GROUND — Ground Truth · Indonesian SME Credit NPL 2026

OJK Portal Data Sektor Jasa Keuangan Terintegrasi Small + Medium only · micro excluded May 2026 · Actual, not estimate

Of Indonesia's small & medium business credit —

0.00%

— has stopped performing. Rp 42.01 trillion of sour loans on Rp 848.4 trillion outstanding, measured from the regulator's own cube.

NPL = kurang lancar + diragukan + macet. Chain of actuals: 4.72 (Jun-25) → 4.85 (Jul-25) → 4.91 · 4.93 · 4.83 · 4.86 · 4.95 (Jan–Mei 26). Validated against OJK's published aggregates to rounding.

Remove the micro segment and the number gets worse, not better — micro is currently the cleanest tier. The problem sits with medium-sized enterprises, at 5.73% — the worst reading since July 2023 — and never below 4.8% since 2021.

§1

How the number is made

SPI Tabel 3.21.a → portal cube

STEP 01

Split the book

Start from SPI Tabel 3.21.a — the only table splitting UMKM credit into micro / small / medium, with both outstanding balance and NPL, per bank group.

STEP 02

Sum the groups

Add state banks, regional development banks, national private banks and foreign branches into one national figure per segment.

STEP 03

Drop micro

The brief is small and medium only. Micro rows out.

STEP 04

Weight, don't average

Divide total bad loans by total outstanding. The small book is ~1.5× the medium book — a simple average of percentages would give the wrong answer.

VALIDATION — rebuilt totals vs SPI printed row, Jun-2025: Rp 1,503,625 mrd = 1,503,625 NPL, Jun-2025: 66,313 vs 66,314 (one miliar of rounding) 2026 cube vs OJK press, all-UMKM with micro included — the only cut OJK publishes: 4.59/4.67/4.59/4.62/4.68 ≈ 4.60/4.68/4.60/4.62/4.68, max deviation 1.1 bps. Our small+medium headline excludes micro and has no public comparator.

§2

Five years of the series

2021 → May 2026

NPL % — small + medium, micro excluded

SME (K+M) Menengah Kecil Mikro (ref)

Medium is the persistent problem. Worst segment in every period since 2021, never below 4.8%, and 5.73% in May 2026 — its worst reading since July 2023. At the last SPI print, June 2025, about Rp 215 trillion of a then Rp 332 trillion medium book — 65% — sat with national private banks, in larger single-borrower loans exposed to construction, wholesale trade and manufacturing. No bank-group split has been published since; the medium book itself is now Rp 337.4 trillion.

Micro, though excluded, is the one to watch. It nearly doubled — 2.18% in 2021 to 4.32% now — as KUR loans age and pandemic-era restructuring winds down. Still the cleanest tier, but on its current path it crosses the small book. The gap micro creates in this report's headline is narrowing because micro keeps deteriorating. That is itself a signal.

§3

The micro paradox

May 2026 · per tier

BD = baki debet, outstanding balance, May 2026. Kecil Rp 511.0 T · Menengah Rp 337.4 T — the medium book shrank month-on-month — · Mikro Rp 664.8 T. Non-UMKM corporate credit runs 1.64% for scale.

Interactive — one filter, one finding

Remove micro

4.68%

All-UMKM NPL, May 2026. Micro's Rp 665 trillion included.

§5

The big-book league

Books ≥ Rp 30 T · Mei 2026

Ranked across all 34 provinces, Jawa Barat comes 6th worst — but everything above it is small. Maluku, at the top of the table, holds Rp 2.8 trillion; Jabar holds Rp 88.9 trillion. Limit the comparison to SME books of Rp 30 trillion or more, and Jawa Barat stands alone at the top, 110 basis points clear of the next big book. It holds 10.5% of the country's SME loans and 13.8% of its SME bad loans.

§6

Momentum & the full ranking

Jan → Mei 2026

Big books, five months

worseninghealing

Jawa Barat worsened 45 basis points — the only big book deteriorating materially. In money: its stock of bad SME loans grew Rp 574 miliar in five months, 5,243 → 5,817 miliar, with the medium segment doing the damage (7.51% → 8.61%).

DKI is down 36 points for the year to date — but it gave a quarter of that back in May, after April printed a 41-month series low. Across those 41 months DKI has never left a 4.34–5.20 band: it is not healing, it is range-bound. Jateng moved +3 and Jatim +10 — for books that size, close to standing still. Banten rose 88 points, still below national but the fastest riser in Java; worth checking again in three months.

Maluku tops the whole table at 9.32% — on a Rp 2.8 trillion book, where seventy per cent of the move came from a single Rp 0.6 trillion medium-enterprise portfolio. That is a concentration event, not a provincial trend. The meaningful small-province readings are Nusa Tenggara Timur, down 79 basis points to 8.29% on Rp 7.5 trillion, and Lampung, up 86 to 7.45% on Rp 15.3 trillion.

Δ = change Jan→Mei 2026 in basis points. Bars scaled to Maluku's 9.32%. The 34 provinces sum back to the national figure — Rp 42.01 T of bad loans on Rp 848.4 T — exactly. A residual “Lainnya” bucket of Rp 24 miliar is excluded.

§7

The industry cut

Sector × quality · K+M only

National — sectors ≥ Rp 2 T, ranked

Construction is the worst sector at 9.89%, almost exactly double the SME average — an Indonesia problem, not a Jabar one. Trade is the weight: Rp 392.2 trillion, nearly half the national book, supplying 50.0% of the national SME NPL stock. Agriculture is the cleanest large sector at 2.79% — remember that number.

Jawa Barat — gap vs national, pp

worse in Jabarcleaner

Manufacturing is the smoking gun: 9.53% vs 5.41% national, +4.12 points on the province's second-largest SME book, alone a quarter of Jabar's bad-loan stock. Agriculture runs nearly twice its national rate. The clean spots are real too — education 0.95%, professional services 1.13%, healthcare 1.66%, all better than national. The consistency with the Bandung–Bekasi–Karawang manufacturing belt is our reading, not something the data states.

The engine cell

Medium-segment trade, West Java

Inside Jabar's medium segment: manufacturing 13.64%, construction 12.14%, trade 8.03%. The trade cell deserves the closest attention — it sits on the largest medium book in the province and it is accelerating. That single cell is the main engine behind Jabar's medium ratio moving 7.51 → 8.61.

6.30% → 8.03%

in five months

+173 bps

Rp 14.02 T medium-trade book

Mix or performance? — shift-share decomposition

axis zoomed — bars start at 4.4% · components computed on the sector cut and may not sum exactly at two decimals

If each of Jabar's sectors performed at its national rate, the province's industry mix would produce just 5.19% — composition explains only a quarter-point. The remaining 1.35 points is the same industries performing worse inside West Java. The problem is local execution and local borrowers, not portfolio shape.

§8

The bottom line

for anyone lending into SME

4.95% — SME NPL, small and medium only, May 2026. An actual from the regulator's portal, not an estimate — essentially flat since 2023, and up about 35 basis points on its 2021 level.

Excluding micro raises the ratio. The risk sits in the medium segment at 5.73% — and micro is catching up fast.

Jawa Barat is the real finding. Worst big-book credit quality in the country, 6.55% and still deteriorating, its medium segment at 8.61% and accelerating. On the last full SPI print its UMKM book was the only large one that shrank — down 1.3% between 2023 and June 2025, on project location — though on the portal cube's narrower small-and-medium cut it has grown again since, Rp 86.0 to 88.9 trillion over January–May 2026. And its corporate book is no alibi. The 1.45% often quoted for Jawa Barat is an NPL net reading. On the gross basis this report uses, OJK's own Jawa Barat office puts the province's all-credit NPL at 3,44% (March 2026, by project location) against a national 2,17%. On our cube's bank-location basis its non-UMKM book runs 3.05% against 1.64% nationally. Jawa Barat is worse across every segment — not just the small ones.

By industry the problem has a name: manufacturing (+4.1 pp vs national) and medium-segment trade (+173 bps in five months). Shift-share says it's performance, not mix.

Underwrite West Java medium tickets to a visibly tougher standard, and overlay agriculture there despite its clean national profile.