both online and offline.
This individual level data is the first step to unlocking actionable insights. The next step is attribution.
What is attribution? What can it achieve?
Individual level data is frequently used to determine the attribution of different channels i.e. the contribution each event has made to the sale.
Attribution gives powerful insights into the relative importance of different channels and how marketing return on investment can be optimised by evolving the spending mix.
Theses insights can be generated at channel or campaign level, and for different segments or groups of prospects.
What are the limitations of attribution?
Attribution is powerful, but it doesn’t tell you the specific action that you should be taking right now to try and convert a prospect. This needs a slightly different question to be asked – and some sophisticated modelling.
How can you evaluate the impact of each potential action?
Choosing the best action to take is simple in principle.
For each prospect, simulate all available actions one at a time and evaluate how much uplift they give to the chance of success. Then compare these uplifts against cost to determine the best action or set of actions to maximise conversion.
However, accurately estimating these uplifts is (inevitably!) less straightforward. It all depends on the model you use.
What should I look for in a predictive model?
The model you use to estimate uplifts needs to be able to:
- Measure the incremental uplift in the likelihood of success
- Take account of events and actions that have already taken place
- Factor in the time and order in which these actions occurred
- Predict the likely outcome if no further actions occur
Traditional regression modelling techniques are ill equipped to handle the complexity of these calculations. Machine-learning algorithms are much better suited to these types of problems.
Even with powerful machine-learning solutions, though, the model design is critical to ensure that the uplifts can be predicted accurately—in particular for simulated journeys that may never have been experienced before.
To find out more about attribution, download our data-driven attribution brochure.