We all know the money that we spend on Facebook and Google to drive performance and branding marketing activities. During the first few months of launch, we need to aggressively bring in traffic that will help in driving acquisitions. This is also very important for the first 18-24 months when we actively seek out in the digital space to build our customer base. Post which we might transfer this spend to the existing base in driving loyalty and repeatability.
The Customer Acquisition Cost (CAC) is really high during this 0-24 months window and in fact very expensive during the first 0-12 months. Reason being we have no clue as to whom should we acquire, what are the characteristics that best fit our Target Group, uncertainty on the quality of such acquisitions. All this, in turn, bring up a high Cost to Income Ratio (CIR). The goal is primarily having a closer to 0% CIR. In reality, it stays around >100% in the initial months and later hovers around 80-100%. The closer you bring it down to 0% the better your Target groups are responding to your digital marketing efforts and spend.
So the question is how you optimize your costs if you have High CIR’s.
Problem 1: Whom should we acquire?
To start off, from an analytics point of view, you need to take your Transactional data, Logistics data and website data.
- Identify the cities from where most of your orders are placed and the cities from where most of your orders are delivered.
- Identify the cities from where most of the cancellations, returns and Return to Origins happen.
- Identify the pin codes that are serviceable as well as not serviceable.
- Identify the delivery partners who have high delivery ratio and very low returns/cancellations or return to origin ratio.
- Identify the cities from where most of your traffic comes from. Also include the page views, transactions and conversion ratios as well.
Now that you have all these data points with you. Start with a scoring methodology that assigns weights to all the parameters identified above. Transaction is a metric that would be given at least 50% weightage. Once you build your algorithm around this. The cities with higher weightage would form the basis for both performance and branded marketing.
This would help in optimizing your spends by targeting only those cities that drive traffic to your site and as well contribute to significant conversions. Naturally your CIR’s fall. Good Job!!
This concludes our first problem, whom should we acquire. Geographically we are sorted out.
Problem 2: What are the characteristics that best fit our Target Group?
Considering we have data for the last 8-12 months, it will be good to start in understanding the characteristics of those who have transacted. This will help us in narrowing the Target Group with a greater affinity towards conversion. Greater the affinity, lesser the spends. Lesser the spends, better the CIR’s.
- Website data
Pull out the customers who have been acquired thru the paid marketing channels. Include the source and medium to have enough information about these customers along with other Geographic’s, demographics.
- Transactional data
Pick all available variables that pass thru test of correlation and overcome multicollinearity as well as have a good significant impact on the outcome. The outcome is identifying the characteristics of customers who transact on your platform acquired thru digital paid marketing.
Having done all the preliminary statistical checks, build a Single View of the Customer (SVoC) basis the above data. Run a classification algorithm to learn the characteristics of the customers. Using this you could define your Target Group for performance marketing.
Run marketing activities for 2-3 months to study the impact and optimize or fine tune your classification algorithm to have much better CIR’s than the ones observed thru Problem 1.
Go ahead and build a self-learning algorithm which would tune your algorithm with an increase in data volume and learnings of various campaigns/hypothesis.
This concludes our second problem, what are the characteristics that best fit our Target Group. Identifying the probable Target group is sorted out.
Problem 3: Uncertainty on the quality of such acquisitions
The chosen Target Group can be optimized further to a larger extent. Most of the acquisitions are discount driven, even the repeat ones are to an extent are discount driven. With very little margins in hand, driving revenue numbers is a hectic task. However, to be in the game, you need to continue your acquisitions as well as drive repeat rates.
Quality of customers is dependent on the Customer Lifetime Value (CLV). Customers with higher CLV are the ones that are going to help your business grow. Identifying these customers based on the CLV is the final step. Higher the CLV, better the Quality of Customers.
Calculate the CLV of customers that meet your business goals and superimpose on the Target Group identified in Problem 2. Now, this is your best Target Group. The one that is going to give you the best CIR. Keep tuning your algorithms as customer patterns change with time. Rethink your strategy and validate your hypothesis once in every 3 months.
What are you waiting for, ask your data science team to bring down your existing CIR’s. It is time to optimize your spends and acquire better customers who will help you grow your business.
Happy Spending, Happy Saving. Better Acquisition, Higher Retention.
Try deep learning using MATLAB