Summary. Xendit, a business-to-business payments company headquartered in Indonesia, reported a 50% reduction in average resolution time, a 112% increase in agent productivity over 14 months and CSAT of about 94% after replacing a spreadsheet-based support process with a connected, multi-channel helpdesk [1]. The case is useful less for the headline figures than for the sequence of changes behind them: remove manual data entry, shorten internal escalation, bring messaging channels into the ticket record, and measure each stage.

Source: Zendesk customer story [1]. All figures are vendor-published and unaudited.

About Xendit

Xendit lets businesses accept payments through cards, e-wallets, bank transfers and retail outlets via a single integration and API [1]. It describes itself as building payments infrastructure for Southeast Asia, a region where payment methods are fragmented and card-centric models do not translate well [2].

Its growth explains why support became a scaling problem. By the time of its $300 million Series D, led by Coatue and Insight Partners, it had raised a $150 million Series C at a billion-dollar valuation only nine months earlier [3]. In May 2022 it reported more than 3,000 customers, including Samsung Indonesia, GrabPay, ShopBack and Ninja Van Philippines, and annualised transactions that had grown from 65 million to 200 million within a year [2].

The challenge

Until 2017, Xendit tracked customer enquiries in spreadsheets and created tickets by hand [1]. Its Customer Success Manager described the workflow as manual end to end: an agent created the ticket, tracked it in a spreadsheet, then went back and forth with the product team to resolve it [1].

Three problems follow from that setup, and they compound as volume grows:

• Effort that adds no value. Agents spent time on administration (creating, copying, tracking) rather than on resolving the issue.
• Slow hand-offs. Resolution depended on coordination with product teams, with no structured path for escalation.
• No visibility. Xendit said it needed to record performance, capture feedback and analyse results, none of which a spreadsheet supports well [1].

What changed

After comparing several options, Xendit selected Zendesk, citing flexibility and ease of use, and reported that onboarding needed little vendor support [1]. The changes that the story links to its results fall into four groups:

WhatsApp: closing a tracking gap

Live chat was the largest source of tickets, but customers also asked for WhatsApp [1]. Before the WhatsApp API was connected through Zendesk’s Sunshine Conversations, agents copied customers’ WhatsApp messages into the ticketing platform by hand, and managers had no reliable way to track team performance on that channel [1].

The channel choice reflects the market. Roughly 91% of Indonesian internet users aged 16 and over use WhatsApp, the highest reach among the social and messaging platforms in the 2024 survey [7]. Technically, Zendesk documents that messages arriving through Sunshine Conversations channels become tickets in the Agent Workspace, with the same business rules and reporting as native channels [8]. That is the mechanism behind “less mishandling”: the conversation lives in the system that is measured.

Reported results

The published figures are as follows [1]:

• Resolution time: a 50% reduction in average resolution time since May 2020.
• Productivity: a 112% increase over 14 months, which the story attributes to insights from analytics, achieved while ticket volume spiked during the pandemic.
• Satisfaction: CSAT of 94.3% across an average of more than 10,000 tickets a month.
• Speed: average first reply time under 30 minutes.
• Self-service: a rise in help-centre views of 293%, described elsewhere on the same page as “almost tripled”.

Reading the numbers critically

The case is a vendor customer story, so the figures deserve the same scrutiny as any marketing claim. Five points limit how far they can be generalised:

• Relative changes without baselines. A 50% cut from a 48-hour average is very different from a 50% cut from a 4-hour average. The story gives neither starting nor ending resolution times.
• Undefined productivity. A 112% gain could mean tickets per agent, tickets per hour or something else. The metric is not defined, and a rise in volume handled per agent could coincide with changes in ticket mix.
• Concurrent events. The measurement window begins in May 2020, during the pandemic, when both volumes and customer behaviour shifted. No control group or before-and-after cohort comparison is reported, so the tooling cannot be isolated as the sole cause.
• Inconsistent wording. A 293% increase would mean views rose to about 3.9 times the original level, whereas “almost tripled” implies roughly 2.9 times. The page does not reconcile the two, so this figure is the least reliable of the set.
• Missing outcomes. The story reports no cost per ticket, retention, churn or revenue effect, so it does not show a financial return.

How the numbers compare with industry data

Support benchmarks are inconsistent across vendors: definitions of first response time, treatment of business hours and whether automated replies count all differ, and some benchmarks report medians while Xendit reports averages [6]. The comparison below is therefore indicative only.

Reading: Xendit’s reported first reply time is faster than the all-industry ticket median, but it is an average from a different dataset and period, and a fast first reply does not by itself indicate fast resolution. Resolution time and re-contact rate need to be read alongside it.

Xplere interpretation for fintech teams

Whatever the exact size of the gains, the case shows that service outcomes are largely determined by workflow design. In a payments business, customers contact support about failed transactions, settlement delays and integration problems, where unresolved issues carry direct commercial cost for the customer. Reviewing the path from an incoming request to resolution tends to expose avoidable effort, including repeated data entry, delay in assigning an owner and time spent waiting for another team.

Measuring those stages separately gives a fintech a clearer basis for deciding what to fix first:

Three practical points follow from the case and the caveats above:

• Set a baseline before changing anything. Record absolute times and the metric definitions, so that later improvements can be stated with confidence rather than as unanchored percentages.
• Bring every customer channel into the measured record. Xendit’s WhatsApp gap shows that an unmeasured channel hides performance and increases mishandling.
• Report medians and tail percentiles, not only averages. Averages can conceal the slowest cases, which are often the ones customers remember.