> ## Documentation Index
> Fetch the complete documentation index at: https://docs.variable.global/llms.txt
> Use this file to discover all available pages before exploring further.
# 2026
## Verify locked versions with the Verifier role
Teams can now sign off on locked versions. A new **Verifier** role grants
read-only access plus one capability — marking a locked version as
**Verified**, the highest-confidence state, recorded in the audit history with
an optional note. Assign the role from Users & roles, then use **Mark as
verified** on a locked version. See
[Product and material versioning](/docs/help/versioning).
Feature flag
## Set a dataset on an element through the API and MCP
You can now bind an emission-factor dataset to a sourced element — material,
energy, or process — in the same call that creates or updates it, instead of
only reading the binding back. Send `dataset: { uuid }` (plus an optional
`datasetAmount`) to point the element at a dataset; binding one makes the
element external (`dataSource: external`), and sending a different dataset on
update re-points it. Use `/v1/database` to find a dataset to bind.
These writes are available through MCP as well, alongside the rest of the
element write contract: `weight`, `packagingWeight`, `recycledPercent`,
`dataSource`, `supplierPartId`, `cpcCode`, and `dataQualityIndicators`. Custom
transport modes accept their backing dataset as `dataset` too (the older
`footprint` field still works but is deprecated). See [Datasets](/docs/dataset)
and the [Material](/api-reference/v1/material/post) reference.
## Justify your data quality ratings
You can now record a free-text justification alongside your data quality
ratings — why those ratings were chosen, which production years and background
datasets the figures rest on, and any caveats a reviewer should know. It reads
and writes through the API and MCP as `dataQualityIndicators.description`, on
both products and materials, and is editable in the Variable app next to the
ratings it explains. See [Data quality](/docs/help/data-quality).
## Set data quality indicators through MCP
Data quality indicators are now available through MCP, matching what you can
already do in the Variable app. When you create or update a product or
material, you can record the documentation year, the share of the footprint
backed by primary data, the share backed by supplier-specific data, and the
five data quality ratings (technological, temporal, geographical,
completeness, and reliability). The average data quality score is still
calculated for you. See [Data quality](/docs/help/data-quality).
## Bundle or unbundle optional benefits (D)
Bill-of-materials lines placed at **optional benefits (D)** can now toggle
between bundled and unbundled, the same as every other stage group. Bundled
(the default) imports the source's cradle-to-gate (A1–A3) total stamped onto
module D; unbundled imports the source's own module-D value — its recycling or
recovery credit — into stage D. Existing inputs are unchanged: the default
stays bundled. See [Life-cycle stages explained](/docs/help/life-cycle-stages).
## Richer dataset reference on custom transport modes
A custom transport mode is backed by a single freight emission-factor dataset,
and its `dataset` reference now carries the same detail as external elements:
alongside `uuid` and `name`, read responses include the `source` database and
the `declaredUnit` (a freight unit — tonne-kilometre or tonne-mile). Built-in modes have no backing dataset
and omit it. There's no `datasetAmount` — a mode's factor is a unit quantity in
its declared unit, so it adds nothing beyond `declaredUnit`.
The enriched reference also surfaces through the MCP `get_transport_mode` and
`list_transport_modes` tools. See the
[Transport mode](/api-reference/v1/transport/mode/list/get) reference.
## See the dataset behind an external element
External elements — materials, energy, and processes that get their impact from
a single emission-factor dataset (`dataSource: external`) — now tell you which
dataset that is. Read responses include `dataset` (the backing dataset's `uuid`,
`name`, `source` database, and `declaredUnit`) and `datasetAmount` (the quantity
and unit consumed). Both are absent for modeled products. If you narrow a
response with `return`, list `dataset` / `datasetAmount` there too or they're
omitted.
The same fields surface through the MCP element tools (`get_material`,
`list_materials`, and the energy/process element tools). See the
[Material](/api-reference/v1/material/get),
[Energy](/api-reference/v1/energy/get), and
[Process](/api-reference/v1/process/get) references.
## Absolute impact contributions in the inventory tab
In the app, a product's **Inventory** tab now shows each element's **absolute**
contribution to the model — its intensity scaled by quantity into the model's
stages — instead of a per-unit intensity, so you can see what every material,
process, and transport segment actually adds to the product's footprint. The
GWP-fossil contribution shows for everyone; a missing value stays a dash and a
real zero shows zero.
With all impact indicators enabled, a selector above the table swaps the impact
column across every EN 15804+A2 indicator the model exposes — each in its own
unit — defaulting to GWP-fossil. See
[Environmental impact indicators](/docs/help/impact-indicators). Preview
## Attach transport to a model input
You can now attach a freight
[transport lane](/api-reference/v1/transport/lane/list/get) to a model input
through the public API and MCP — modeling the transport leg of a product's bill
of materials end-to-end without dropping into the app. Send a `transport` block
on `POST` / `PATCH` (and the bulk input endpoint) with the `lane` to follow and
either the material input it carries (`for`) or an explicit cargo `weight`; send
`transport: null` to detach.
When you set `for`, the leg takes its lifecycle stage and cargo weight from the
carried material, so `weight` is derived rather than sent. Reads echo the lane,
weight, carried material, and the derived tonne-kilometres (`tkm`); the leg's
carbon footprint flows into the model total. The MCP `create_model_input` and
`update_model_input` tools gain the same field. See
[Models & Inputs](/api-reference/v1/model/inputs/post).
## Deleting an import undoes the field updates it made
Deleting a product or material import no longer just removes the items it
created — it also **reverts the scalar updates** that import made to
pre-existing items. Fields it overwrote are restored to their prior values, and
blanks it filled in are cleared again. The revert is guarded: any field you (or
a later import) changed after the import is left untouched, so a delete never
clobbers newer edits. When some fields are skipped for that reason, the app
tells you how many couldn't be reverted. See
[Undoing an update](/docs/import/import-behavior#undoing-an-update).
## Material imports default to filling in blanks
Material imports now default to **Fill missing only**, so re-importing a file
backfills blank fields — weights, CPC/HS codes, descriptions, notes — on
materials you already have, without overwriting anything you've already set. A
stored `0` or empty value still counts as set and is kept. Product and other
import types are unchanged and still default to **Create new only**; you can
switch either to any mode on the upload screen. See
[Import match mode](/docs/import/import-behavior#import-match-mode-products-and-materials).
You can also see which existing products an import updated, not just the ones it
created: an import's detail view now lists updated products under their own
**Updated** tab.
## Energy and process elements in the public API
You can now manage energy and process elements through the public API and MCP,
the same way you already manage materials: list them, create and update them,
upload an image, and delete them — one at a time or in bulk. New `/v1/energy`
and `/v1/process` endpoints mirror `/v1/material`.
An energy or process element behaves like any other element; what makes it
"energy" or "process" is simply its place in the taxonomy. Creating one files it
under the matching category automatically, and you can pass a `taxonomy` to pick
a more specific subcategory.
`GET /v1/material` now returns materials (and any uncategorized elements), so
energy and process elements no longer appear there — reach them through their own
endpoints. See the [Energy](/api-reference/v1/energy/list/get) and
[Process](/api-reference/v1/process/list/get) references.
## Update existing materials and products on import
Material and product imports now have a **match mode** that decides what happens
when a row matches an item you already have. Pick it on the upload screen:
**Create new only** (the default — matched rows are skipped, today's behavior),
**Overwrite existing** (the imported value wins), or **Fill missing only** (fill
blanks, keep what you've already set). Use it to backfill enriched source data —
weights, CPC/HS codes, descriptions, notes — onto items you created earlier,
without re-keying them by hand.
Only the item's own fields are updated; its relationships — supplier, assigned
datasets, and bill-of-materials inputs — are left untouched in every mode. Under
**Fill missing only**, a stored `0` or empty value counts as set and is kept:
Variable treats zero as a value and only a truly missing field as missing.
Re-imports still match by SKU (or name), so they never create duplicates. See
[Import match mode](/docs/import/import-behavior#import-match-mode-products-and-materials).
## Set data quality indicators on materials
Material elements now have an editable data quality panel beside the dataset
passport, so you can set all five pedigree-matrix dimensions — technological,
temporal, and geographical representativeness, completeness, and reliability —
directly on a material, not only on the dataset it points to. The aggregate
data quality rating folds in all five and updates as you edit.
## Public API: set temporal representativeness directly
`temporalDQR` is now a writable data-quality indicator on product create and
update, alongside the other pedigree-matrix ratings (`technologicalDQR`,
`geographicalDQR`, `completenessDQR`, `reliabilityDQR`). Send `1` (Good — same
reporting year), `2` (Fair — less than 5 years old), or `3` (Poor — more than
5 years old).
Previously `temporalDQR` was read-only and inferred from `documentationYear`;
it's now set directly, matching the data-quality panel in the app.
`documentationYear` remains a separate, writable field for reference.
## Public API & MCP: recycled content on products
Product and material responses now carry recycled-content percentages.
`recycledPercent` is the value you declare (0–100) — a `0` is a declared zero,
while an undeclared value is absent — and `modeledRecycledPercent` is the
server-computed weight-weighted average of the contributing inputs' recycled
content, the recycled-content counterpart to `modeledWeight`.
Each model input also carries its own `recycledPercent`, inherited from the
input's source product — the source's modeled recycled content when available,
otherwise its declared value, or `0` when there is no source product or the source
has neither — so you can trace a model's recycled content line by line.
`modeledRecycledPercent` is derived from the bill of materials, read-only
(ignored if sent on create or update), and present only once a model has
contributing inputs. The same fields flow through every MCP tool that reads a
product or material.
## Public API & MCP: total modeled weight on products
Product and material responses now also carry `modeledTotalWeight` — the gross
modeled weight, the sum of `modeledWeight` and `modeledPackagingWeight` — in the
same `Amount` shape (`{ quantity, unit }`).
It's derived at read time from the two component weights rather than stored, so
it can't drift from them: a missing addend contributes nothing, the unit follows
the first present addend, and the field is present whenever at least one of the
two is. Like its parts it's read-only (ignored if sent on create or update). The
same field flows through every MCP tool that reads a product or material.
## Filter the product inventory by dataset status
The product inventory now has a **Dataset status** filter so you can narrow the
list to items that already have a dataset assigned (**Assigned**) or still need
one (**Unassigned**). Use it to find the products missing footprint data before
a bulk update. The filter applies to assignable items (products and energy);
composed (live) items count as assigned.
## Public API & MCP: modeled packaging weight on products
Product and material responses now also carry `modeledPackagingWeight` — the
server-computed sum of a model's A1–A3 packaging input weights — alongside the
declared `packagingWeight`, in the same `Amount` shape (`{ quantity, unit }`).
It's the packaging counterpart to `modeledWeight`: compare it against the
`packagingWeight` you declared to mass-balance packaging the way an EPD reviewer
would. It's derived from the bill of materials, read-only (ignored if sent on
create or update), and present only once a model has qualifying packaging inputs.
The same field flows through every MCP tool that reads a product or material.
## Public API & MCP: modeled weight on products
Product and material responses now carry `modeledWeight` — the server-computed
sum of a model's A1–A3 product-group material input weights (excluding packaging
and ancillary inputs) — alongside the existing `weight` field, in the same
`Amount` shape (`{ quantity, unit }`).
Use it to compare what a model actually adds up to against the weight you
declared: `weight` is the value you set, while `modeledWeight` is derived from
the bill of materials and is read-only (ignored if sent on create or update).
It's present only once a model has qualifying inputs. The same field flows
through every MCP tool that reads a product or material.
## Auto-assign transport across a bill of materials
Variable can now fill in transport for you. Instead of hand-attaching a lane to
every material in a model, use **Auto-assign transport** on a product — or select
several products on the product list and run it in bulk. For each material,
Variable reads the facility it's made at (or its supplier's location) and the
facility your product is made at, finds an existing transport
lane that connects those points within a radius you choose, and creates the
transport input using the material's own weight.
The closest matching lane wins; a tie with no clear winner is left for you to
resolve, and transport you've already set is never overwritten. When the run
finishes you get a summary grouped by result — assigned, no match, ambiguous, and
the rest. The default radius is 50 km, adjustable per run. Preview
## Public API & MCP: categorize model inputs
You can now read and write an input's **material group** through the
[API](/api-reference/v1/model/input/post) and MCP. Each input carries a
`materialGroup` array — common values are `product`, `packaging`, and
`ancillary`, and custom values (for example from a PCR document) are allowed.
Set it on create or update and it comes back on every read (single input and
list). This lets integrations model packaging and ancillary inputs end-to-end
instead of relying on the value Variable infers. The legacy `isPackaging` flag
stays consistent automatically.
## Public API & MCP: write a dataset's impact indicators
You can now set a dataset's environmental impact indicators through the
[API](/api-reference/v1/dataset/impacts/put) and MCP — the write counterpart to
reading them. `PUT /v1/dataset/{uuid}/impacts` takes the same keyed shape the
read endpoints return: each EN 15804+A2 indicator code (e.g. `GWP-fossil`,
`ODP`) maps to its life-cycle stage values (`A1`–`A5`, `B1`–`B7`, `C1`–`C4`,
`D`). Totals like `A1_A3` and `totalCarbonFootprint` are computed for you, and
the dataset's footprint is recalculated on write.
Indicators you omit are left unchanged, so you can fill in a dataset
incrementally. The dataset must be in draft state (newly created datasets are).
Using an MCP client, this means an assistant can now build a fully populated
dataset from a source document in one flow — create the dataset, then call
`set_dataset_impacts` to write its indicator values.
## Public API & MCP: all impact indicators in responses
Products, materials, datasets, switches, and [database search](/api-reference/v1/search/list)
results can now return their full set of [EN 15804+A2 impact indicators](/docs/help/impact-indicators)
— acidification, ozone depletion, water and energy use, waste, and the rest — not just carbon.
Impacts come back under a new top-level `impacts` field, keyed by indicator code (for example
`GWP-fossil`, `ODP`, `AP`).
Opt in with the `impacts` query parameter: `?impacts=all` for everything, or
`?impacts=gwp-fossil,odp,ap` to filter to a subset. Without it, responses return `GWP-fossil`
only, so existing integrations are unaffected. The same data flows through every MCP tool that
reads these resources.
The existing `footprint.CO2e` field is unchanged and still works, but is now **deprecated** in
favor of `impacts["GWP-fossil"]`. See [Impacts](/docs/impacts) for the response shape and query
semantics.
## Public API & MCP: manage your locations
You can now manage your company's locations — factories, warehouses, offices,
and ports — through the [API](/api-reference/v1/location/list) and MCP, not just
in the app. List the sites you already have, add new ones, update their details,
and remove the ones you no longer need. Pin a site to a point on the map so it
carries real coordinates, or keep it as a plain address. You can also set up a
location on behalf of a supplier you work with, not only your own company.
Once a location exists, reuse it anywhere you pick a place — for example the
start or end of a [transport route](/api-reference/v1/transport/lane/post). A
location that's still in use is protected: remove the places that reference it
before you can delete it.
Using an MCP client, just ask your assistant to list, create, update, or delete
a location.
## Public API & MCP: fetch a supplier by internalId
`GET /v1/supplier/{id}` now resolves a supplier by its `uuid`, `syncId`, or
`internalId` — the same `internalId` the API returns on the supplier — so you
can read a supplier back by the identifier you already have. The MCP
`get_supplier` tool accepts all three. `PATCH` and `DELETE` continue to require
the `uuid`.
## Public API & MCP: model freight transport
Build and reuse freight routes through the API. **Transport lanes**
(`/v1/transport/lane`) are ordered legs, each with a mode, an origin, and a
destination; distances auto-compute from the endpoints, and a lane reports its
carbon intensity (`CO2ePerTkm`) rather than an absolute total — the total
materialises once cargo weight is applied. **Transport modes**
(`/v1/transport/mode`) cover the system-curated built-ins (`truck`,
`container-ship`, `train`, the `air-*` haul bands, …) plus custom modes you
create from your own freight emission-factor datasets. Lanes support full
CRUD; built-in modes are read-only, while the custom modes you create are
fully editable.
The MCP server exposes the same operations as transport-lane and
transport-mode tools, so an agent can assemble routes end to end.
## Public API: read and write product notes
Products now expose their `notes` field through the API and MCP. `GET`
responses include `notes`, and you can set it on `POST`/`PATCH` (or via the
`create_product`/`update_product` MCP tools) alongside `description`.
## Public API: resolve addresses into transport-leg locations
Two new endpoints turn a free-text address into a referenceable location, so you
can set transport-leg origins and destinations programmatically instead of
pre-seeding them in the app. `GET /v1/geolocation?search=...` geocodes a string
like `Syracuse, NY` into ranked, un-persisted candidates; `POST /v1/geolocation`
persists a chosen candidate (by `googleMapsId`) or a manual point
(`name` + coordinates) and returns it with a `uuid`. Reference that `uuid` — or a
candidate's `googleMapsId` directly — on a transport leg.
The MCP server exposes the same surface as `geocode_location` and
`create_geolocation` tools.
## Terrestrial eutrophication now reported in mol N eq
The EP-terrestrial (Terrestrial eutrophication) impact indicator is now
reported in `mol N eq`, matching EN 15804+A2. It was previously labelled
`mol H+ eq` (the acidification unit) — a mislabelling inherited from
ecoinvent, which documents the same error in its
[v3.11 known issues](https://support.ecoinvent.org/ecoinvent-version-3.11).
The stored values are unchanged — only the unit label was corrected.
## Inputs editor: enter quantities in alternative units
The input amount field now groups units by dimension and lets you enter
a quantity in any unit the source product supports — e.g. enter a board
as `2.4 m` against an `m²`-declared product, and Variable converts to the
declared unit using the product's conversion factors. The source
product's length, volume, and area conversion factors are now editable
directly from the product preview sidebar. This is the UI counterpart to
the `declaredAmount` API shape shipped on May 22.
## Location picker: pick a location without an address
Picking a `Location` that has no attached `GeoLocation` in the
transport-leg location picker now works — previously the click was
silently dropped, blocking address-less custom locations from being used
as a segment endpoint.
## Public API: machine-readable taxonomy path on responses
Taxonomy references on `/v1/product`, `/v1/material`, and `/v1/activity`
responses now include a `path` field (e.g. `material`, `energy/electricity`)
alongside `uuid` and `name`. The path is a stable, machine-readable grouping key
— unlike the human-readable `name` — so you can reliably branch or aggregate on a
category. Purely additive; existing `uuid`/`name` readers are unaffected.
## File import: BOM rows in mixed units
BOM imports now accept an input quantity in any unit the source product supports,
even when it differs from that product's declared unit (e.g. listing a board as
`2.4 m` against an `m2`-declared product). Variable converts the value to the
declared unit for impact calculations using the product's conversion factors and
keeps your entered value in the API. Rows whose unit has no matching conversion
factor are skipped. See the
[BOM preparation guide](/docs/import/prepare-bom) for details.
## Suppliers can have their own locations
A supplier can now own one or more named locations — a factory, warehouse, or
port, each with its own address — managed from the new **Locations** tab on the
supplier page. Mark one as the default (the first one added is set automatically),
and it's used wherever a supplier needs a single address. Supplier locations also
appear anywhere you pick a place, such as the origin of a transport leg, searchable
by supplier name, so you can reuse *"Acme — Rotterdam warehouse"* instead of
re-typing the address.
## Public API: fetch a single model input
New `GET /v1/model/{modelId}/input/{inputId}` returns one input on a given model.
Responds with `404` when the input does not belong to the model (or does not
exist for this caller). The response carries the same `quantity` / `unit` /
`declaredAmount` shape as the list endpoint.
The MCP server exposes the new endpoint as a `get_model_input` tool, alongside
refreshed `list_model_inputs`, `create_model_input`, and `update_model_input`
descriptions that surface the alternative-unit semantics introduced last week.
## Owner-controlled AI features toggle
Company owners can now enable or disable AI features for their company directly
from **Company Settings** with the new "Enable AI features" switch. Previously
this required Variable support to flip the company-level `DisableAI` flag.
## Activities anchored in your company timezone
Activity dates now honor a `Company.timezone` (IANA, defaulting to
`Europe/Oslo`), so partial inputs like `"2026"` or `"2026-03"` resolve in the
company's timezone instead of UTC. Period filters (YTD, MTD, specific years)
bucket from the same timezone, and each `Activity` exposes a new `timezone`
field recording the zone used on write.
## Public API: alternative-unit response shape on `/v1/model/{id}/input`
GET, POST, and PATCH responses on `/v1/model/{id}/input` now return the `quantity`
and `unit` you supplied, even when those units are in a different dimension than the
source product's declared unit (e.g. sending `kg` for an `m3`-declared source product
with a weight conversion factor). A new read-only `declaredAmount` field carries the
canonical converted amount in the source product's declared unit — this is what
impact calculation uses. `declaredAmount` is present only for cross-dimensional
inputs; same-dimension entries (e.g. `l` against an `m3`-declared product) are
unchanged. `declaredAmount` is ignored if supplied on POST/PATCH.
## Bulk delete activities from the activity list
Activities can now be deleted in bulk directly from the Activity list. Select activities on the
Draft or Rejected tab and use the new Delete button in the floating bulk action bar.
## Admin-managed maintenance windows for self-hosted
Self-hosted operators can now schedule and run maintenance windows from
**Admin -> Maintenance windows** instead of toggling an environment variable
and redeploying. Configure a recurring window (daily, weekly, or monthly in
any IANA timezone) with an `HH:mm` start and a duration up to seven days, or
hit **Start now** for a one-off window.
## Offline ISO-3166-1 country coverage for self-hosted
The seeded `GeoLocation` dataset now covers the full ISO-3166-1 country and territory set (plus
Kosovo, which Google Maps treats as `XK`). Self-hosted instances can resolve any country offline
without a Google Maps API key. Each row ships with the same `googleMapsId`, lat/lng, and bounds
that Google's Geocoding API returns, so locations created via seed match locations created via
online lookup.
## BOM import: supplier matching and XLSX precision fixes
BOM imports now match existing suppliers by `supplierId` (internalId) alone when no
`supplierName` is provided, so id-only rows correctly link to suppliers already in your
account instead of falling through and creating products without a supplier link. The
supplier lookup also no longer caps at the first page, so accounts with more than 250
suppliers match reliably.
XLSX imports now preserve full numeric precision (previously, cells like `0.0004`
displayed as `0.000` were silently truncated to the display string), normalize date
cells to ISO strings, and drop spreadsheet error sentinels (`#REF!`, `#N/A`, `#VALUE!`,
etc.) instead of importing them as literal strings.
## Bulk location import
Users can now bulk-create or update Locations from CSV or XLSX files. The Locations page
exposes an Import action that maps name, address, and type columns; re-imports merge on
name within the company so existing values are preserved when a column is left blank.
Addresses are geocoded in the background after import.
## SmartEPD integration
Companies can now connect Variable to SmartEPD v2 and push EPDs directly from the platform.
Owners configure the connection in Company Settings → Integrations by providing their SmartEPD
API key and selecting a workspace. On each EPD, a "Sync to SmartEPD" action lets users choose a
PCC and project (creating a new project if needed) and submit the EPD to SmartEPD. A sync status
badge on the EPD page shows the current state (pending, syncing, synced, failed, or out of date),
and a "Check status" action polls SmartEPD for the latest review state. Once an EPD is approved
or published in SmartEPD, Variable automatically locks the EPD to Verified or Published to mirror
the remote lock. If the EPD is edited locally after its last successful sync, the auto-lock is
skipped and the record is marked `out_of_date` instead.
## Dataset resource in the public API
The public API now supports managing datasets in your account via `/v1/dataset`. You can list, fetch, create,
update, upload images, and delete datasets - whether they were user-created, added from the Variable Database,
or shared by a supplier. Deleting a dataset that came from the Variable Database or a supplier removes it from
your account while leaving the original in place; user-created datasets are permanently deleted.
New `DATASET_READ`, `DATASET_CREATE`, `DATASET_UPDATE`, and `DATASET_DELETE` permissions gate these routes.
The same operations are exposed as MCP tools (`list_datasets`, `get_dataset`, `create_dataset`, `update_dataset`,
`delete_dataset`), and dataset payloads now use dedicated `Dataset`, `NewDataset`, and `EditDataset` schemas in
the OpenAPI spec instead of reusing `Material` / `NewProduct` / `EditProduct`.
`dataSource` is no longer part of the Dataset surface. The previously documented `dataSource: "database"` value
has been removed from dataset responses, and sending a `dataSource` field in a dataset request body has no effect
* the server always treats the resource as a dataset based on the route. Clients that read or write `dataSource`
on the `/v1/dataset` endpoints should drop the field.
The `ProductDataSource` enum used on `/v1/product` and `/v1/material` is narrowed to `"model"` and `"external"`.
The `"database"` value is no longer accepted on those routes - datasets now have their own endpoint.
## Organization assignment on user invite
Admins can now assign an organization when inviting new users. The org selector appears in the invite
modal and the backend creates the membership in the same transaction as user creation.
## Role-based access control
A new [Access Control](https://app.variable.global/access-control) page gives a clear overview of what each role
* Owner, Admin, Contributor, and Viewer - can access across your organization. Permissions are organized into five
categories:
* **Products** - LCAs and Declarations (All roles)
* **Inventory** - Materials, Energy, Transport, Process, Datasets, and Activities (All roles)
* **Exchange** - Elements, Suppliers, Customers, Requests, and Deliveries (Owner and Admin only)
* **Reports & Plans** - Analyze, Report, and Plan (Owner and Admin only)
* **Account** - Org, Users, Locations, Company Settings, and more (Mixed access)
## Extended org access control
Organization-level access control now covers transports, datasets, parts, and EPDs in addition to products and
materials.
## Self-service dataset copying
Users can duplicate database datasets into their own company, enabling custom modifications. Preview
## Second indicator in modelling
The modelling screen now supports a secondary impact indicator alongside GWP, allowing side-by-side comparison of
environmental impacts. Sorting by the second indicator is also supported.
## Taxonomy enforcement and dataset alignment
Taxonomy alignment is now enforced when assigning datasets to inventory items. A mismatch indicator shows
when a dataset's taxonomy differs from the input and offers a one-click update action.
## PERM/PENRM calculation for classified materials
Impact calculations now include PERM (primary energy from renewable materials) and PENRM (primary energy from
non-renewable materials) for material inputs classified with flow types.
## BCCP and BCCAP rules
The impact calculation service now applies biogenic carbon content rules (BCCP for products, BCCAP for packaging)
based on functional group classification, per EN 15804+A2.
## Geographic filtering
Search results can now be filtered by geography, making it easier to find region-specific datasets and products.
## A1-A3 entry mode toggle for custom datasets
Custom datasets now support toggling between individual A1, A2, A3 stage entry and a combined A1-A3 entry mode.
## Decimal separator override for data imports
Data imports now support explicit decimal separator selection, resolving ambiguity with European-format CSV files.
## Self-hosted data package delivery
Self-hosted instances now receive emission factor data packages via mounted volumes, simplifying data updates.
## Versioning in Public API
The public API now supports versioning for products and materials. Create new versions, list version history,
and fetch specific versions via the API. Bulk endpoints updated accordingly. Preview
## `varId` added to Public API
All API responses now include the `varId` field - a human-readable, short identifier for each entity.
Preview
## BOM import adapters
A new adapter pattern for BOM imports allows external formats (e.g., SAP S/4HANA) to be automatically
transformed into the expected CSV structure on upload.
## Bulk endpoints in Public API
New bulk create, update, and delete endpoints added for products, materials, and activities, allowing efficient
batch operations via the API.
## Manual geolocation entry
Users can manually enter geographic coordinates with source tracking, useful when automated geocoding is
unavailable.