Open rates, click-through rates, bounce rates, and conversion percentages have long been the name of the game. We’ve built entire strategies around improving these numbers. 

And, barring opens, they’re still the most important metrics for any business anywhere. 

But at the same time, it is important to realize that the shift to agentic marketing, where agents handle dynamic, two-way conversations, demands an entirely new analytics paradigm.

One that also takes charge of attribution. 

In response to both these realities, Salesforce built two interconnected platforms: 

• Marketing Intelligence for unified campaign analytics
• Agentforce Observability for conversation-level insights

Is Agentforce integral to your MOps? In that case, you need to rethink your approach to analytics. Let us first consider in more detail why you need Agentforce analytics. 

The problem with limitations of traditional metrics

Traditional metrics were built for a broadcast model: you send a message, measure who opened it, track who clicked, and count conversions. For marketing that was exclusively about pushing content to an audience, you could not ask for a more sensible and practical measurement model. 

And mind you, these still give the best 360-degree hilltop view of your campaigns. 

According to a survey of 400 marketers by Adobe Business (2025), CTR and bounce rates rank just below brand awareness metrics in importance. But with the advent of agentic, two-way (email) marketing, the number of facets have increased. There’s a lot more information, more interaction layers, and more dimensions to the overall customer experience. Today, for you to know the complete story, you’ll have to train your sights at:

• Conversation quality: A customer might reply to an email, but did the agent understand their intent? Did it provide value?
• Multi-turn engagement: The conversational interface requires you to track the depth of every interaction an agent has with a customer. 
• Autonomous resolution: Did the agent successfully complete a goal without escalating to its human counterpart?
• Sentiment evolution: How did the customer sentiment shift during an interaction? 

Now one might wonder, Why do I need these if traditional metrics can still get the job done? 

Because the scope of customer experience is expanding at the speed of light. The journey from cold to click now includes way more touchpoints than ever before. 

As Sandeep Garg pointed out three years ago, “A decade ago, we may have collected surveys for product feedback or brand perception and looked at NPS to get an understanding of loyalty. We now have terabytes of data and feedback coming in from social media, reviews, call center discussions, chat conversations and recordings that give us an abundance of information about how a customer feels about a brand or whether they are happy or not.”

(NPS, for example, has been one of the most active fault lines in marketing in recent years.)

Agentforce gives you the toolkit to traverse the extended scope of customer experience. Today, you need to go granular.  

Agentforce Observability, the main pillar of Agentforce analytics, allows you to do that. 

What is Agentforce Observability?

Agentforce Observability is a monitoring and diagnostic suite designed specifically for AI agents. It provides deep visibility into how individual agents are performing. 

Salesforce has organized Agentforce Observability into three distinct functional areas:

1. Agent Analytics: Provides a comprehensive view of how every agent is performing in real customer interactions, and surfaces KPI trends over time. 
2. Agent Optimization: Records the steps in the agent’s reasoning chain, allowing you to learn why and how a particular agent behaved in a certain way. 
3. Agent Health Monitoring: Gives you real-time visibility into agent health. You can visualize performance trends instantly and get alerted to ‘silent failures’ when they happen. 

With Agent Health Monitoring, you can configure custom thresholds for any of your core metrics.

Agentforce Observability Features
Feature	Primary focus	Capabilities	Key metrics
Agent Analytics	High-level business impact and ROI	Tracks usage and effectiveness trends over time; identifies ineffective topics/flows; provides a comprehensive view of all deployed agents	Deflection Rate, Abandonment Rate, Escalation Patterns, Volume Trends, Quality Scores
Agent Optimization	Session-level diagnostics and reasoning-chain clarity	Records every LLM call, tool invocation, and guardrail check; groups conversations by intent/sentiment; full session traces with step-by-step latency	Session-level visibility, User Utterances, Tool Invocations, Automated Intent Clustering
Agent Health Monitoring	Real-time reliability and proactive alerting	Eliminates black box guesswork; identifies “silent failures” instantly; provides a transparent dashboard for immediate troubleshooting	Agent Error Rate, Average Interaction Latency, Escalation Rate

When you receive an alert, Agent Health Monitoring allows you to investigate using investigation flows based on session traces. Metrics are built on top of the Session Trace Data Model. You can drill down into the actual interaction logs. You can review the breakdown across all topics, steps, and across your agents and channels to identify exactly where the conversation failed. 

In Agentforce Observability, a deflected session is not a failure. It refers to a session that was resolved successfully by the agent without requiring escalation to a human. In fact, deflection can also mean the agent redirected a query it couldn’t handle  toward a relevant FAQ, then answered from there. 

With 1 in 3 agents lacking customer context and 1 in 5 consumers seeing no benefit from AI bots, the observability suite becomes critical. It tracks deflection quality to determine whether agents are truly helpful or simply passing the buck.

Understanding conversation moments and quality scoring

Agentforce Optimization breaks down conversations into discrete interactions within a session, or “moments”. Imagine an agent experience where customers can ask diverse questions in a single session, from account inquiries to technical issues and future service needs. Agentforce Optimization allows you to evaluate the quality of such multifaceted sessions and identify specific areas for improvement. Here’s how moments are evaluated:  

• Intent Recognition: Did the agent correctly understand what the customer wanted?
• Response Relevance: Was the agent’s response directly relevant to the question?
• Action Completion: Did the agent successfully execute the required action?
• Sentiment Trajectory: How did customer sentiment evolve during the conversation?
• Guardrail Compliance: Did the agent stay within defined policy boundaries?

Now then, with Agentforce Observability features, you are well positioned for success in customer experience. However, success must be measured accurately and credited to the right teams. This brings us to how Agentforce addresses attribution.

The problem of attribution in marketing 

Attribution has beenone of the biggest thorns in the side of marketing. According to the State of Marketing Attribution (2024) by Revsure.ai, lack of resources (46%) and complexities (46%) are the topmost challenges that marketers face in attribution. The challenge has only increased in recent years, although the core issues remain more or less the same: 

• Marketing data is scattered across multiple platforms with their own attribution logic. As a result, you get conflicting insights and never a unified view of performance. 
• Attribution models are imperfect. They simplify reality and, willy-nilly, introduce bias. 
• Modern customer behavior is multi-channel, cross-device, and influenced by aspects both measurable and unmeasurable. 
• Teams operate in siloes and compete for credit. Stakeholders lack understanding or trust in attribution logic. As a result, attribution mutates into an organizational problem. 
• Performance is influenced by factors that attribution models struggle to capture, e.g. offline channels, seasonal factors, economy, weather, etc. 

Even when marketers succeed in breaking silos and gaining more visibility into the sales pipeline, the priority shifts to achieving end-to-end attribution and accurately measuring ROMI. 

Marketing Intelligence is Salesforce’s answer to the long-standing demand for unified analytics.

What is Marketing Intelligence?

Marketing Intelligence is Salesforce’s next-generation, AI-driven analytics platform built natively on Data Cloud and the Agentforce 360 Platform. It brings together the performance of Hyperforce, data ingestion and unification of Data Cloud, AI capabilities of Agentforce, and the visualizations of Tableau Next. 

The 4 components of Marketing Intelligence
Component	Function
Hyperforce	Delivers scalable, cloud-native performance
Data 360 (Data Cloud)	Enables real-time data ingestion and unification
Tableau Next	Provides next-gen visual dashboards with smarter filtering
Agentforce	Drives autonomous, AI-powered orchestration with proactive insights

Marketing Intelligence unlocks a new interface paradigm: analytics by conversation. Users can describe reporting requirements in natural language, and the system will

• build dashboards
• generate performance summaries
• suggest optimizations

End-to-end marketing attribution in Salesforce

Once thought of as the holy grail of marketing analytics, end-to-end attribution is one of the key selling points of Marketing Intelligence. Previously, analytics platforms have focused strongly on either first-party or third-party data sources, forming an incomplete view of performance at best.

Marketing Intelligence has the full range of paid and organic marketing actions coupled with customer and sales funnel insights originating from Salesforce CRM and Data Cloud. This means Marketing Intelligence is able to provide marketing attribution and ROMI.

Now, please understand that Marketing Intelligence does not entirely solve the problem of attribution. It acts as a data unification and integration layer, bringing together data across Salesforce products via a shared semantic layer. It enables you to ingest, enrich, and visualize unified marketing data, which is a prerequisite for better attribution.

Salesforce employs a number of attribution models, but they all have their own limitations. In the experience of Romain Blanc, founder of Heeet, these include: 

• Significant manual effort, which is ironic before anything else. 
• Complex dashboards that hinder the discovery of insights. 
• Enterprise-grade attribution capabilities gated behind high costs. (Marketing Intelligence typically requires both a Data Cloud license and a Tableau Next license, making it a more substantial investment compared to a standalone Marketing Cloud Intelligence license.)

“Salesforce provides essential attribution tools, without the enormous cost of custom attribution, but it is up to your team to integrate them,” Blanc points out. 

What MI excels at is delivering a unified view of analytics. It’s a critical advantage, given that less than 40% of marketers even know where their customer data is stored! 

How to build your analytics stack in Salesforce?

Transitioning from traditional analytics to agentic analytics requires a phased approach. Here is a roadmap for you to build your analytics stack in Salesforce. 

Agentforce analytics stack roadmap
Phase	Focus	Key actions
1	Data foundation	• Audit data sources and identify gaps. • Enable Data Cloud and configure streams. • Connect 3rd-party platforms via connectors. • Set governance and quality standards.
2	Marketing Intelligence setup	• Configure Marketing Intelligence & attribution models. • Define business-specific KPIs and goals. • Build Tableau Next dashboards. • Enable Agentforce Paid Media Optimization.
3	Implement Agentforce Observability	• Deploy low-risk agents. • Enable Agent Analytics in Agentforce Studio. • Configure session tracing and custom tags. • Define quality thresholds and escalation triggers.
4	Baseline and optimize	• Collect metrics across all platforms. • Identify weak conversations/campaigns. • Refine agent instructions via session traces. • Update knowledge bases and train teams.
5	Scale and improve	• Expand agents to complex use cases. • Establish a weekly performance review cadence. • Use AI insights for spend allocation. • Monitor health trends and adjust guardrails.

Step into Agentforce analytics with Mavlers!

To those coming from a deep background in legacy Marketing Cloud Intelligence, rest assured, the business knowledge you built is entirely transferable. 

The most practical first step is to gain a working understanding of Data Cloud before stepping into Marketing Intelligence. Given that MI is built on Data Cloud’s semantic layer and relies on its connection and identity resolution infrastructure, a foundational Data Cloud certification will give you the architectural fluency to use MI effectively from the word go. 

If you need guidance with AI marketing analytics in Salesforce, we can help. With over 10 years of experience in serving more than 800 SFMC clients, we can be your go-to execution partner. 

Book a free, no-obligation call with one of our Salesforce Marketing Cloud experts.