Market Performance Group
How Claude turned MPG's own market data into a cross-sell engine
Industry
Consumer goods services
Engagement
Project Compass
Stack
Claude + Salesforce
Model
Claude API for batch scoring; Claude Projects & Skills with Salesforce MCP
The challenge
What they came to us with
MPG sells across multiple service verticals, and its biggest growth opportunity sits inside its existing book: accounts active in one vertical that should be in six.
Identifying that whitespace meant manually cross-referencing Salesforce vertical data against MPG SmartSights™ market intelligence — a consolidated dataset no rep had time to read account by account, and no rules engine could reason about.
Prospecting had the same gap. MPG's data could describe its ideal customer precisely, but nothing turned that profile into a ranked list of net-new targets with a reason attached.
Our approach
How the work was sequenced
01
Claude as the scoring core
Claude reads each account's CRM vertical penetration alongside SmartSights™ market intelligence and returns a 1-100 score plus a written justification naming which verticals to pitch and why. The judgment lives in the model, not in a brittle scoring formula.
02
Batch scoring on the Claude API
Scoring runs as a batch job against the Claude API on MPG's data cycle, so every account carries a fresh score and rationale before a rep ever opens Salesforce — and rep-side retrieval costs nothing.
03
Claude Projects & Skills over Salesforce MCP
The rep-facing workflow is a Claude Project with purpose-built Skills connected to Salesforce through MCP. Reps see their five best cross-sell accounts one at a time for accept, reject, or hold, and Claude writes decisions straight back to the CRM.
04
Prospecting from the best customers
Claude derives an ICP from MPG's strongest accounts, finds matching non-customers in the SmartSights™ data, and creates accepted prospects directly as Salesforce Accounts.
05
Matching reps can trust
Identity resolution stays deterministic: a strict three-tier match on real identifiers rather than letting the model guess at names — a wrong-account match costs more trust than a missing row. Rejections route to vertical VPs through Salesforce dashboards instead of disappearing.
The results
What changed
Every account scored by Claude
With a written justification, so a rep gets an opening line rather than just a ranking.
Whitespace identification
Manual cross-referencing replaced by a ranked query against Claude-scored penetration and market data.
Rep usage consumed
Claude runs in batch on the data cycle; retrieval is a plain Salesforce query.
What's next
Extending Claude's scoring core to more verticals and deepening the rep-facing workflow in Claude Projects and Skills over the Salesforce MCP integration.
Want results like this?
Tell us what's broken. We'll tell you how to fix it.
Book a free strategy call