Predicted to Purchase Product: Propensity v2

Predicted to Purchase Product Propensity v2

We talk to a lot of retailers. We mean…a lot. They all have some form of product propensity. Some of them are basic historical models, others have some genuine predictive analytics, and fewer of them are still ML-driven.

It’s not the methodology in need of the upgrade,

it’s the use case

The funny thing is, though – methodology doesn’t matter as it’s not the path to the answer – but the question being asked that is due for an upgrade.

Our use case LTV, not a transaction, not a recommendation.

Huh, what??? It’s odd, so let me explain. Most product affinity use cases begin and end with a product your shoppers are most likely to buy. Our angle is unique in that it includes further “qualifying questions.”

What product are you likely to buy that:

  • Is a product/product line you have already purchased (continuity)
  • Is part of a “product theme” you are already purchasing in (continuity/expansion)
  • Is a product line that is “new to you” product (expansion)
  • Will lead to the most downstream purchases (Journey Graphing)

You see, the above isn’t really about a product but speaks to the nature of the relationship a customer has with the product lines…it sets the stage for customer optimization and conversational marketing.

So, what else makes our approach different…read on.

We predict the entire catalog – every product line

Z72eDEA9yHZCPLv guFh0Nc mJNgJh9uhk77QXudhWV774kVjaQobvwi6QcnoF0qOyGQgcbEjYj4L3ScKStD7nmYQA0E 1cJEzuWW2tmsaDtLctxHZqSdZPsZ98KCqos4TKS qOi=s0
Line level 30-day predicted purchase w/ (300k prospects)

Why? From recognizing prospect opportunities (expansion) to managing a customer shift of spend between lines, to the basics of prioritizing email participation – without a prediction against the entire catalog – you have an incomplete picture.

We predict categories and product themes

Managing a customer relationship means maximizing the value you create for her within the product lines you carry. Our approach allows you to optimize her value not just in the Nybok shoes product line but across the entire product category/theme (running).

Modeling the entire product catalog allows you to understand the customers’ shifting relationship with product categories and themes.

We predict product thematic based on the global product taxonomies we create with our clients, allowing us to shift with her product interests and manage the relationship when it’s at its most volatile.

We predict every customer identity in the DB

With few exceptions, all customers receive a prediction. From digital anons to in-store churned – scoring the entire customer community is essential for optimizing initial conversion through realizing winback opportunities.

We train and classify on cohorts

Meaningfully different audiences will have meaningfully different predictions. Our models will train/predict against discrete audience populations to create a more accurate result. We also bin scores to create High to Low distributions (more marketer-friendly). That binning is done against cohorts, oftentimes focused on the lifecycle phase (newly acquired, churn risk levels, attributed).

We predict against multiple date ranges into the future

We create date range-based predictions based on your business. If it’s grocery, your prediction windows may be in the next 3 days, the next week, and the next 2 weeks. If your apparel, your prediction window may be the next 30, 60, or 90 days.

Why? Predicting the future allows us to respond to it – both strategically and tactically.

We predict future seasons

For Open clients, seasons are your best weapon for maintaining and expanding customer value (we have around 40+ programs built around this). Propensity starts with the system identifying seasonal products. No, not products purchased during the season, but how the season impacts the products purchased. Note: The difference is critical.

Next is the prediction itself. Depending on the season we can identify a shopper’s likelihood to transact in early vs. deep season vs. EOS. This gives you a sense of “who” you are dealing with.

Finally, is the management of the seasonal lifecycle. Predicting ahead of the seasonal allows us to “pretarget” customer with advertising that secures “share-of-mind.” Identifying customers who lapsed this season allows us to make your EOSS (End of Season Sale) a bit more strategic.

In Closing: Propensity v2

Again, propensity v2 really isn’t about being descriptive for merchandising and prescriptive to your business. It’s the strategic alternative to the question, “What products are a customer most likely to purchase right now.” It’s about having a propensity model(s) focused on the business of optimizing the relationship between customers, products and our retail brand. It’s about

Leave a Reply

Your email address will not be published. Required fields are marked *