Skip to content

Rate observatory · Consumer and ordinary credit

Usury rate.

Tracking of the rate certified by the Financial Superintendence of Colombia and a projection of next month built from 107 million Form 414 disbursement records.

See all projectsHow it was built
The problem
The SFC certifies the usury rate every month; until then, the finance team couldn’t tell which way the ceiling on its rates was moving.
What it enables
Projecting the usury rate before the official publication and anticipating its effect on income tied to that ceiling, such as credit cards.
Impact
2 bp average error vs. the official SFC rate

  • Current usury rate · Oct 2026

    28.59%

    TIBC 19.06 % · −65 bp vs. previous month

  • Projected vs. published · Oct 2026

    −0.8bp

    projected 28.58 % · published 28.59 %

  • TIBC base · Oct 2026

    $ 36.1tn

    Sep 4 – Sep 25 · 18.4M loans

Trend

Usury rate over time

+423bp

From October 2025 to October 2026

By effective month · % EA

25.0 %27.5 %30.0 %Oct 25Feb 26Jun 26Oct 2628.59 %Published28.58 %Projected

TIBC base by week

$ 9.7tn

Week of Sep 25, 2026

Products in the SFC methodology · COP billions

2,5005,0007,500Sep 25Jan 26May 26Sep 265,507Ordinary commercial4,205Consumer

Weighted rate of the base

18.69%

Week of Sep 25, 2026 · both types, unadjusted

Without model adjustment, by loan type · % EA

15 %20 %25 %Sep 25Jan 26May 26Sep 2623.52 %Consumer15.00 %Ordinary commercial

How the model reproduces the official rate.

Open data does not exactly match the certified rate. An optimization model adjusts each category's weight until it reproduces the published history; the same factors project the next month.

History, reproduced

13 of 13months

reproduced with error below 0.01 bp

Published, unadjusted and model TIBC · % EA

16 %18 %20 %Oct 25Feb 26Jun 26Oct 2619.06 %Published19.0600 %Model18.97 %Unadjusted

Adjustment by dimension

8.6%

Product is the most adjusted dimension

Mean weight change, amount-weighted · %

  • Product8.6 %
  • Loan type6.3 %
  • Term3.2 %
  • Rate type2.8 %
  • Amount range0.6 %
  • Institution0.5 %

Factors by product

+14.5%

Libranza otros · largest weight change

The 15 categories with the largest amount · weight change %

  • Libranza otros+14.5 %
  • Empresarial en pesos+11.0 %
  • Factoring+9.5 %
  • Oficial o Gobierno+7.9 %
  • Vehículo+7.4 %
  • Tarjeta de crédito para ingresos superiores a 2 SMMLV+5.5 %
  • Leasing financiero - financiación inmuebles diferente de habitacional+4.7 %
  • Leasing financiero - financiación maquinaria y equipo+3.9 %
  • Comerciales con recursos provenientes de redescuento+1.3 %
  • Tarjeta de crédito para ingresos hasta 2 SMMLV+1.1 %
  • Financiero e institucional−0.4 %
  • Leasing financiero - otros−0.6 %
  • Leasing financiero - financiación vehículos−0.6 %
  • Crédito rotativo−7.7 %
  • Libre inversión−12.0 %

How it was built.

A Python pipeline turns the Superintendence's massive CSV files into an incremental Parquet data lake, calibrates the model against the official releases and publishes this view.

  • Data

    107M

    weekly Form 414 records

  • Storage

    24 → 0.7GB

    from CSV to columnar, partitioned Parquet

  • History

    222weeks

    of cutoffs in the data lake

  • Model

    134factors

    calibrated against the official TIBC

  • Python
  • DuckDB
  • Parquet
  • NumPy · SciPy
  • Plotly
  • Streamlit
  • pytest