diff --git a/MEMORY.md b/MEMORY.md
new file mode 100644
index 00000000..5229429e
--- /dev/null
+++ b/MEMORY.md
@@ -0,0 +1,22 @@
+# Project Memory
+
+## HubSpot New Field Addition Process
+
+When adding a new field from HubSpot to the deal table, touch these 4 locations in order:
+
+1. **DB model** — `backend/app/db/models/organisation.py`
+   Add `field_name: Optional[str] = Field(default=None)` to `HubspotDealData`, near related fields.
+
+2. **HubSpot API fetch** — `etl/hubspot/hubspotClient.py`, `from_deal_id_get_info()`
+   Add the HubSpot internal property name string to the `properties=[...]` list.
+
+3. **Soft check** — `etl/hubspot/hubspotDataTodB.py`, `update_deal_with_checks()`
+   Add a `soft_assert(deal_in_db.field_name == hs_deal.get("hs_property_name"), "field_name mismatch")` entry to the `checks` list.
+
+4. **Upsert** — `etl/hubspot/hubspotDataTodB.py`, `upsert_deal()`
+   - **Update branch**: add `"field_name": deal_data.get("hs_property_name")` to the attr dict
+   - **Insert branch**: add `field_name=deal_data.get("hs_property_name")` to the `HubspotDealData(...)` constructor
+
+**Notes:**
+- Date fields: wrap with `self._parse_hs_date(deal_data.get(...))` in steps 3 & 4.
+- If HubSpot property name differs from DB column name (e.g. `coordination_status__stage_1_` → `coordination_status`), use the HS name in `.get()` and the DB name as the key/attr.