diff --git a/CHANGELOG.md b/CHANGELOG.md
index ae550df..7a434b2 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,19 +7,26 @@ inside each artifact (`schema_version`).
 
 ## [Unreleased]
 
-### Changed
-- Documentation reads beyond credit risk without hiding where the project
-  started. A new *Where it fits* page states the origin in credit, the general
-  boundary — tabular models making decisions about individual cases that must
-  be explained, reproduced, audited or run outside Python — which domains fit
-  strongly or weakly and why, and a table reading the library's vocabulary
-  (`pd`, bands, reason codes, `risk_increasing`) in fraud, clinical triage and
-  churn terms. No code or artifact names change. The README's *Why this
-  exists* now opens with the credit origin — risk practitioners and data
-  scientists each right about what the other gives up — and states the choice
-  in terms any domain facing it will recognise. The docs home, fairness and
-  reason-code guides gain a sentence each; the fairness guide keeps lending as
-  its worked lens and says where its regulatory references come from.
+## [0.8.0] - 2026-09-13
+
+What compilation cost where decisions are made — and what to do about it.
+
+A portfolio retention figure is an average, dominated by the easy
+separations, while a decision is made in the narrow band of risk where a
+segment's cutoff sits. `retention_by_segment` reports, per segment, what
+compiling cost there: retention, agreement with the teacher across a PD
+cutoff range, and whether the band ladder can hold a cutoff inside that
+range. When one segment pays, `sample_weight` directs the whitebox's budget
+toward it, and the tuning guide says when that works and when a segment needs
+its own model instead.
+
+The documentation now reads beyond credit risk without hiding where the
+project started, and GitHub Discussions is open for questions, design debates
+and results on your own data.
+
+A minor release because `retention_by_segment`, `segments` and
+`sample_weight` are new public API. The runtime, exporters and artifact schema
+are unchanged.
 
 ### Added
 - `compileml.tune.retention_by_segment`
@@ -48,6 +55,18 @@ inside each artifact (`schema_version`).
   calibration table, whose emitted PD is only its raw score rescaled.
 
 ### Changed
+- Documentation reads beyond credit risk without hiding where the project
+  started. A new *Where it fits* page states the origin in credit, the general
+  boundary — tabular models making decisions about individual cases that must
+  be explained, reproduced, audited or run outside Python — which domains fit
+  strongly or weakly and why, and a table reading the library's vocabulary
+  (`pd`, bands, reason codes, `risk_increasing`) in fraud, clinical triage and
+  churn terms. No code or artifact names change. The README's *Why this
+  exists* now opens with the credit origin — risk practitioners and data
+  scientists each right about what the other gives up — and states the choice
+  in terms any domain facing it will recognise. The docs home, fairness and
+  reason-code guides gain a sentence each; the fairness guide keeps lending as
+  its worked lens and says where its regulatory references come from.
 - GitHub Discussions is open for questions, design discussions and results on
   your own data; the new-issue page links to it, and CONTRIBUTING says what
   goes where. CONTRIBUTING's starting points no longer list work that has
diff --git a/CITATION.cff b/CITATION.cff
index 74af896..b8976fa 100644
--- a/CITATION.cff
+++ b/CITATION.cff
@@ -19,7 +19,7 @@ authors:
 repository-code: "https://github.com/orgoca/CompileML"
 url: "https://github.com/orgoca/CompileML"
 license: Apache-2.0
-version: 0.7.1
+version: 0.8.0
 date-released: "2026-09-13"
 keywords:
   - credit risk
diff --git a/README.md b/README.md
index 879e463..d4b4243 100644
--- a/README.md
+++ b/README.md
@@ -408,7 +408,12 @@ src/compileml/runtime/
 
 ## Roadmap
 
-Recently shipped: exact attribution aggregated per tree, so explaining a decision no longer grows with feature count, and a fairness audit in `compileml.fairness` (0.5); exact drift decomposition, band calibration and baseline staleness in `compileml.monitor` (0.6).
+Recently shipped:
+
+* exact attribution aggregated per tree, so explaining a decision no longer grows with feature count, and a fairness audit in `compileml.fairness` (0.5);
+* exact drift decomposition, band calibration and baseline staleness in `compileml.monitor` (0.6);
+* the calibrated PD and reason codes from the COBOL and SQL exports (0.7);
+* retention by segment and across PD cutoff ranges, and weighting a whitebox toward the segment that pays (0.8).
 
 The current priorities are:
 
diff --git a/src/compileml/__init__.py b/src/compileml/__init__.py
index 4dbb2c6..52b5774 100644
--- a/src/compileml/__init__.py
+++ b/src/compileml/__init__.py
@@ -16,6 +16,6 @@
 # this attribute at build time (setuptools dynamic version), so the wheel
 # metadata, `compileml.__version__`, `compileml inspect`, and the
 # `compileml_version` recorded inside every artifact can never disagree.
-__version__ = "0.7.1"
+__version__ = "0.8.0"
 
 __all__ = ["decide", "load_artifact", "verify_artifact", "__version__"]