MT362 Unit 2 AI Use Case Assessment Example

Problem Before Technology: An MT362 Unit 2 AI Use Case Assessment for a Pet Chain's Subscription Cancellations

[Student Name]

Purdue University Global

MT362: Artificial Intelligence Applications for the Marketing Professional

Unit 2 AI Use Case Assessment

[Instructor Name]

[Date]

The chain, survey, quotes and estimates are composite; each figure is labeled as quoted, estimated or assumed, and the research cited is real.

The Business Problem

Fernbrook Pet is a composite chain of 43 pet supply stores in California and Oregon. Its Repeat Delivery program ships food and supplies to about 71,400 subscribers on a schedule they set. About 4,200 subscribers cancel each month. The outcome the chain needs is fewer cancellations net of whatever it spends to prevent them. That outcome, not any technology, is the standard every option below is held to. Davenport (2018) argues that companies get the most from artificial intelligence when they begin with a specific business problem rather than with the technology, and this assessment follows that order.

Why Customers Leave

The chain's exit survey asks every canceling subscriber for one main reason. Over the past year, 2,860 cancelers answered. Food piling up at home was the most common reason, at 31 percent. A pet that had died or been rehomed accounted for 19 percent. Price accounted for 17 percent, a switch to another retailer for 14 percent, a vet-prescribed diet the chain does not carry for 11 percent and other reasons for 8 percent. The survey is answered by about two thirds of cancelers, so it may undercount customers who leave without explanation, but its pattern has held steady across four quarters.

Six Reasons, Sorted by What Changes Them

The reasons fall into three groups, and sorting them does most of the analytical work.

Fixable by process: food piling up. A subscriber whose food arrives faster than the pet eats it does not need a discount; the subscriber needs a better schedule. Together with the prescription diet gap, which a merchandising decision could address, this group accounts for 42 percent of cancellations.

Outside marketing's reach: a pet's death or rehoming. No offer should be sent to these subscribers, and any system that sends one would cost goodwill no retention figure could recover. This group accounts for 19 percent.

Responsive to an offer: price and switching to a competitor, 31 percent together. These are the only cancellations a well-timed offer might change, and so the only ones where predicting who is about to leave could help.

A churn model can only save the customers an offer could keep, and at this chain that is about 31 percent of those who leave.

Food Piling Up Is a Scheduling Problem

A 15-pound bag of food for a 12-pound terrier lasts far longer than the program's default four-week cadence assumes, so the bags accumulate in a closet until the subscriber cancels. A schedule calculator using the bag size and the pet's weight, both of which subscribers already enter, would set the delivery interval at sign-up and suggest a change whenever a subscriber edits the order. The web team estimates about three weeks of work, roughly 120 hours, with no vendor involved. If it prevented even a third of schedule-related cancellations, it would keep about 430 subscribers a month, more than any estimate for the model.

Pet Loss Is Excluded From Every Option

Subscribers who cancel citing a pet's death receive no save offer under any option. This is written as a design rule rather than left to a model's judgment, because no prediction should decide whether a grieving customer receives a discount.

Three Options on the Same Cost Basis

Each option is laid out by first-year cost in dollars and analyst hours, the data it needs and the share of cancellations it could plausibly reach.

The schedule calculator costs about 120 hours of internal web work and no vendor fee. It needs bag size and pet weight, already collected. It could reach the 31 percent of cancellations caused by food piling up.

A skip rule, offering a one-time skip whenever a subscriber edits the schedule, costs about 40 hours to configure in the existing email platform. It needs only schedule edits. It reaches the same schedule-related group, plus some price-sensitive subscribers who would rather skip than cancel.

The vendor's churn model costs $48,000 a year as quoted, about 220 analyst hours as estimated to connect order and schedule data, and an assumed discount spend on subscribers it flags. It needs order history, schedule edits and possibly more, which the Unit 3 data inventory will review. Its reach is capped at the 31 percent of cancellations driven by price and competition, and inside that cap Ascarza (2018) warns that targeting by risk alone can spend offers on customers whose decisions the offer does not change.

What Could Go Wrong With Each Option

Each option carries its own risk, and the assessment states them so the cheap options are not mistaken for safe ones. The calculator depends on pet weights that many subscribers entered when their animals were young and never updated; a puppy's weight would set a schedule far too slow for the grown dog, so the calculator must ask for a current weight before suggesting a change. The skip rule could teach subscribers to edit their schedules just to see an offer, the same lesson the Unit 1 post worried about, though a skip costs the chain far less than a discount. The model carries the largest risks: a contract signed before its value is known, customer data shared with a vendor under terms not yet reviewed and discounts that may flow to customers who were never going to leave. None of these risks rules an option out, but each one shapes how it should be tested.

What the Model Might Be Worth

If the model saved between 5 and 15 percent of the roughly 1,300 price and competitor cancellations each month, an assumed range, it would keep 65 to 195 subscribers a month. At an estimated $120 in margin from each retained subscriber, that is $93,600 to $280,800 a year before the discounts it spends and its fee. Lemmens and Gupta (2020) show that retention programs should be judged by the profit they produce after the cost of the incentives, not by how accurately the model predicts, and the range above is too wide to justify a contract without a test.

Recommendation

Sequence the work, cheapest first. Build the schedule calculator and the pet-loss suppression now. Add the skip rule next. Carry the churn model forward only as one arm of a controlled pilot, compared against the simpler rule, where its added value can be measured rather than promised. The vendor should be told that the chain will decide after the pilot, not before.

References

Ascarza, E. (2018). Retention futility: Targeting high-risk customers might be ineffective. Journal of Marketing Research, 55(1), 80-98. https://doi.org/10.1509/jmr.16.0163

Davenport, T. H. (2018). The AI advantage: How to put the artificial intelligence revolution to work. MIT Press.

Lemmens, A., & Gupta, S. (2020). Managing churn to maximize profits. Marketing Science, 39(5), 956-973. https://doi.org/10.1287/mksc.2020.1229