@@ -4,15 +4,15 @@ Three real-world stories that show P2Predict on data anyone can download and ver
 
 | Case study | Audience | Data source | Demonstrates |
 |---|---|---|---|
-| [Battery Management ICs](battery-management-ics/) | EE procurement engineers, electronics category managers, hardware BOM owners | Octopart / Mouser / DigiKey API | The classic parametric procurement use case — `--explain` and `--whatif` for spec-driven cost trade-offs |
-| [Used vehicles](used-cars/) | Anyone — the universally-relatable tutorial | Kaggle (CC license) | Log-target SHAP + multiplicative explanations on prices that span orders of magnitude |
-| [Aerospace fasteners](aerospace-fasteners/) | Fastener / MRO category managers, federal procurement, cost engineers | DLA PUB LOG / FLIS catalog (public domain) | **Detecting noisy data** — measuring a model's R² ceiling before tuning, plus log-target multiplicative SHAP on a heavily-skewed price |
+| [Battery Management ICs](battery-management-ics/) | EE procurement engineers, electronics category managers, hardware BOM owners | Octopart / Mouser / DigiKey API | The classic parametric procurement use case, `--explain` and `--whatif` for spec-driven cost trade-offs |
+| [Used vehicles](used-cars/) | Anyone, the universally-relatable tutorial | Kaggle (CC license) | Log-target SHAP + multiplicative explanations on prices that span orders of magnitude |
+| [Aerospace fasteners](aerospace-fasteners/) | Fastener / MRO category managers, federal procurement, cost engineers | DLA PUB LOG / FLIS catalog (public domain) | **Detecting noisy data**, measuring a model's R² ceiling before tuning, plus log-target multiplicative SHAP on a heavily-skewed price |
 
 ## Why these, in this order
 
 - **Battery Management ICs** is the closest-to-target audience. Procurement engineers in EE will recognise the methodology immediately and pay attention.
 - **Used vehicles** is the warm-up tutorial. Everyone gets the problem, the data is clean, and it's the strongest visual demo of P2Predict's log-target multiplicative SHAP attribution.
-- **Aerospace fasteners** is the *honesty* story. It's the one case study where the model is deliberately weak — and the value is the diagnostic that proves it's the data, not the model. It teaches how to measure a noise floor and know when to stop tuning, the skill that matters most on messy real-world data. The dataset (DLA PUB LOG, public domain) is also a credibility signal to the federal-procurement and cost-estimating audience.
+- **Aerospace fasteners** is the *honesty* story. It's the one case study where the model is deliberately weak, and the value is the diagnostic that proves it's the data, not the model. It teaches how to measure a noise floor and know when to stop tuning, the skill that matters most on messy real-world data. The dataset (DLA PUB LOG, public domain) is also a credibility signal to the federal-procurement and cost-estimating audience.
 
 The three different doors into the same product matter: an EE engineer finds #1, a curious developer finds #2, a fastener category manager finds #3. None overlap.
 
@@ -23,14 +23,14 @@ Each subfolder has the same four files:
 ```
 case-studies/<name>/
   README.md               ← the procurement story, results, takeaways (5-minute read)
-  fetch_data.py           ← pulls data from the source — credentials in env vars
+  fetch_data.py           ← pulls data from the source, credentials in env vars
   train.py                ← trains a P2Predict model (or just documents the CLI command)
   predict_examples.py     ← runs sample predictions with --explain and --interval
 ```
 
 The pattern is deliberate: anyone reading the case study's `README.md` should be able to reproduce the result end-to-end in under 15 minutes if they have the right API credentials (for #1) or just an internet connection (for the public-data studies).
 
-## Reproducibility — what's checked in and what isn't
+## Reproducibility, what's checked in and what isn't
 
 | Type | Checked in? | Why |
 |---|---|---|