Get startedGet started for free

Marketing examples

1. Marketing examples

This chapter focuses on making data-driven decisions in a marketing context.

2. 4 P's of marketing

Marketing is characterized by the 4 P's. Promotion, Pricing, Product, and Placement. Let's see how to leverage data to make decisions related to the first three.

3. To grow or not to grow

Often you must choose to grow your business through costly promotion or cheaply through word of mouth improving profitability. You can run ads on TV to get new customers but it's expensive. Or you can run no ads, and any new customers you get, by word of mouth, cost less to acquire making them more profitable. The trade-off is that TV ads reach more people so you will get more new customers quickly versus word of mouth.

4. Modeling customer behavior

Data scientists will build a customer propensity model providing a likelihood a potential customer will accept the offer. This lets you target the most likely respondents so your spend is efficient rather than blasting ads everywhere. For example, when you see an online ad banner, a model has scored your likelihood to click it, based on your browsing history.

5. Modeling customer behavior

If your search history included cute dogs, your banner may be from a pet store. The model is predicting your likelihood of a click based on your historical behavior.

6. Understanding a propensity function

When you build a customer propensity model the outcome is binary, classified as 0 or 1 based on inputs like income, age, or marital status among others. Rather than focusing on the outcome, it's the underlying probabilities between 0 and 1 that are insightful. This graph has probabilities between 0 and 1 on the x-axis. The y-axis shows a fictitious number of responses. Many people are not likely to respond because the majority of the curve lies below 0.50 on the x-axis. You should focus on the most likely responses in the blue box, representing the top 10% of likely respondents.

7. Identify the tail, most likely responses

A common mistake is to choose the most overall accurate model for customer propensity. If 95% of the time customers don't click on the banner, then the most accurate model would be one that predicts no-clicks! Accuracy doesn't help you focus on the most likely respondents to your offer. Since you have a limited budget, you need to be more targeted. Investigating how accurate the model is within the top 10% of likely respondents is more important than overall accuracy.

8. Model predictions by expected responses

This curve represents how many responses you can expect at each probability percentile, ranging from 0 to 100% of possible ads to purchase. If you focus on the first percentile, extremely low response probabilities, no one will respond, this is the 0-0 point in the graph. If you send ads to every person, all 100% you will spend a lot, but there is a maximum number of responses at the blue dot. Somewhere along this curve is the optimal number of ads to place based on the profit and expenses.

9. Return on investment

The orange point, where you show ads to the first 1%, when no one responds, means every dollar is wasted, your return is -1, each dollar spent yields 0 new revenue. At the green dot, you are spending a lot on ads, reaching many people but at that point, fewer are responding compared to higher percentiles. You have lots of responses with new revenue but the spend isn't efficient. To the right of the blue dot, you are very efficient but it's not an optimized return. This is because you have fixed costs. The cost of art designers whether you send one postcard or one million is the same. At this point, you are attracting new customers but the fixed costs deteriorate your return.

10. Costs and revenue

Costs include fixed and variable aspects. You have to spend on a designer for your ad no matter what, so it's fixed. However each new impression costs some amount so it goes up. Impressions are a variable cost. The blue line y-intercept is your total fixed cost and the slope is the variable cost. When you start with just the lowest 1% of likely respondents, you will have no responses, but you still have costs. Similarly, when you reach 100% of the prospects, at the blue dot the impression cost is more than the revenue.

11. Profit curve

Profit is the expected revenue for each successful ad, minus the fixed and variable costs. Anything above the break-even blue line means the ad campaign is a success. The right crossover is the last percentile where you aren't losing money. This means you have as many new customers as possible without losing money. The left-most crossover is the point your ad spend becomes profitable. You are reaching the minimal amount of people to make it worthwhile. A choice must be made between the crossovers balancing growth, and ad spend efficiency.

12. Off to it!

Now you have an understanding of customer propensities, let's apply that knowledge.

Create Your Free Account

or

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.