🛒 Part II: Going Beyond MarTech —

Propensity Score Matching to Decode Cart Abandonment cases

Devranjan Dash

🎯 Why Should Marketers Care?

In customer engagement and campaign planning, we often ask:

• “Did this loyalty offer actually increase purchases?”
• “Did our push notification/email/ WhatsApp boost retention?” or “Increased Conversion”
• “Which is this segment responding better to our engagement campaigns?”

The problem? Not all customers are the same to begin with. Some are more likely to engage or purchase regardless of your campaign.

🧪 Real-World Marketing Use Case:

Campaign: Push Notification for Abandoned Carts

• Group B didn’t ( 5% of the Base which qualified based on Abandoned Cart Event)
• Group A received the notification ( randomly set at 95% of the base which Qualifies based on Abandoned cart event)

📊 Conversion Breakdown

➕ Group A (Received Push Notification)

• 📩 95 users received the push
• ✅ 10 users converted
• 🎯 Conversion Rate: 10 / 95 = ~10.5%

➖ Group B (Did Not Receive Push Notification)

• 💤 5 users didn’t receive the push
• ✅ 1 converted
• 🎯 Conversion Rate: 1 / 5 = 20%

😬 Wait… 20% > 10.5%?

💡 Marketing Insight:

Just because a user converts after a push doesn’t mean the push caused it.
And just because someone converts without a push doesn’t mean you should have sent it.

• Campaigns must beat the natural conversion rate,
• Pushes should be sent strategically, not blindly.
• Was it worthwhile to send push notifications/ Email/ WhatsApp/SMS?
• Wouldn’t Group A users would have anyway converted as already more active… was it really the push that worked?

So is “ Spraying and Praying” a tactics at all to reduce Cart Abandonment?

Welcome Propensity Score Matching !

🧠 Propensity Score Matching (PSM):for Marketers

📌 What is Propensity Score Matching?
Propensity Score Matching is a statistical technique used to match individuals with similar characteristics across different groups, helping to target specific campaign, offer, or intervention.

Get Devranjan’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

In plain marketing terms:
👉 It’s how you make sure you’re comparing apples to apples — not apples to watermelons — when measuring customer behavior.

🧠 How It Works (In Simple Terms)

1. 🔍 Score each customer by how likely they are to convert (propensity)
2. 🧩 Match customers with similar scores across groups (like twin users)
3. 📈 Target users which are look-alike you isolate the effect of your marketing

The r-code for Analytics

Getting Started

Setting Av=bandoned Case as 1 and Converted as 0

The Propensity Matching

The Factors of Importance which form the Backbone for designing of the Content for Engagement.

The shaded Areas represent the Matched Audience and the Green area that is the unshaded part represents unmatched Audience

Now since we know who are More and Least Likely should the Engagement not be differentiated

• Most likely to convert has to be engaged differently ( Drive Immediacy i.e PN/ Exit banners/In-apps )
• While who is less likely should be engaged differently ( WhatsApp/Emails/Sms)

✅ What It Enables

• 🎯 Smarter Targeting: Segment based on Likelyhood or immediacy
• 💸 Better Budget Allocation: Invest where it truly moves the needle
• 🔄 Refined Campaign Strategy: Type of Content and Design of Engagement Journey.
• ✅ Personalized experiences based on behavioral likelihood, not assumptions

🔍 TL;DR for Marketers:
Propensity Score Matching filters the noise.
It ensures your engagement strategy isn’t driven by coincidence, but by real, attributable impact.
Think of it as your analytics truth serum.

Marketing is a Science to Get the Art Right , for getting the Science Right Use Data .