From $60M in Aged Debt to 70% Email Open Rates: How Skit.ai Used AI for Debt Collections to Transform Recovery for a Global E-commerce Marketplace

How Skit.ai transformed a $60M post charge-off portfolio into a precision recovery operation using AI for debt collections for a global e-commerce marketplace.

The Company

A leading global e-commerce marketplace that connects buyers and sellers across 190+ markets worldwide operates at a significant scale. With hundreds of millions of active users and billions of transactions annually, even a small percentage of unresolved debt translates into substantial portfolio exposure.

Its collections challenge was specific to the nature of e-commerce debt: accounts with an average size of $2,300, aged between 3-4 years, and spread across a consumer base with highly varied language preferences and optimal contact windows. Traditional outreach approaches were consistently underperforming against this profile.

The Problem

A significant capacity gap meant a massive volume of aged retail debt was managed through manual outreach, where a small team struggled to consistently engage a large account base, leading to declining recovery on older accounts. Growth remained tied to headcount, with the existing team unable to effectively handle a 1.5M+ account base without scaling resources—an unsustainable approach. Outreach was further hindered by fragmented, resource-intensive efforts across isolated channels, preventing coordinated multi-channel campaigns at scale. At the same time, a shift in consumer preferences toward email and SMS over phone calls exposed gaps in the existing infrastructure, causing many reachable customers to be missed. Additionally, payment hesitation driven by misconceptions around lump-sum requirements discouraged a large segment of consumers from engaging altogether, suppressing overall recovery potential.

Context Building: Understanding the eBay Borrower

Before deploying a single campaign, Skit.ai’s Collections Intelligence engine contextualized the portfolio end to end — mapping engagement patterns by language preference, geography, optimal contact times, and past payment behaviour. Every design choice flowed from this foundation.

What the data showed

Dimension	Insight
Portfolio profile	Accounts averaging $2,300 in balance, aged 3–4 years, segmented by outstanding balance, debt age, and payment history.
Engagement pattern	Consumer contact windows and channel responsiveness varied significantly by language preference and geography — making uniform broadcast outreach ineffective.
Key behavioural barrier	Widespread assumption that only full lump-sum payment was accepted, suppressing engagement from consumers who could have paid in instalments.

Every insight became a campaign decision

What We Learned	What We Changed
Consumers ignore unknown phone calls	Shifted primary outreach to email and SMS; voice reserved for targeted escalation
Engagement varies by language and geography	Campaigns segmented and timed by consumer language preference and optimal contact window
Lump-sum assumption blocked engagement	Flexible installment options surfaced prominently and early in every outreach flow
Campaigns need to learn and improve	Campaign outcomes fed directly back into AI models for continuous performance improvement

The Solution

Skit.ai deployed a unified omnichannel debt collection management software built around three priorities: Precision Contact, Scaled Reach, and Resilient Engagement. The approach marked a shift from generic communication to a personalised model where each interaction was tailored based on existing consumer data.

Precision — Understanding the Consumer First

The Collections Intelligence engine modelled the portfolio on outstanding balance, debt age, and past payment behaviour. Engagement patterns were mapped by language preference and optimal contact times based on geography.

Outreach campaigns were highly targeted and timed to match consumer segment behaviour rather than being broadcast uniformly. Messaging was updated to prominently feature flexible installment plans, removing the assumption that full payment was required.

Scale — Reaching More Consumers, More Consistently

Strategy and outreach intensity were customised by debtor responsiveness. Each segment received a personalised, compliant approach via voice, SMS, and email — reaching more consumers, more consistently, without adding headcount.

A catalog of 23 high-performing email templates was built to cover key touchpoints across the customer journey. Campaign outcomes fed directly back into the AI models, enabling performance to improve with every cycle rather than remaining static.

Resilience — Higher Quality Conversations

Two specific strategies improved contact quality: identifying the best call times for each consumer segment via a multichannel approach increased connectivity by 19–20% among previously hesitant cohorts. Introducing flexible payment plan options at the right stage of the conversation converted hesitant consumers into paying ones.

The Results

What Actually Worked

Approach	What Changed	Outcome
Language & Timing Precision	Campaigns segmented by language preference and optimal contact windows before any outreach began.	Higher engagement across all consumer segments.
Flexibility Messaging	Installment options surfaced early and prominently in every outreach flow, removing the lump-sum assumption.	19–20% payment lift observed among previously hesitant cohorts.
Email Template Library	Built a catalog of 23 high-performing email templates covering key customer touchpoints.	70% email open rate achieved across campaigns.
AI Feedback Loop	Campaign outcomes fed directly back into AI models after every cycle, enabling continuous improvement.	Performance improved cycle-over-cycle without manual intervention.
Compliance Automation	Every interaction met regulatory requirements automatically — disclosures, frequency caps, consent management, and full audit trails. This automated collections software approach ensured compliance at scale while reducing operational overhead.	Compliance overhead removed from the operations team entirely.

To Summarize

This leading global e-commerce marketplace came in with a large, ageing retail debt portfolio and an outdated outreach model, reliant on phone calls, fragmented channels, and messaging misaligned with how consumers actually prefer to communicate. The result was low recovery on accounts with real resolution potential.

Skit.ai rebuilt its debt collection management strategy from scratch through portfolio modelling, multichannel precision, flexible payment messaging, and a reinforcement learning loop that improved with every campaign cycle. The outcome was a clear success — $770K resolved, $300K collected, a 19–20% payment lift among hesitant cohorts, and a 70% email open rate.