Version 2026.45
4 breaking25 new17 changed2 retiring36 fixed
Reddit Ads reporting API changes¶
Reddit Ads datastreams that select more than one of Keyword, Community or Interest, or combine them with DMA, Region, Gallery Item ID, Carousel Card or Country, are now blocked; reconfigure them. Saved datastreams using app_install_roas now use app_install_roas_double, and new app_install_reddit_* fields are available. For more information, see Reddit Ads: Setup guide.
Schedule fields removed from MCP datastream updates¶
Datastream write updates through MCP no longer accept schedule fields or the schedules list, and requests that include them are rejected. Use create_schedule to add a cron schedule and update_schedule to change one schedule. Update any calls that send schedules. For more information, see Using Adverity MCP.
API keys revoked when a user is deactivated¶
Deactivating a user now permanently revokes all of their API keys, including after reactivation, so integrations using those keys stop working. Workspace administrators can list, view and revoke keys issued in their workspaces. Create new API keys for affected integrations. For more information, see Authorizing Management API.
Notification subscription user can no longer be changed¶
API keys can no longer change the user of a notification subscription, and such requests now fail. Notification administrators and API keys with user write permission can update and delete the email, Slack and Microsoft Teams subscriptions they can see. Remove user changes from requests. For more information, see Configuring notifications in Management API.
CitrusAd and Pacvue connectors¶
CitrusAd (via BigQuery) and Pacvue (via BigQuery or Snowflake) are now available as standalone connectors. Set them up directly by name; data arrives through the BigQuery or Snowflake delivery route.
Criteo Retail Media offsite report¶
Criteo Retail Media has a new offsite report type with five placement dimensions: media type, ad format size, inventory type, publisher and supply-side platform. Use it alongside the existing onsite metrics to see offsite delivery. For more information, see Criteo Retail Media: Setup guide.
Facebook Ads 1-day click attribution windows¶
Facebook Ads datastreams can now select the 1d_click_all_conversions and 1d_click_first_conversion options in the Action attribution windows setting. For more information, see Facebook Ads: Setup guide.
Facebook Ads ad bid type field¶
Facebook Ads datastreams can now include the ad_bid_type field, which reports each ad’s bid type, such as ABSOLUTE_OCPM, CPC or CPM. For more information, see Facebook Ads: Setup guide.
Facebook Ads campaign-level frequency caps¶
Facebook Ads now reports campaign-level frequency caps, previously available only at ad set level. For Advantage+ campaigns, frequency settings no longer read as empty. For more information, see Facebook Ads: Setup guide.
Lazada date filter and local time¶
Lazada Order, Order Items and Products reports have a Date filter with Created at and Updated at options. Dates and request windows now follow the seller’s local time: Products dates shift by up to 8 hours, and Thailand, Vietnam and Indonesia request windows by one hour.
Lazada Sponsored Discovery campaign level¶
The Lazada Sponsored Discovery report now offers Campaign as a level option. Select it in the Level field to get campaign-level data.
Pinterest Audience Name field¶
The Pinterest Advertiser Delivery Metrics report has a new Audience Name field at the advertiser, campaign, ad group and pin promotion targeting levels. The ad group auto-targeting flag is available at the ad group targeting level. For more information, see Pinterest: Setup guide.
Pinterest Subscribe conversion metrics¶
The Pinterest Ads Analytics report type now offers the total, click-through, view-through and web Subscribe conversion metrics. For more information, see Pinterest: Setup guide.
Roundel placement level and search activity reports¶
Roundel has two new report subtypes. Placement Level returns Meta placement types, and Search Activity provides page type and page category dimensions. Campaigns without the needed data are filtered out and listed as warnings in the report.
Roundel search keyword level report¶
Roundel has a new Search Keyword Level report subtype. It provides keyword text and match type dimensions, which were not available in the connector before.
Spotify Ads artist fields¶
Spotify Ads has new Artist Id, Artist Name and Artist Image Url fields on the Ad Sets breakdown. They show the artists each ad set targets, with the largest available portrait image. For more information, see Spotify Ads: Setup guide.
Teads Ad Manager budget delivered fee value metric¶
Teads Ad Manager datastreams can now select the budget delivered fee value metric. Existing datastream defaults are unchanged.
MCP documentation search area filter¶
The MCP documentation search and answer tool now accepts an optional area value (all, product, mcp or api) to limit a search or answer to one part of the documentation. It returns Adverity Connect documentation only. For more information, see Using Adverity MCP.
Month offset for schedules in MCP¶
Custom schedules in MCP now include a run offset in months, from -12 to 24, applied together with the day offset. For more information, see Using Adverity MCP.
Monitor filter for notification subscriptions¶
Notification settings have a Monitors filter, so email, Slack and Teams subscriptions can receive data quality issues from selected custom monitors only. If the last selected monitor is deleted, the Data Quality topic is removed from the subscription. For more information, see Managing user notifications.
Run offset in months for custom schedules¶
Custom scheduling has a new Run offset (months) option next to the day offset. It shifts execution back by whole calendar months, accepts values from -12 to 24, and applies together with the day offset. For more information, see Manual, Smart, and Custom fetches.
Search in Column statistics¶
The Column statistics column list has a search field that filters columns by name, ignoring case. Ticked columns stay selected while hidden by the search, and Re-load selection still loads them.
Starting week for two-week and four-week schedules¶
Schedules set to Every 2 weeks or Every 4 weeks have a starting option in the custom schedule. It lists each possible cycle with its first runs, and a preview shows the next runs before saving. Schedules without a chosen week keep running in the same weeks. For more information, see Manual, Smart, and Custom fetches.
Terms of Use acceptance for new users only¶
Users who existed before this release are not asked to accept the Terms of Use. Only users created after the release are asked to accept them on first login.
Transformation audit trail in extract metadata¶
Extract metadata now records which jobs, configurations and tasks processed each extract. This helps trace extract processing when debugging; no action is required.
Used enrichments for conditional enrichments¶
Conditional enrichments now record which other enrichments they reference in a new used-enrichments field. Cloning an enrichment keeps these dependencies but does not copy datastream assignments.
Monitors field on notification subscriptions¶
Email, Slack and Microsoft Teams notification subscriptions have a new monitors field that limits data quality notifications to selected custom monitors. For more information, see Configuring notifications in Management API.
Month offset field for fetch schedules¶
Fetch schedules have a new offset_months field that shifts execution back by whole calendar months, from -12 to 24. It applies together with the day offset. For more information, see Scheduling fetches in Management API.
Starting week stored in not_before_date¶
For Every 2 weeks and Every 4 weeks schedules, the starting week is stored in not_before_date and does not delay runs. The API now clears not_before_date and not_before_time on other recurring schedules, where they had no effect. For more information, see Scheduling fetches in Management API.
Apple Search Ads multi-select Group by¶
Apple Search Ads Group by is now a multi-select, so Admin Area and Locality can be combined with the levels Apple requires. Saving names any missing or conflicting value. Datastreams grouped by Admin Area or Locality alone already fail and need Country Code added. For more information, see Apple Search Ads: Setup guide.
LinkedIn Ads API version 202609¶
LinkedIn Ads requests now use LinkedIn API version 202609 instead of 202602. No action or data migration is needed, and existing datastreams keep their configuration. For more information, see LinkedIn Ads: Setup guide.
Facebook Ads guidance for reach and frequency in MCP¶
The Facebook Ads guide in MCP now recommends a separate reach and frequency datastream, because they are not additive. It also recommends raw metrics (clicks, costs, impressions) instead of CTR, CPC and CPM for ad-level datastreams. For more information, see Using Adverity MCP.
Authentication error page for JWT sign-in¶
The page shown when a JWT sign-in link fails is now at /public/authentication-error/ and matches the sign-in page design. The old /go/error/ address redirects to it, and its Go to login page link opens the sign-in page.
Connection deletion refused with protected references¶
Deleting a connection that a protected reference still uses is now refused, even when deletions are scheduled. Large authorizations, workspaces and OAuth2 services now delete without failing, and delete previews show per-type counts beyond 1,000. Remove the reference first.
Current datastream label in Custom Script¶
In the Custom Script datastream picker, the datastream a transformation runs on is now labelled Current datastream, with __self__ as its caption. Picking it still inserts __self__, so existing scripts are unaffected. For more information, see Using custom script transformations.
Datastream deletion refused with protected dependencies¶
Deleting a datastream is now refused when a protected object depends on it, and the delete preview lists everything the datastream takes with it. Preview rows show entity type and connector. Remove the dependency before deleting.
Inline editing in the authorization Settings card¶
On an authorization’s details page, the Settings card now edits inline. Changed fields show a save bar with the number of changes, plus Save and Discard buttons. Pressing Enter saves, and leaving with unsaved changes asks for confirmation.
Legacy authorization pages removed¶
The legacy Reauthorize and Authorization Invite pages are removed, and their old addresses redirect to the matching authorization pages. The re-authorize dialog on the authorization details page and the request-access step replace them. No data is deleted.
NoCode inline workflow retries¶
Inline workflows in NoCode connectors no longer occupy a worker while waiting between retries. Connection errors and timeouts are still retried; a batch that fails for any other reason is skipped with a warning.
Webhook Payload URL hint¶
The webhook Payload URL field in Notifications now shows a hint to enter a public http or https address, and to contact Adverity Support for an internal host. For more information, see Managing webhooks.
Workspace switching without page reload¶
Switching workspace from the header now happens without a page reload, and the header and sidebar stay in place. The header’s workspace list stays current after a workspace is created, cloned or deleted.
API reference covers billing bodies¶
The API reference now describes the public billing request and response bodies and the billing objects CSV export. This is a documentation change.
API reference covers type-based endpoint bodies¶
The API reference now describes the request and response bodies of the type-based connection, datastream, destination and mapping endpoints. Regenerate clients built from the reference to pick up these types. For more information, see Getting started with the Management API.
API reference shows actual field types¶
The API reference now lists column data types and cron types as the labels the API sends and accepts, and shows the real types of datastream, job, extract, error and connection fields, including nullable ones. Behaviour is unchanged; regenerate clients built from the reference. For more information, see Getting started with the Management API.
Connection deletion returns 409 for protected references¶
The API connection delete now returns 409 for a connection that a protected reference still uses, even when deletions are scheduled. Remove the reference before deleting the connection. For more information, see Creating and configuring authorizations in Management API.
Datastream deletion refused in the API¶
Deleting a datastream through the API is now refused when a protected object depends on it. Remove the dependent object before sending the request. For more information, see Configuring datastreams in Management API.
Salesforce Audience Studio connector deprecated¶
The Salesforce Audience Studio connector is deprecated. Customers using it should plan to move off it.
Webtrends connector deprecated¶
The Webtrends connector is deprecated. Review existing Webtrends setups.
Adobe Analytics 2.0 metadata updates¶
Adobe Analytics 2.0 metadata updates now work for every report suite, refreshing metrics, dimensions and segments.
Adobe Customer Journey Analytics calculated metrics¶
Calculated metrics, including those created in other data views, can now be selected in the Adobe Customer Journey Analytics Metrics list.
BigQuery and Google Cloud Storage service account validation¶
BigQuery and Google Cloud Storage service account validation now accepts only allowed token URIs, so service account files pointing elsewhere are rejected.
Facebook Ads metadata with null characters¶
Facebook Ads metadata updates no longer fail on ad creatives whose texts contain null characters.
File destination JSON fields as nested JSON¶
File destinations with JSON or NDJSON output now write fields mapped to a JSON target field as nested JSON, not a string; values that are not valid JSON remain strings.
Fluent Commerce and WooCommerce authentication requests¶
Fluent Commerce and WooCommerce connectors no longer send requests to internal network addresses during authentication and credential validation. No setting needs to change.
Google Search Ads 360 report request timeout¶
Google Search Ads 360 report requests with no response now time out after one hour instead of four, so stalled fetches finish about three hours sooner.
Maileon connection timeout¶
Maileon extractions no longer fail on a transient connection timeout; the connection is retried instead.
OAuth2 authorization URL validation¶
OAuth2 connections can no longer be set up with attacker-controlled authorization URLs that make the authorization flow send requests to arbitrary external addresses.
SFTP file destination size mismatch¶
SFTP uploads to file destinations no longer fail with a size mismatch when the server moves or empties the file on arrival; the import finishes with a warning. Other size mismatches still fail the import.
Shopee fetches with empty pages¶
Long Shopee fetches no longer fail when empty pages succeed on retry, because only retries that make no progress count towards the limit.
Sklik fetch error message¶
When Sklik rejects a fetch, the fetch now fails with Sklik’s own error message instead of an internal error.
Snapchat interrupted report batches¶
Snapchat report batches that the API cuts off mid-response are now retried instead of failing the whole fetch.
TikTok Ads Smart+ ad names¶
Smart+ ad names now appear in the TikTok Ads Metadata report instead of being blank for Smart+ campaigns.
TikTok GMV Max default fields¶
New TikTok GMV Max datastreams at Product and Creative levels now save with their default fields; campaign metadata fields are no longer selected by default there.
Web Connect 404 error reporting¶
A Web Connect URL that returns HTTP 404 is now reported as a datastream configuration error for you to fix, and the message names the host that returned 404.
MCP authorization deletion outcome¶
Deleting an authorization through MCP now states that it may still appear in listings until the nightly cleanup, and the delete preview no longer claims a deletion goes through when another object uses it.
MCP authorization invite defaults¶
Inviting someone to authorize a connection through MCP no longer fails when recipients or the authorization name are omitted; recipients are notified by default and the authorization receives a generated name.
Missing implementation guide in MCP¶
Asking the MCP documentation tool for an implementation guide that does not exist now returns the list of available guides instead of an error.
Administration sidebar highlighting¶
The Administration sidebar now highlights only the current page, instead of keeping the first opened Administration page highlighted.
Bulk Retry for Match & Map warnings¶
Bulk Retry on the issues tables no longer discards a finished processing job when retrying a Match & Map warning.
Cancelled fetches recorded as cancelled¶
A fetch cancelled while its task is running is now recorded as cancelled, not failed, and no issue is shown for it.
Data exports with time values¶
Data exports containing time values no longer fail, so downloads of those exports succeed.
Deleting destinations from settings pages¶
The delete button on settings pages of destinations such as Google BigQuery, Azure Blob and File no longer fails, so these destinations can be deleted again.
Dropdown overlap and Clear button¶
Dropdown option lists no longer cover their own field when space below is limited, and the extract preview Clear button unmaps a target field without pressing Esc first.
Invalid credentials message for Lumar, MikMak and Moloco¶
Entering invalid credentials when setting up a Lumar, MikMak or Moloco authorization now shows an invalid credentials error on the form.
Job cleanup affecting other tenants¶
The hourly job cleanup no longer drops task tracking for other tenants’ running jobs, so healthy rescues are no longer failed as stalled and retry indicators show correctly.
Jobs after worker crash¶
When a worker crashes, the job is now rescued or failed even if the crashed task lost track of its job group, instead of staying in running.
Multi-select conditions in forms¶
Forms now evaluate “does not contain” and “selection count” conditions correctly for multi-select values, and an empty selection counts as containing none of the listed values.
Optional filters for fetches¶
The Optional Filters for this Fetch section works again on the Initial Fetch and Fetch Data pages, including Google Display & Video 360 Advertisers scoped to the selected Partners.
Private Custom Script operations¶
Custom Script instructions that use a private operation (names starting with !) are now reported as private and not configurable in the editor, not as no longer available.
Published extract counts on job groups¶
Job group published extract counts are now correct, so historical successful jobs no longer appear as jobs with no data.
Smart Schedule cadence in timeliness issues¶
Timeliness issues in the issue drawer, issues table and notifications now show the datastream’s actual Smart Schedule cadence instead of a fixed daily line.
Stalled rescue timeout extended¶
Rescued jobs are now failed as stalled only after six hours without activity instead of one, so rescues that are still waiting or finishing are no longer failed early.
Worker crash handling¶
When a worker crashes, only the job whose task crashed is handled, once; finished or discarded jobs are no longer changed to failed by a later crash handler.
API retry of Match & Map warnings¶
The API retry endpoints no longer discard a finished processing job when retrying a Match & Map warning, and return 409 instead of 500 when a message cannot be retried.