Databricks on Thursday, November 19, 2026, introduced CustomerLake Conversions, a feature designed to let brands tell advertising platforms which of their conversions are worth the most, using purchase history, margin and predicted lifetime value already held inside the Databricks data platform. Nine advertising platforms are named as launch partners: Amazon Ads, Google, LinkedIn, Meta, Microsoft Ads, Reddit, Snapchat, The Trade Desk and TikTok. Three agency groups sit alongside them: Epsilon, Merkle and Omnicom.

The announcement came in a blog post by Tasso Argyros, Katy Yuan and Cindy Jiang. It presents the feature as a new capability of CustomerLake Campaign Agents, part of the agentic customer data platform Databricks launched in June. There is no release date. According to Databricks, "When available, CustomerLake Conversions will be accessible to all customers using Campaign Agents in CustomerLake." Pricing is not mentioned.

In Short

Databricks has announced a tool meant to let companies tell ad platforms such as Meta, Google and Amazon Ads which sales and leads matter most, based on customer data the companies already keep in Databricks. It matters to brands and agencies that feed conversion data into automated bidding systems, because the feature moves that work to the place where customer records and prediction models already sit. There is no release date or price yet, so for now the practical change is a partner list and a described workflow, not a product that can be switched on.

What Databricks announced

The post opens with a familiar problem in performance marketing. Two purchases can look the same in a campaign report: one from a buyer who may become a loyal regular, the other a first-time order prompted by a deep discount. "Both count as conversions, but their long-term value can be very different," according to Databricks. The data that separates them, the post argues, sits with the advertiser: purchase history, margin, loyalty, engagement and predicted lifetime value.

CustomerLake Conversions is the company's attempt to carry that distinction into paid media. Databricks describes it as a capability that "connects first-party data and context directly to advertising activation and measurement" and says it "helps brands pursue more valuable conversions and improve return on ad spend (ROAS), while controlling the data they share."

The summary at the top of the post lays out the mechanics in one paragraph. Marketers define value-based conversion signals, such as a first purchase from a customer with high predicted lifetime value. Those signals, along with audiences, go to ad platforms through Conversions API integrations. Campaign performance data comes back into Databricks through OpenSharing and Lakeflow Connect. Brands and agencies work from one governed foundation in Unity Catalog, "with delegated permissions and an audit trail of what was sent where."

Role Partners named by Databricks
Advertising platforms (9) Amazon Ads, Google, LinkedIn, Meta, Microsoft Ads, Reddit, Snapchat, The Trade Desk, TikTok
Agency partners (3) Epsilon, Merkle, Omnicom

The components the post names, and the job each one does in the workflow, are set out below.

Component Function, according to Databricks
CustomerLake "the Agentic CDP built natively in Databricks"
Campaign Agents Build audiences in natural language, activate across channels, power 1:1 personalized experiences
Conversions tab Select conversion data, configure destinations, manage delivery
Audience and Conversions API (CAPI) integrations Deliver selected audiences and conversion signals to advertising platforms
Native Reverse ETL connectors The delivery route for those audiences and signals
OpenSharing integrations and Lakeflow Connect Bring campaign insights back into Databricks
Unity Catalog Secures and governs customer profiles and business context
Audit trail Records what was sent and where

How CustomerLake Conversions works

According to Databricks, the feature extends Campaign Agents to advertising performance by connecting three steps. The post describes them as follows.

  1. Define what matters. Teams use first-party data and machine learning models to build audiences and identify "meaningful conversion signals, such as a qualified lead or a first purchase from a customer predicted to be high value."
  2. Activate across paid media. Delivery runs through what Databricks calls Audience and Conversions API (CAPI) integrations, which "deliver selected audiences and conversion signals, through native Reverse ETL connectors, to advertising platforms."
  3. Bring results back. "OpenSharing integrations and Lakeflow Connect bring campaign insights back into Databricks, where teams can evaluate advertising performance alongside customer and business context."

The operating surface is a dedicated Conversions tab inside CustomerLake Campaign Agents. According to the post, the tab "lets marketers select conversion data, configure destinations, and manage delivery within the Campaign Agents workflow."

A diagram in the post, titled "How CustomerLake Conversions works," fills in the plumbing. On the left sits "Customer and business context," broken into customer profiles, transactions and customer-value models. That feeds two boxes inside Campaign Agents: Audiences, labelled "Choose who to reach," which flows out through "Audience sync," and Conversions, labelled "Define conversion events and values," which flows out through "CAPI integrations." Both lines end at a single box of nine advertising platforms that handle "Campaign delivery" and "Targeting and optimization." Campaign results then return "Via OpenSharing or Lakeflow Connect" to inform "the next audiences, signals, and media strategy." The caption reads: "The brand controls what is shared and with whom. The agency does its best work on better data."

The diagram draws both the audience line and the conversion line to all nine platforms as a group. It does not show which platforms receive which feed, and the post does not break that down by partner.

The argument: signals as the last lever

Behind the feature is a claim about where control now sits in automated advertising. "As advertising platforms automate more bidding, targeting, and campaign decisions, the only lever advertisers fully control is what signals they feed them," the post states. "That puts a premium on first-party data, and on the business context that explains what it means."

That is a position many in paid search and social would recognise. Value-based bidding prices each auction on predicted conversion value rather than on a count of conversions, so the quality of the value assigned to each event shapes where the money goes.

Databricks then describes the friction. Getting richer business context into paid media, according to the post, often requires "manual data extracts, custom integrations, and handoffs across marketing, analytics, engineering, and agency teams." Audiences can drift apart across platforms, and campaign reporting arrives late and disconnected from customer and revenue data. "That means attribution is fuzzy, acquisition costs creep up, and teams can't say with confidence which paid channels are converting and which aren't."

Two short examples follow. A retailer, according to the post, can use predicted lifetime value to adjust lookalike targeting quickly. A business-to-business company can use "qualified leads or closed won revenue to represent progress beyond a submitted form." The post adds a qualifier: "Marketers still define success; CustomerLake helps translate those priorities into signals advertising platforms can use."

The retailer example, before and after

The longest illustration in the post follows a retailer whose goal is to acquire more customers likely to buy again.

Stage Before, according to Databricks With CustomerLake Conversions, according to Databricks
Signals sent to platforms Purchase events and order values Completed first purchases from customers the lifetime value model flags as likely repeat buyers
Where the value model lives In the data platform, maintained by the data team The same model in Databricks, used directly by the paid media team
Delivery Data extracts, platform-specific uploads and separate delivery workflows Configured in the Conversions tab and sent to supported platforms through CAPI
Audiences Not described Built and synced from the same customer data
Evaluation The agency reconciles campaign reports with revenue data Campaign insights flow back into Databricks, where retailer and agency compare acquisition spend with customer value

The outcome Databricks claims is qualitative. "The ad platform can better optimize what a good customer looks like," the post says, and the retailer "gets a faster, clearer view of which campaigns acquire the customers it wants to grow with." No figures accompany the example, and no customer is named.

Agencies inside the client's account

A large part of the post is aimed at agencies, which, according to Databricks, provide "core strategy, identity assets, models, and media expertise" for many enterprises. The central promise: "Agencies get governed access to customer context in the lakehouse, without copying or duplicating data." The brand decides what is shared and with whom.

The post sets out what that could look like in three steps.

  1. Configure delivery. Agencies can set up destinations and conversion definitions in the client's Conversions tab, "using permissions the brand delegates, with an audit trail of what was sent where."
  2. Apply unique intelligence. Agencies can combine their own identity data and predictive models, such as lifetime value and propensity scores, with a brand's first-party data. The post singles out one partner: Epsilon's identity graph, which according to Databricks "can connect conversion events to customer identities even when the original event has limited identifying information." Adding those identifiers to CAPI payloads, the post says, "can help ad platforms match more conversions."
  3. Measure results. "Campaign insights from the platforms feed the agency's MMM and incrementality studies," the post states, referring to marketing mix modelling and experiments that test what a campaign actually caused.

The agency names carry weight of their own. Epsilon belongs to Publicis, which on November 2 closed its acquisition of LiveRamp and launched an engine combining LiveRamp's platform with Epsilon's identity assets. Publicis had agreed in May to buy LiveRamp at an equity value of $2.5 billion. Omnicom absorbed IPG after the FTC finalized restrictions on the $13.5 billion acquisition in September 2025.

The Epsilon example also points to a trade-off the post does not dwell on. Enriching conversion events with identifiers from an agency's graph is meant to raise match rates, and it does so by sending more identity data to the platforms. That is the stated purpose. How consent is checked, whether identifiers are hashed, and how long platforms keep the data are not covered.

Who controls what is shared

Databricks lists four design commitments under the heading "Customer context, shared on your terms." Because they define what the product is meant to allow, they are reproduced here word for word.

Commitment Databricks wording
Governance "Your data stays governed. Customer profiles and business context are secured and governed by Unity Catalog."
Destinations "You choose the destinations. You decide which audiences and conversion events go to which platforms."
Visibility "You can see what was sent. An audit trail provides visibility into what was sent and where."
Human oversight "Humans stay in the loop. Your team sets the strategy, goals and guardrails. Agents help scale execution."

These are commitments about the Databricks side of the pipe. Once a conversion event reaches a platform, the platform's own terms apply, and the post does not address that handover.

Where CustomerLake came from

Databricks launched CustomerLake on June 16, 2026 as an agentic customer data platform that handles identity resolution, audience segmentation and campaign activation on governed data inside Databricks, without copying records into a separate system. IAS was the launch partner, linking first-party records to its contextual and attention signals through clean-room connections. PPC Land's June coverage listed Lakeflow, Unity Catalog, Genie, a Real-Time Profile API, Agentic Identity Resolution and a third-party data marketplace among the components, and it described "infinity campaigns" as agent-driven engagement loops. Conversion APIs did not feature.

The November post closes that gap and returns to the earlier phrase. CustomerLake Conversions, according to Databricks, "brings paid media into the broader vision of Infinity Campaigns: continuous engagement that learns and adapts as customer context changes."

The post also makes a claim about speed. "A single campaign can take 6+ teams and 60+ days to launch. By then, customers have moved on." Databricks gives no source or method for those figures.

A crowded pipe

None of the nine platforms needs Databricks to receive server-side conversions. Each runs its own route, and PPC Land has tracked several of them.

Platform or body What PPC Land reported Date
LinkedIn Conversions API for online and offline conversions, data held for at most 180 days November 2023
Google Data Manager API covering Google Ads, Google Analytics and Display & Video 360, up to 2,000 conversion events per request December 2025
Meta Free one-click Conversions API, web only, mirrors Pixel events April 2026
IAB Tech Lab ECAPI 1.0, 27 standard event types plus 15 additional May 2026
Microsoft Conversions API gated behind per-account pilot approval, batches of up to 1,000 events September 2026

The pressure to wire up these feeds has been building. An IAB guide in late 2025 found that 72% of publishers cited technical complexity as a major obstacle to conversion APIs, while two-thirds of advertisers reported improved ROAS after adopting them. Streaming has joined in too: Netflix announced its own conversion API in March and in October expanded it to more than 20 lower-funnel events, in the same announcement that gave Trade Desk buyers more than 4 million data segments for Netflix ads. In early November, IAB Tech Lab published final guidelines for sending ECAPI events into data clean rooms.

Warehouse-native activation is not new either. StackAdapt in July 2025 unveiled a Snowflake Native App that let marketers map Snowflake tables to its platform and build segments for retargeting, lookalikes and suppression without moving the data out first.

So what does Databricks add? A claim to the layer in between: one place where value definitions are built, sent to many platforms and measured afterwards, with agencies working inside the brand's environment rather than receiving exports.

Why this matters for marketers

For advertisers who already store customer data in Databricks, the pitch is consolidation. The model that predicts lifetime value, the decision about which purchase counts as a valuable conversion, and the campaign results used to judge spend would sit in one governed system rather than across exports, uploads and agency spreadsheets. That is the gap the post describes, and it is a real one for teams that run value-based bidding across several platforms at once.

The partner list is broad. It covers search, social, professional networking, retail media and an independent demand-side platform, which suggests Databricks has secured at least a working relationship with each. Yet a launch partnership is not the same as a shipped integration, and the post describes availability only in the conditional.

For agencies, delegated permissions inside the client's account change the shape of the work. The agency configures delivery and applies its own identity data and models, but on infrastructure the brand owns, with an audit trail the brand can read. That arrangement favours groups with their own identity assets, such as Epsilon, which the post names directly.

For the platforms, the stated benefit is more and better-matched conversion data. That is also where the open questions sit. The more identity data and business context flows into platform optimisation systems, the more those systems learn about a brand's best customers.

What the post does not say

The post gives no release date, no price, no performance results and no named customers. It does not say which Google product or API the integration uses, which of the nine platforms receive audiences, conversions or both, or how consent, hashing and data retention are handled for the identifiers it describes adding. OpenSharing is named but not explained, and the "6+ teams and 60+ days" figure carries no source.

Key facts

  • Product: CustomerLake Conversions, a capability of CustomerLake Campaign Agents in Databricks CustomerLake
  • Announced: Thursday, November 19, 2026
  • Source: Databricks blog post "Introducing CustomerLake Conversions: Improve advertising performance with first-party customer context," by Tasso Argyros, Katy Yuan and Cindy Jiang, category Platform, tags Marketing and Announcements
  • Source URL: https://www.databricks.com/blog/customerlake-conversions
  • Availability: no date; "When available," for all customers using Campaign Agents in CustomerLake; access through the Databricks account team
  • Price: not stated
  • Advertising platform partners (9): Amazon Ads, Google, LinkedIn, Meta, Microsoft Ads (shown as Microsoft Advertising in the partner graphic), Reddit, Snapchat, The Trade Desk, TikTok
  • Agency partners (3): Epsilon, Merkle, Omnicom
  • Outbound routes: Audience sync and Conversions API (CAPI) integrations via native Reverse ETL connectors
  • Return routes: OpenSharing integrations and Lakeflow Connect
  • Governance: Unity Catalog, delegated permissions, audit trail
  • Interface: Conversions tab within CustomerLake Campaign Agents
  • Databricks claim on campaign setup: "6+ teams and 60+ days" for a single campaign, no source given
  • Parent product launch: CustomerLake, June 16, 2026

Timeline

Summary

Who: Databricks, through a blog post by Tasso Argyros, Katy Yuan and Cindy Jiang, with nine advertising platform partners (Amazon Ads, Google, LinkedIn, Meta, Microsoft Ads, Reddit, Snapchat, The Trade Desk, TikTok) and three agency partners (Epsilon, Merkle, Omnicom).

What: CustomerLake Conversions, a capability of CustomerLake Campaign Agents that lets marketers define value-based conversion signals from first-party data and models, send them and audiences to ad platforms through CAPI and Reverse ETL connectors, and bring campaign results back into Databricks through OpenSharing and Lakeflow Connect, under Unity Catalog governance with delegated agency permissions and an audit trail.

When: Announced Thursday, November 19, 2026. No release date was given; the post says the feature will reach all Campaign Agents customers "when available." CustomerLake itself launched on June 16, 2026.

Where: Inside the Databricks data platform, in a dedicated Conversions tab within CustomerLake Campaign Agents, with outbound connections to the nine partner platforms.

Why: Databricks argues that as ad platforms automate bidding and targeting, the signals advertisers send are the main lever left to them, and that moving signal creation to where customer data and lifetime value models already live replaces manual extracts, platform-specific uploads and delayed reconciliation. The post offers no performance data, pricing or customer examples to support the case.