Lead Matching

Match sales leads across Salesforce, HubSpot, and LinkedIn Sales Navigator to eliminate duplicates, consolidate lead intelligence, and route the richest profile to your sales team.

The Problem

Your sales team uses three lead sources: inbound leads captured in HubSpot from marketing campaigns, outbound prospects in Salesforce from the SDR team, and enrichment data from LinkedIn Sales Navigator. The same prospect often enters all three systems:

• A prospect fills out a form on your website (HubSpot) and an SDR independently reaches out on LinkedIn (Salesforce). Neither team knows the other has made contact.
• LinkedIn enrichment data uses the prospect's full legal name ("Robert James Smith"), while the CRM has the name they signed up with ("Bob Smith").
• Company names vary between systems: "Acme Corp" in the CRM, "Acme Corporation" in LinkedIn, "ACME" in the marketing form.

Without deduplication, your SDR sends a cold email to someone who is already an active marketing lead. The prospect sees two outreach emails from the same company and loses trust.

The Data

Salesforce (Outbound Prospects)

lead_id	name	email	company	title	phone	lead_source	created_date
sf_lead_001	Bob Smith	[email protected]	Acme Corp	VP Engineering	+1-555-0201	outbound	2026-01-08
sf_lead_002	Jenny Lee	[email protected]	Globex	Product Manager	null	outbound	2026-01-10
sf_lead_003	Michael Chen	[email protected]	Initech	CTO	+1-555-0203	referral	2026-01-12
sf_lead_004	Sarah Connor	[email protected]	Cyberdyne Systems	Director of Ops	+1-555-0204	outbound	2026-01-14
sf_lead_005	Dave Wilson	[email protected]	Umbrella Corp	Head of Data	+1-555-0205	inbound	2026-01-15
sf_lead_006	Amy Zhang	[email protected]	Soylent Corp	Data Engineer	null	outbound	2026-01-16
sf_lead_007	Tom Harris	[email protected]	Wayne Enterprises	VP Sales	+1-555-0207	referral	2026-01-17

HubSpot (Inbound Marketing Leads)

contact_id	full_name	email_address	company_name	job_title	phone_number	lead_score	form_submitted
hub_44201	Robert Smith	[email protected]	Acme Corporation	VP of Engineering	null	85	demo_request
hub_44202	Jennifer Lee	[email protected]	Globex Corporation	PM	+1-555-0302	72	whitepaper
hub_44203	Mike Chen	[email protected]	Initech Inc	Chief Technology Officer	+1-555-0203	91	demo_request
hub_44204	S. Connor	[email protected]	Cyberdyne	Dir. Operations	+1-555-0204	45	newsletter
hub_44205	Lisa Park	[email protected]	Oscorp Industries	Data Analyst	+1-555-0305	68	whitepaper
hub_44206	Amy Z.	[email protected]	Soylent	null	+1-555-0306	34	newsletter
hub_44207	David Wilson	[email protected]	Umbrella Corporation	Head of Data Engineering	+1-555-0205	77	case_study

LinkedIn Sales Navigator (Enrichment)

profile_id	display_name	work_email	organization	headline	location	connections
li_prof_901	Robert James Smith	[email protected]	ACME	VP Engineering at ACME	San Francisco, CA	1247
li_prof_902	Jennifer Lee	[email protected]	Globex Corporation	Product Manager at Globex	New York, NY	832
li_prof_903	Michael J. Chen	[email protected]	Initech, Inc.	CTO at Initech	Austin, TX	2103
li_prof_904	Sarah J. Connor	[email protected]	Cyberdyne Systems Inc	Director of Operations	Los Angeles, CA	654
li_prof_905	Thomas Harris	[email protected]	Wayne Enterprises	VP Sales at Wayne Enterprises	Chicago, IL	1876
li_prof_906	Patricia Nguyen	[email protected]	Stark Industries	Head of Analytics	Seattle, WA	943

Data Challenges by Prospect

Prospect	Salesforce	HubSpot	LinkedIn	Challenge
Bob/Robert Smith	[email protected]	[email protected]	[email protected]	3 different emails, name as "Bob" / "Robert" / "Robert James"
Jenny/Jennifer Lee	[email protected]	[email protected]	[email protected]	3 different emails, "Jenny" / "Jennifer" / "Jennifer"
Michael Chen	[email protected]	[email protected]	[email protected]	3 different emails, "Michael" / "Mike" / "Michael J."
Sarah Connor	[email protected]	[email protected]	[email protected]	3 different emails, "Sarah" / "S." / "Sarah J."
Dave/David Wilson	[email protected]	[email protected]	not in LinkedIn	2 different emails, "Dave" / "David"
Amy Zhang	[email protected]	[email protected]	not in LinkedIn	2 different emails, name truncated in HubSpot
Tom Harris	[email protected]	not in HubSpot	[email protected]	2 different emails, "Tom" / "Thomas"
Lisa Park	not in Salesforce	[email protected]	not in LinkedIn	Only in HubSpot (singleton)
Patricia Nguyen	not in Salesforce	not in HubSpot	[email protected]	Only in LinkedIn (singleton)

Not a single prospect has the same email address across all their sources. Exact email matching alone would fail to deduplicate any of these leads.

The Spec

api_version: kanoniv/v1
identity_version: "1.0"

entity:
  name: lead
  description: "Sales lead deduplication across CRM, marketing, and enrichment platforms"

sources:
  - name: salesforce
    adapter: csv
    location: data/salesforce_leads.csv
    primary_key: lead_id
    attributes:
      name: name
      email: email
      company: company
      title: title
      phone: phone
      lead_source: lead_source
      created_date: created_date

  - name: hubspot
    adapter: csv
    location: data/hubspot_contacts.csv
    primary_key: contact_id
    attributes:
      name: full_name
      email: email_address
      company: company_name
      title: job_title
      phone: phone_number
      lead_score: lead_score
      form_submitted: form_submitted

  - name: linkedin
    adapter: csv
    location: data/linkedin_navigator.csv
    primary_key: profile_id
    attributes:
      name: display_name
      email: work_email
      company: organization
      title: headline
      location: location
      connections: connections

rules:
  # Email exact match. Despite different email addresses per person,
  # some pairs may share an email (e.g., future data loads).
  # Weight is highest because email is the most reliable identifier.
  - name: email_exact
    type: exact
    field: email
    weight: 1.0

  # Fuzzy name matching is critical here because every prospect
  # has name variations. Jaro-Winkler at 0.88 catches:
  #   "Bob Smith" vs "Robert Smith" -> 0.78 (MISS -- below threshold)
  #   "Jennifer Lee" vs "Jenny Lee" -> 0.89 (HIT)
  #   "Michael Chen" vs "Mike Chen" -> 0.86 (MISS -- just below)
  #   "Michael Chen" vs "Michael J. Chen" -> 0.93 (HIT)
  # Some pairs rely on corroborating rules to bridge the gap.
  - name: name_fuzzy
    type: similarity
    field: name
    algorithm: jaro_winkler
    threshold: 0.88
    weight: 0.6

  # Company name fuzzy match corroborates name matches.
  # "Acme Corp" vs "Acme Corporation" -> 0.92 JW (HIT)
  # "Acme Corp" vs "ACME" -> 0.76 JW (MISS)
  # "Globex" vs "Globex Corporation" -> 0.85 JW (MISS)
  # "Initech" vs "Initech Inc" -> 0.91 JW (HIT)
  - name: company_fuzzy
    type: similarity
    field: company
    algorithm: jaro_winkler
    threshold: 0.90
    weight: 0.3

  # Phone exact match as a strong corroborating signal.
  # Bridges records when email and name both differ.
  - name: phone_exact
    type: exact
    field: phone
    weight: 0.8

blocking:
  strategy: phonetic
  keys: [name]

decision:
  scoring: weighted_sum
  thresholds:
    match: 0.85
    review: 0.55
  review_queue:
    enabled: true
    sla_hours: 336    # 14 days

survivorship:
  default: most_complete
  overrides:
    - field: email
      strategy: source_priority
      priority: [salesforce, hubspot, linkedin]
    - field: company
      strategy: source_priority
      priority: [linkedin, salesforce, hubspot]
    - field: name
      strategy: source_priority
      priority: [linkedin, salesforce, hubspot]

metadata:
  owner: [email protected]
  tags: [lead-matching, sales, dedup]
  description: "Cross-platform lead deduplication with review queue"

How the Rules Score Each Prospect

Bob/Robert Smith (Acme)

Pair	email_exact	name_fuzzy	company_fuzzy	phone_exact	Total	Decision
SF / HubSpot	0.0	0.0 (0.78 JW)	0.3 (0.92 JW)	0.0 (null)	0.3	Reject
SF / LinkedIn	0.0	0.0 (0.72 JW)	0.0 (0.76 JW)	0.0 (no field)	0.0	Reject
HubSpot / LinkedIn	0.0	0.6 (0.91 JW)	0.0 (0.76 JW)	0.0 (no field)	0.6	Review

"Bob Smith" vs "Robert Smith" fails the name threshold. But "Robert Smith" vs "Robert James Smith" scores 0.91. The HubSpot-LinkedIn pair enters the review queue at 0.6, and a human reviewer can confirm the match and pull in the Salesforce record transitively.

Jenny/Jennifer Lee (Globex)

Pair	email_exact	name_fuzzy	company_fuzzy	phone_exact	Total	Decision
SF / HubSpot	0.0	0.6 (0.89 JW)	0.0 (0.85 JW)	0.0 (null)	0.6	Review
SF / LinkedIn	0.0	0.6 (0.89 JW)	0.3 (0.92 JW)	0.0 (no field)	0.9	Match
HubSpot / LinkedIn	0.0	0.6 (0.95 JW)	0.3 (0.92 JW)	0.0 (no field)	0.9	Match

"Jenny Lee" vs "Jennifer Lee" scores 0.89 in Jaro-Winkler. LinkedIn and both CRM sources match. All three records cluster together.

Michael Chen (Initech)

Pair	email_exact	name_fuzzy	company_fuzzy	phone_exact	Total	Decision
SF / HubSpot	0.0	0.0 (0.86 JW)	0.3 (0.91 JW)	0.8	1.1	Match
SF / LinkedIn	0.0	0.6 (0.93 JW)	0.3 (0.91 JW)	0.0 (no field)	0.9	Match
HubSpot / LinkedIn	0.0	0.0 (0.82 JW)	0.3 (0.91 JW)	0.0 (no field)	0.3	Reject

Phone match bridges SF-HubSpot despite "Michael Chen" vs "Mike Chen" being below the name threshold. SF-LinkedIn matches via "Michael Chen" vs "Michael J. Chen". HubSpot-LinkedIn ("Mike Chen" vs "Michael J. Chen") rejects, but transitive closure through Salesforce clusters all three.

Sarah Connor (Cyberdyne)

Pair	email_exact	name_fuzzy	company_fuzzy	phone_exact	Total	Decision
SF / HubSpot	0.0	0.0 (0.71 JW)	0.0 (0.87 JW)	0.8	0.8	Review
SF / LinkedIn	0.0	0.6 (0.92 JW)	0.3 (0.95 JW)	0.0 (no field)	0.9	Match
HubSpot / LinkedIn	0.0	0.0 (0.68 JW)	0.0 (0.87 JW)	0.0 (no field)	0.0	Reject

"Sarah Connor" vs "S. Connor" scores only 0.71. But phone bridges SF-HubSpot into review, and SF-LinkedIn matches cleanly. The review queue catches the HubSpot record.

The Result

Match Summary

Sources:              3
Total records:        20
Clusters:             9
  - 3-record clusters: 3  (Jennifer Lee, Michael Chen, Sarah Connor)
  - 2-record clusters: 3  (Dave Wilson, Amy Zhang, Tom Harris)
  - Singletons:        3  (Lisa Park, Patricia Nguyen, Bob Smith partially)
Match pairs:          7
Review queue:         4
Rejected pairs:       9

Matched Leads (Automatic)

Cluster	Lead	Sources	Confidence	Match Path
1	Jennifer Lee	SF + HubSpot + LinkedIn	0.90	name_fuzzy + company_fuzzy
2	Michael Chen	SF + HubSpot + LinkedIn	0.93	phone_exact (SF-HB) + name_fuzzy (SF-LI)
3	Sarah Connor	SF + LinkedIn	0.90	name_fuzzy + company_fuzzy
4	Dave Wilson	SF + HubSpot	0.86	phone_exact + company_fuzzy
5	Tom Harris	SF + LinkedIn	0.88	name_fuzzy + company_fuzzy

Review Queue

Candidate Pair	Score	Reason for Review	Signal
Bob Smith (SF) / Robert Smith (HubSpot)	0.60	company_fuzzy only	Same company, name below threshold
Robert Smith (HubSpot) / Robert James Smith (LinkedIn)	0.60	name_fuzzy only	Name match, company below threshold
Sarah Connor (SF) / S. Connor (HubSpot)	0.80	phone_exact only	Phone match, name abbreviated
Amy Zhang (SF) / Amy Z. (HubSpot)	0.56	phone_exact only	Phone match, name truncated

Consolidated Lead View (Jennifer Lee Example)

After survivorship with most_complete strategy and field overrides:

{
  "canonical_id": "canon_lead_002",
  "entity_type": "lead",
  "canonical_data": {
    "name": "Jennifer Lee",
    "email": "[email protected]",
    "company": "Globex Corporation",
    "title": "Product Manager",
    "phone": "+1-555-0302",
    "location": "New York, NY",
    "lead_score": 72,
    "connections": 832,
    "lead_source": "outbound",
    "form_submitted": "whitepaper"
  },
  "field_provenance": {
    "name": { "source": "linkedin", "strategy": "source_priority" },
    "email": { "source": "salesforce", "strategy": "source_priority" },
    "company": { "source": "linkedin", "strategy": "source_priority" },
    "title": { "source": "hubspot", "strategy": "most_complete" },
    "phone": { "source": "hubspot", "strategy": "most_complete" },
    "location": { "source": "linkedin", "strategy": "most_complete" },
    "lead_score": { "source": "hubspot", "strategy": "most_complete" },
    "connections": { "source": "linkedin", "strategy": "most_complete" },
    "lead_source": { "source": "salesforce", "strategy": "most_complete" },
    "form_submitted": { "source": "hubspot", "strategy": "most_complete" }
  },
  "cluster": {
    "size": 3,
    "sources": ["salesforce", "hubspot", "linkedin"],
    "confidence": 0.90
  }
}

The golden record combines the best data from each source: LinkedIn's authoritative name and company, Salesforce's email for outreach, HubSpot's engagement data, and LinkedIn's enrichment metadata.

Key Takeaway

Fuzzy matching plus a review queue catches matches that exact matching misses, while keeping humans in the loop for uncertain cases. In this dataset, zero email addresses match exactly across sources. Without fuzzy name and company matching, every single lead would remain a duplicate. The review queue with a low threshold (0.55) ensures borderline cases like "Bob" vs "Robert" are flagged for human decision rather than silently missed or incorrectly auto-merged.

The most_complete survivorship strategy is ideal for lead matching because you want the richest possible profile, with every field filled in from whichever source has it.

Variations to Consider

If you want to catch "Bob" vs "Robert" automatically, use Soundex or Metaphone:

- name: name_phonetic
  type: similarity
  field: name
  algorithm: soundex
  threshold: 1.0
  weight: 0.4

Soundex maps "Robert" and "Bob" differently (R163 vs B100), so this would not help for that specific case. For nickname resolution, consider a preprocessing step that normalizes common nicknames before matching.

If you need stricter matching (e.g., B2B where false positives are costly):

decision:
  thresholds:
    match: 0.95    # very high bar for auto-match
    review: 0.70   # send more to review

If you want to score by lead engagement, use an aggregate survivorship override:

survivorship:
  default: most_complete
  overrides:
    - field: lead_score
      strategy: aggregate
      function: max

Related

• Spec Reference: Rules: Algorithm comparison for name matching
• Spec Reference: Survivorship: most_complete strategy details
• Tuning Match Quality: Diagnose and improve match quality
• Tutorial: Lead Matching: Step-by-step walkthrough with Python SDK