LOS Pega — From Ping-Pong to One Clean DecisionLOS Pega — Dari Bolak-Balik ke Satu Keputusan Bersih
Redesigning the credit-decision engine for an OJK-regulated multifinance lenderMerombak mesin keputusan kredit untuk lender multifinance yang diatur OJK
"Faster and safer shouldn't be a trade-off." — the principle that guided the redesign.
.png)
Snapshot
| Role | Product Manager — owned the redesign end-to-end |
| Company | PT Sinarmas Multifinance — OJK-regulated multifinance lender |
| Timeline | 2025 |
| Team | Cross-functional — Product, Engineering, Credit, Risk & Compliance |
| Domain | Credit & Risk Systems (LOS / credit engine) |
TL;DR
Every loan application at Sinarmas runs through LOS Pega, the credit engine where Credit Analysts (CA) make the call. But before an analyst could even begin, applications flagged for a deviation (any exception to standard credit terms) bounced back to the sales team for a separate sign-off — a ping-pong that stretched end-to-end processing to as long as 7 days. Worse, the final decision sat with the same people carrying sales targets.
I redesigned the flow so deviation approval lives inside one approval chain, decision authority moved to Credit, and the manual checks got automated. End-to-end SLA dropped 7 → 3 days (-57%), credit-analysis-to-decision TAT 1–3 days → 1 day, First Installment Default ~6% → ~4%, and ~IDR 130B of stuck applications were unblocked.
This case study is less about the workflow tool and more about the product decisions: finding the real bottleneck, and choosing to move authority — not just automate.
1. Context
Sinarmas Multifinance is an OJK-regulated lender. LOS Pega sits at the heart of the lending funnel — every application passes through it on the way to a yes or no. The company already had analysts, SOPs, and an approval hierarchy. What it didn't have was a flow that let a good decision happen quickly without weakening control.
2. The Problem — "We weren't slow at deciding. We were slow at getting to the decision."
When I traced how an application actually moved, the same pattern showed up everywhere: - Deviation lived outside the credit decision. A flagged application was returned to Sales for deviation sign-off before the CA could analyze it — then came back for analysis. Two loops where there should have been one. - Manual, low-control checks. Blacklist screening was done "by eyeball"; there was no automated scoring. - Authority sat in the wrong place. The final call rested with Sales — the team that owns the sales KPI — so risk control and sales incentives pulled against each other.
"I don't spend my day deciding — I spend it waiting for the file to come back from Sales." — a Credit Analyst, on the back-and-forth
The real problem statement wasn't "make approvals faster." It was: "remove the hand-offs, and put the decision where the risk actually lives."
3. Discovery — Following one application end to end
Instead of starting from the tool, I followed real applications across the journey and sat with CAs and Sales. - The bottleneck wasn't analysis — it was the hand-offs. The actual credit assessment was fast; the waiting happened between teams. - The conflict of interest was structural, not personal. Good people made defensible calls — but asking the sales owner to also own the credit decision was a system design flaw, not a behavior problem. - Speed and control were being treated as a trade-off. Every past fix optimized one at the expense of the other. The opportunity was to win both at once.
4. The Strategic Bet — Fix the flow and the authority, not just the speed
The easy options were to add reminders, SLAs, or a status dashboard — all of which dress up the existing two-loop process. I argued against them as the primary bet. The real leverage was structural: 1. Collapse deviation into one approval chain so an application is decided once, in order, after analysis. 2. Move decision authority to Credit, separating the call from sales incentives. 3. Automate the highest-friction manual checks so analysts spend time on judgment, not lookups.
5. Prioritization & Trade-offs — Deciding what to change first
Decision 1 — Longest-chain approval over parallel tracks. When an application needs, say, Credit Dept Head but is flagged as a deviation requiring Credit Div Head, the chain routes straight to the highest required authority and runs after analysis — no mid-flow return to Sales. Trade-off: a slightly longer chain for flagged cases, in exchange for eliminating an entire rework loop. Worth it.
Decision 2 — Move authority to Credit, despite the change-management cost. This was the politically hardest change. I sequenced it deliberately and tied it to the risk outcomes leadership already cared about, so it landed as a control improvement, not a turf move.
Decision 3 — Automate blacklist first; defer scoring sophistication. Automated internal + external blacklist screening (plus an appeal mechanism and one-obligor exposure calculation) closed the biggest control gap immediately. A more sophisticated scoring layer stayed on the roadmap rather than blocking the release.
6. Execution — Shipping inside a regulated lender
- Risk and Compliance were partners, not gatekeepers. Because the change touched who decides credit, I brought them into the design — what's automated, what stays human, and where the guardrails sit.
- Change management was the real work. Moving authority out of Sales needed careful framing and rollout, not just a config change.
7. Impact
- End-to-end SLA: 7 → 3 days (-57%)
- Credit-analysis-to-decision TAT: 1–3 days → 1 day
- First Installment Default: ~6% → ~4% over the year
- ~IDR 130B of previously stuck applications unblocked
8. Reflections — What I'd carry forward
- The bottleneck is rarely the step everyone watches. It was the hand-offs between teams, not the analysis itself.
- Authority is a product decision. Aligning the decision with the risk owner drove as much impact as any automation.
- You don't have to trade speed for control. Framed right, the same redesign delivered both.
Prepared by Hamdan Fauzi — Product Manager. Some figures generalized for company confidentiality. Happy to walk through the full version in conversation.
"Cepat dan aman seharusnya bukan pilihan salah satu." — prinsip yang memandu perombakan ini.
Snapshot
| Peran | Product Manager — memimpin perombakan end-to-end |
| Perusahaan | PT Sinarmas Multifinance — lender multifinance yang diatur OJK |
| Timeline | 2025 |
| Tim | Lintas fungsi — Product, Engineering, Credit, Risk & Compliance |
| Domain | Sistem Kredit & Risiko (LOS / credit engine) |
TL;DR
Setiap pengajuan kredit di Sinarmas melewati LOS Pega, credit engine tempat Credit Analyst (CA) mengambil keputusan. Tapi sebelum analis bisa mulai, pengajuan yang terindikasi deviasi (penyimpangan dari syarat kredit standar) dikembalikan ke tim sales untuk persetujuan terpisah — bolak-balik yang membuat waktu proses end-to-end molor sampai 7 hari. Lebih parah lagi, keputusan akhir ada di tangan orang yang memikul target penjualan.
Saya merombak alurnya sehingga persetujuan deviasi berada di dalam satu chain approval, otoritas keputusan pindah ke Credit, dan pengecekan manual diotomatisasi. SLA end-to-end turun 7 → 3 hari (-57%), TAT analisa-sampai-keputusan 1–3 hari → 1 hari, First Installment Default ~6% → ~4%, dan ~Rp 130 miliar aplikasi tertahan ter-unblock.
Studi kasus ini lebih soal keputusan produk ketimbang tool-nya: menemukan bottleneck yang sebenarnya, dan memilih memindahkan otoritas — bukan sekadar otomasi.
1. Konteks
Sinarmas Multifinance adalah lender yang diatur OJK. LOS Pega berada di jantung funnel lending — setiap pengajuan melewatinya menuju ya atau tidak. Perusahaan sudah punya analis, SOP, dan hierarki persetujuan. Yang belum ada: alur yang memungkinkan keputusan baik terjadi cepat tanpa melemahkan kontrol.
2. Masalah — "Kami bukan lambat memutuskan. Kami lambat sampai ke keputusan."
Saat saya menelusuri pergerakan satu pengajuan, polanya konsisten di mana-mana: - Deviasi berada di luar keputusan kredit. Pengajuan yang terindikasi deviasi dikembalikan ke Sales sebelum CA bisa menganalisa — lalu balik lagi untuk dianalisa. Dua putaran, padahal harusnya satu. - Pengecekan manual & kontrol rendah. Screening blacklist dilakukan "by eyeball"; belum ada scoring otomatis. - Otoritas di tempat yang salah. Keputusan akhir ada di Sales — tim yang memegang KPI penjualan — sehingga kontrol risiko dan insentif penjualan saling tarik-menarik.
"Saya nggak menghabiskan hari buat memutuskan — saya menghabiskannya buat menunggu berkas balik dari Sales." — seorang Credit Analyst, soal bolak-balik itu
Rumusan masalah sebenarnya bukan "percepat persetujuan." Melainkan: "hilangkan hand-off, dan taruh keputusan di tempat risikonya benar-benar berada."
3. Discovery — Mengikuti satu pengajuan dari ujung ke ujung
Alih-alih mulai dari tool, saya mengikuti pengajuan nyata sepanjang journey dan duduk bareng CA serta Sales. - Bottleneck-nya bukan analisa — tapi hand-off. Asesmen kreditnya sendiri cepat; yang lama adalah menunggu antar-tim. - Konflik kepentingannya struktural, bukan personal. Orang-orangnya baik dan keputusannya bisa dipertanggungjawabkan — tapi meminta pemilik target penjualan juga memegang keputusan kredit adalah cacat desain sistem, bukan masalah perilaku. - Cepat dan kontrol diperlakukan sebagai trade-off. Setiap perbaikan sebelumnya mengorbankan salah satu. Peluangnya: menang di keduanya sekaligus.
4. Taruhan Strategis — Benahi alur dan otoritasnya, bukan cuma kecepatannya
Opsi gampangnya: tambah reminder, SLA, atau dashboard status — yang semuanya cuma mempercantik proses dua-putaran. Saya menolaknya sebagai taruhan utama. Leverage yang sebenarnya bersifat struktural: 1. Satukan deviasi ke dalam satu chain approval sehingga pengajuan diputus sekali, berurutan, setelah analisa. 2. Pindahkan otoritas keputusan ke Credit, memisahkannya dari insentif penjualan. 3. Otomatiskan pengecekan manual yang paling memakan waktu agar analis fokus pada judgment, bukan pencarian data.
5. Prioritas & Trade-off — Menentukan apa yang diubah lebih dulu
Keputusan 1 — Longest-chain approval, bukan jalur paralel. Bila pengajuan butuh sampai Credit Dept Head tapi terindikasi deviasi yang butuh sampai Credit Div Head, chain langsung diarahkan ke otoritas tertinggi yang diperlukan dan berjalan setelah analisa — tanpa kembali ke Sales di tengah jalan. Trade-off: chain sedikit lebih panjang untuk kasus deviasi, ditukar dengan hilangnya satu putaran rework. Sepadan.
Keputusan 2 — Pindahkan otoritas ke Credit, walau ada biaya change-management. Ini perubahan paling sulit secara politik. Saya urutkan dengan sengaja dan kaitkan ke outcome risiko yang sudah dipedulikan manajemen, sehingga mendarat sebagai perbaikan kontrol, bukan rebutan wewenang.
Keputusan 3 — Otomatiskan blacklist dulu; tunda kecanggihan scoring. Otomatisasi screening blacklist internal + eksternal (plus mekanisme banding dan perhitungan exposure one-obligor) langsung menutup celah kontrol terbesar. Layer scoring yang lebih canggih tetap di roadmap, bukan penghambat rilis.
6. Eksekusi — Merilis di dalam lender ter-regulasi
- Risk dan Compliance jadi mitra, bukan penghambat. Karena perubahan menyentuh siapa yang memutuskan kredit, saya libatkan mereka sejak desain — apa yang diotomasi, apa yang tetap manusia, dan di mana guardrail-nya.
- Change management adalah pekerjaan sesungguhnya. Memindahkan otoritas keluar dari Sales butuh framing dan rollout yang cermat, bukan sekadar ubah konfigurasi.
7. Dampak
- SLA end-to-end: 7 → 3 hari (-57%)
- TAT analisa-sampai-keputusan: 1–3 hari → 1 hari
- First Installment Default: ~6% → ~4% sepanjang tahun
- ~Rp 130 miliar aplikasi yang sebelumnya tertahan ter-unblock
8. Refleksi — Yang akan saya bawa terus
- Bottleneck jarang ada di langkah yang semua orang awasi. Ia ada di hand-off antar-tim, bukan di analisanya.
- Otoritas adalah keputusan produk. Menyelaraskan keputusan dengan pemilik risiko memberi dampak sebesar otomatisasi apa pun.
- Kamu tak harus menukar cepat dengan kontrol. Dengan framing yang benar, satu perombakan memberi keduanya.
Disusun oleh Hamdan Fauzi — Product Manager. Sebagian angka digeneralisasi demi kerahasiaan perusahaan. Dengan senang hati menjelaskan versi lengkapnya secara langsung.