You’re a growth marketer with an analytical mind and a creative side. Each quarter, you lead content strategy, review creative content, allocate budgets, and plan inbound and outbound campaigns.
You’ve taken the first step toward higher customer quality measurement by using customer lifetime value (CLV) rather than the old ad spend (ROAS) metric. CLV measures customer value throughout all time, so it’s a clear way to score customer quality from acquisition to retention.
In this post, you’ll learn how to frame your marketing decisions to optimize for customer value. We do this by sharing a modeling language that will empower your thinking and conversations. Then we share tools and tips for posing helpful questions that are relevant to your business and your role.
Some of these ideas may sound familiar from your strategies using ROAS or revenue. (And using CLV will drive your decision quality even higher.) While we won’t dive into code (yet), this post strives to equip the analytical reader with a modeling mindset.
How to think about your marketing decisions
Every company is different, and your role within it is unique. So it would help to look at your own approach to decision making. Let’s break it down into three steps:
- Identify your decisions
- Prioritize your goals
- Track your signals
Having accounted for these, we’ll be ready to pose the right questions.
Identify your decisions
Let’s start with the decisions you make regularly. These could be (but are not limited to):
- Branding and messaging
- Ad platforms and budgets
- Channel touchpoints and frequency
- Campaigns and promotions
- Customer-level CAC caps
If you do not make these decisions above, then who does? If you use any third party vendors, consider how these improve your workflow and why.
Timing is important. Do you revisit your decision-making weekly, monthly, or quarterly? Think about your iteration cycle and the freedom to (or limitation from) critical thinking and evidence-gathering as you arrive at your decisions.
For example, your CMO may require approval prior to the launch of each campaign. This would likely add significant time to your iteration cycle, and flavor your reasoning to suit her palate. This is all well and good, but it’s important to recognize it for being a factor for success.
Prioritize your goals
Next come your goals and metrics, such as:
- # of conversions
- ad spend
- customer quality
- team size and reputation
- other company KPIs
Some of these will be hard requirements and others would add value to a varying degree.
A wonderful framework for goal articulation is the SMART system, where each goal should be: Specific, Measurable, Action-Oriented, Rewarding, and Trackable. These ideas will recur later in this post.
Break your goals down to their smallest granularity and think about how you measure success. Try to capture the units of measurement, such as count, dollars, or lifetime value. What are the possible outcomes, and what are the rewards for accomplishing them? (A reward can be infinitely negative if your job depends on it.) How does execution risk contribute to your thinking?
In an age of limited budgets, time, and personnel resources, marketers are asked to increase results with reduced spend. It’s more important than ever to strive toward goals you can act on immediately, not after months of collecting customer data.
Track your signals
By now, you must be thinking deeply about information flows in your department. Let’s refer to these as signals, and imagine your decision-making as a white box wired with signals. Some common signals are:
- Cost per acquisition (CPA)
- Return on Ad Spend (ROAS)
- Cohort or segments
Each of these can be measured to a varying degree of granularity, from customer-level to cohort-level to company-level.
Some signals are strongly tied to KPIs. Most signals are rooted in data. Consider your comfort level with data; how do you interface with it: yourself, from a direct report, or through a separate department?
It’s common to overlook and undervalue the importance of signals for decision making. Think carefully about how your signals drive goals, rewards, costs, and risks. Only trust signals that you know to be reliable and well suited to your task. If you haven’t already, question your trust in ROAS for measuring acquisition quality and consider CLV instead.
How to optimize your marketing decisions
By now we’re comfortable with our surroundings and keen to start modeling. But there’s an art to posing goals, and for that we need to define some vocabulary. Once posed we can recommend three methods for solving the problem to a reader with increasing levels of technical expertise.
Learn to speak the language.
In an optimization problem, the variable is an unknown quantity for which you seek a solution. Your decision options are known; what remains is to choose the best one (the optimal solution).
The objective measures outcomes in a way that depends on the variable. Outcomes should be tied to goals and combine rewards with costs and risks into one number. As before, a constraint is a hard requirement for a particular outcome.
Finally, parameters incorporate data and signals into your objective function. Parameters will change over time as you collect new data. Models and assumptions use data to learn parameters that capture the relationship between variables and outcomes.
As an example, suppose you need to decide how much to spend on campaign X for audiences A, B and C. Using the optimization language:
- variable: your budget for each audience
- objective: the CLV per acquisition, which increases (with diminishing returns) as ad spend increases
- constraint: your acquisition quota
- parameters: the trade-offs from paying to impress upon one audience over another
Pose the right question; solve it; repeat.
The beauty of this modeling language is how easy it becomes to pose a relevant question: what variable maximizes the objective subject to the constraints?
Here are three ways to arrive at an answer. (They increase in technical efficiency.)
Level 1: Use exhaustive search to assign an objective value to each variable outcome. This approach is good if you have direct experience to draw from and your resources are limited to back-of-the envelope calculations.
Level 2: Craft logic that maps inputs to outputs in order to capture the objective as a function of the variable. Coding is a very helpful skill here, as it allows you to simulate a large number of outcomes (including unforeseen ones). This approach is well suited for testing innovative ideas and managing risk at scale.
Level 3: Use mathematical modeling to capture all the relevant human-in-the-loop and data-driven parameters. This is very difficult to do well, and Retina can serve as a resource. There are also some open-source software packages that help you do this, such as cvxpy, lifetimes, and BTYDPlus.
There’s no silver bullet for optimal growth. With careful thinking and deliberate actions, you can lead marketing optimizations that drive measurable growth with lasting impact. Don’t forget to iterate frequently, collaborate with your team, and adjust as you go.
How this related to Data Science more broadly?
As our whirlwind tour winds down, let’s reflect on the difficulties. Your challenge lies in (i) how to model the objective as a function of the variable and (ii) how to learn parameters from data.
The former is the art of mathematical modeling. The latter is the science of learning from data. Retina AI is the only vendor in the customer valuation space with experts from both areas. Contact Retina at [email protected] to learn more.
In the next installment of this guide, we will provide some concrete examples of marketing decision optimizations in code.