Agency Implementation Agent

Research Summary for TSM Review
Tuesday, March 3, 2026
1:30 PM • 488 tickets • 11 + 735 transcripts analyzed
488 TCS Tickets
(8 months)
194 With Tech
Questionnaire Data
129 Transcripts Analyzed
(11 deep + 118 Haley/Chris)
73.5h Haley + Chris
Call Time Analyzed
142% Volume Growth
Jul '25 → Feb '26
~35% Tickets That Look
Automatable

What the Data Shows

Top Platforms (from 488 TCS tickets)

Shopify
103
Zapier
48
HubSpot
44
Stripe
37
GTM
36
WooCommerce
23
Impact
23

Top Issues (from 488 TCS tickets)

Redirect Issues
128
Events Not Firing
61
Cross-Domain
44
Duplicate Conversions
41
SSL / iFrame
10

Tracking Types (from tickets)

Server-Side (S2S)
210
Client-Side (JS)
200
Skimlinks
31

TSM Workload

TICKETS (8 months)
Haley Betts
235 (48%)
Chris Chimento
98
Errol Jacobs
82
FATHOM CALLS (Dec–Mar)
Chris Chimento
60 calls · 43.3h
Haley Betts
58 calls · 30.2h
85 Fathom recordings in tickets 26 Zoom links 14 Loom videos 194 questionnaires

Key Learnings from 129 Analyzed Calls

11 manually deep-analyzed scoping calls + 58 Haley Betts calls + 60 Chris Chimento calls (via Fathom API)

What Clients Struggle With

  • Offer vs. Partner vs. Brand confusion — most common across all calls
  • Transaction IDs — clients don't know what they are or how to pass them
  • First-touch vs. last-touch — one call spent 45 min just on this
  • Duplicate conversions — "Allow duplicates" left on by accident (149x dupes in one case)
  • Revenue vs. Payout — when to use CPS %, flat rate, $0, etc.
  • Engineers say it's done, it's not — TSM wastes time debugging ghost setups
  • Platform migrations — Impact/Cake/TUNE clients confused by Everflow terminology differences NEW from Fathom

Where TSM Time Goes

  • #1: Walking clients through UI — every single call, "click here... no, there..."
  • #2: Crafting postback URLs in Zoom chat — deterministic format, same pattern every time
  • #3: Live debugging — 20+ min sessions for known misconfigs
  • #4: Waiting for the right person — tech contact missing from call, no permissions
Haley: 58 calls, 30.2h, avg 31 min — Implementation-focused (41%). 48% of all TCS tickets.
Chris: 60 calls, 43.3h, avg 44 min — Scoping + complex architecture. 13 min longer per call.

Haley & Chris — Fathom Call Profile (Dec 2025 – Mar 2026)

HB
Haley Betts 58 calls · 30.2h · 31 min avg
Implementation
24
Onboarding
13
Scoping
10
61 unique clients · Specializes: Shopify, Ringba, standard S2S
CC
Chris Chimento 60 calls · 43.3h · 44 min avg
Scoping
12
Implementation
10
Onboarding
9
62 unique clients · Specializes: Leadspedia, attribution, complex architecture

Top Gotchas the Agent Could Pre-Check

MisconfigurationWhat Happens
"Allow Duplicate Conversions" left on149x duplicate posts in one case
Partners set up as Brands (or vice versa)Broken reporting
Offer visibility Public instead of PrivatePartners join wrong offers
iframed widgetsTracking breaks completely
Two click scripts on same pageDouble-firing
Stripe can't accept 3rd-party dataMust use email attribution
HubSpot can't pass revenueRevenue missing from reports

Setups That Follow a Template (Automation Candidates)

Shopify standard setup (103 tickets) S2S postback URL generation Skimlinks (26 near-identical) Ringba mapping (formulaic) GoHighLevel webhooks (copy-paste + ID swap) GTM container config (36 tickets) Post-call summary emails Pre-call diagnostic check

How We Got This Data

1. Jira API Fetch
Python script connected to TCS board via API. Pulled all 488 tickets (8 months) with descriptions, comments, statuses, assignees.
2. Pattern Scan
Scanned every ticket's text for platform keywords, tracking types, and common issue patterns. Also found 85 Fathom recording links embedded in tickets.
3. Transcript Analysis
Pulled 11 Fathom call transcripts (6,381 lines). Analyzed for client struggles, TSM process, platform-specific patterns, and automation opportunities.

How This Feeds the Agent — Click Each Step

Step 1
Intake
Questionnaire
→
Step 2
Classify
Simple / Complex
→
Step 3a
Auto-Generate
Script + Config
→
Step 3b
Escalate
TSM + Pre-fill
→
Step 4
Validate
Gotcha Pre-Check
Front-End: Agency-Facing Form

Tech Setup Questionnaire

Found: 2 offers, 14 partners
Back-End: What Happens on Submit
On submit: 1. Save answers to database 2. Auto-lookup NID 3905 in Everflow API → Pull existing offers, partners, events → Check current config state 3. Pass to Step 2 (Classify) with: { "client": "Acme Corp", "nid": 3905, "tracking": "s2s", "platform": "shopify", "domain_changes": 1, "iframes": false, "events": ["click", "purchase"], "revenue": "dynamic", "attribution": "last_touch", "existing_offers": [/* from API */] } Data sources: questionnaire + EF API lookup
Front-End: Complexity Score Card

Acme Corp (NID 3905)

SIMPLE
2 / 10
Known platform (Shopify)+0
S2S tracking+0
1 domain change+1
No iframes+0
Standard events (click, purchase)+0
Dynamic revenue+1
No cross-instance / Leadspedia+0
0-3 Simple
Agent handles
4-6 Medium
Agent drafts + TSM review
7-10 Complex
Route to TSM
Back-End: Scoring Logic
Complexity factors (from our data): +0 Known platform (Shopify, WooCommerce...) +2 Custom-built eCommerce +1 Each domain change beyond 0 +3 iframes present +2 First-touch attribution needed +2 Leadspedia / custom postback chain +1 Dynamic revenue +2 Migration from another network +3 Cross-instance tracking +1 Subscription / rebill model Routing: if score <= 3: → Step 3a (Auto-Generate) if score <= 6: → Step 3a (draft) + TSM review if score > 6: → Step 3b (Escalate to TSM) Weights tuned from 488 tickets + call data. Ceno can adjust as TSM gives feedback.
Front-End: Generated Implementation Package

Acme Corp — Implementation Package

AUTO-GENERATED
Shopify + Server-Side (S2S) + Dynamic Revenue
Shopify Setup
Postback URL
JS Click Script
Event Config
1. Go to Integrations → E-Commerce → Shopify
2. Enter store URL: acme-corp.myshopify.com
3. Map events: Click → Page View, Purchase → Conversion
https://www.everflow.io/s2s?adv_event_id=1&aid={oid}&txn_id={transaction_id}&amount={amount}
<script src="https://www.everflow.io/scripts/sdk/everflow.js"></script> <script>EF.click({ offer_id: 42 });</script>
Duplicates
OFF
Visibility
Require Approval
Attribution
Last-touch
Back-End: Template Engine
Templates built from ticket patterns: SHOPIFY_S2S_TEMPLATE: 1. Generate postback URL from: - offer_id (from NID lookup) - event_ids (from questionnaire) - revenue_param (dynamic/fixed) 2. Generate JS click script 3. Generate Shopify integration steps 4. Set default config: - duplicates: OFF - visibility: REQUIRE_APPROVAL - attribution: questionnaire answer SIMILAR TEMPLATES FOR: - Shopify + client-side only - Skimlinks (26 identical setups) - Ringba call tracking - GoHighLevel webhook - Standard S2S (non-Shopify) - GTM container Each template = one common ticket type. ~103 Shopify + 48 Zapier + 31 Skimlinks = 180+ tickets this covers.
Front-End: TSM Gets Pre-Filled Ticket

Ignite Media (NID 2847)

COMPLEX — 8/10
Leadspedia Multi-offer 2 domain changes
Assigned To
CC
Chris Chimento
Leadspedia / complex integrations
Pre-Filled Context
Platform: Leadspedia
Tracking: Hybrid
Events: Lead, Buyer, Revenue
Revenue: Dynamic per-buyer
Agent's Pre-Check
!
2 domain changes — cross-domain tracking risk
!
Dynamic per-buyer revenue — complex payout split
✓
No iframes detected
✓
Backend supports GET requests
Similar Past Tickets
TCS-785 Leadspedia, resolved by Chris • TCS-634 Typeform iframe, Haley
Back-End: Escalation Logic
Escalation creates: 1. Jira ticket (TCS project) with: - All questionnaire answers - Complexity score breakdown - Pre-check results - Similar past tickets (via text search) - Suggested assignee (by expertise) 2. Assignee logic: Leadspedia/complex → Chris Chimento Shopify/standard → Haley Betts Ringba/call → Haley Betts Attribution issues → Chris Chimento 3. Slack notification to assigned TSM 4. Client gets email: "Your setup has been routed to our Technical Solutions team. Chris will reach out within [SLA]." TSM starts the call with EVERYTHING pre-filled instead of asking from scratch. Saves the first 10-15 min of every call.
Front-End: Validation Report

Pre-Launch Checklist

Acme Corp (NID 3905)
✓
Offer created (ID: 42)
✓
Events configured Click, Purchase
✓
Postback URL responds 200
✓
JS click script found on landing page
✓
Duplicate conversions OFF
✓
Offer visibility: Require Approval
✓
Attribution: Last-touch
!
1 domain change — confirm JS SDK cookie persists across landing.acme.com → checkout.acme.com
Test Results
✓
Test Click
Tracked
✓
Test Conversion
Tracked
✓
Revenue
$49.99 passed
READY TO LAUNCH
Back-End: Gotcha Checks (from our data)
Automated checks from 488 tickets: CHECK_DUPLICATES: → Is "Allow Duplicates" OFF? → (caused 149x dupes in one case) CHECK_VISIBILITY: → Is offer Private or Require Approval? → (partners joining wrong offers) CHECK_CLICK_SCRIPTS: → Only 1 click script on page? → (double-firing from 2 scripts) CHECK_CROSS_DOMAIN: → If domain changes > 0, verify cookie persistence or email fallback CHECK_IFRAME: → If iframe detected, flag for TSM CHECK_POSTBACK: → Hit postback URL, confirm 200 → Verify transaction_id param exists CHECK_STRIPE: → If Stripe, confirm email attribution CHECK_HUBSPOT: → If HubSpot, warn: no revenue passing Every check maps to a real gotcha from our ticket + call analysis.

What we can give Ceno

  • 488 raw tickets with descriptions, comments, statuses
  • 11 deep-analyzed call transcripts showing real scoping conversations
  • Platform patterns — Shopify, Stripe, Ringba, GoHighLevel, etc.
  • Gotcha database — 10+ known misconfigs to check for
  • Questionnaire fields — 15 questions validated from transcripts
  • Template examples — postback URLs, event steps, integration configs
  • Full Fathom API analysis — 736 calls, 735 transcripts, Haley & Chris deep dives (see below)

Estimated Impact

  • ~35% of tickets are templated enough for automation
  • ~20-30 tickets/month off TSM's plate
  • 10-15 min saved per call with pre-filled questionnaire
  • Postback URLs — instant vs. hand-typed in Zoom
  • Pre-checks — catch misconfigs before the call starts
Offer to Ceno: All raw data (tickets JSON + transcripts + analysis) is packaged and ready to send whenever he needs it.

For Discussion

Where's the line between "agent can handle" vs. "needs TSM"?
Which 5 ticket types would you most want automated?
What's missing from the questionnaire draft?
Fathom API connected — 736 calls, 735 transcripts pulled. Can set up webhook for auto-ingest.
Appendix
Full Fathom API Call Analysis
736 calls · 735 transcripts · 293K lines · 29 recorders · Dec 2025 – Mar 2026
Data Source: Fathom API
Fetched Mar 3, 2026 via api.fathom.ai/external/v1

All-Calls Overview

736Total Calls
735With Transcripts
365hTotal Call Time
293KTranscript Lines
29.7Avg Call (min)
29Team Members

Call Types (all 736)

Other / Misc
352
Onboarding
165
Strategy/Review
74
Implementation
67
Internal/Team
47
Scoping
22
Troubleshooting
7

Top Recorders (calls & hours)

Laurie Cutts
74 · 29.9h
Denis Urnikis
65 · 29.3h
Chris Chimento
60 · 43.3h
Lexi Norman
59 · 30.4h
Haley Betts
58 · 30.2h
Irina Sergushina
52 · 20.5h
Isaac Maier
46 · 23.3h
Peter Bloy
41 · 19.2h
Francisco Bervis
39 · 17.9h
Jake Adee
34 · 17.2h

Weekly Call Volume (Last 6 Weeks)

79
Jan 26
80
Feb 2
61
Feb 9
63
Feb 16
73
Feb 23
29
Mar 2*
*Mar 2 = partial week (Mon–Tue only) · ~70–80 calls/week is steady state

What 293K Transcript Lines Reveal

Platforms Mentioned in Calls

Impact
170 calls
Shopify
107
Stripe
80
Zapier
68
HubSpot
59
GTM
48
Cake
43
TUNE
38
WooCommerce
37
Ringba
25
WordPress
23
Salesforce
19
GoHighLevel
15
HasOffers
14
New finding vs. tickets: Impact.com is the #1 platform mentioned in calls (170) but only #7 in tickets (23). This suggests Impact migrations are discussed heavily on calls but don't always generate separate TCS tickets.

Top Issue Keywords (frequency across all transcripts)

Offer
7,057
Partner
6,039
Tracking
4,649
Conversion
3,721
Event
3,182
Payout
2,390
Transaction ID
2,246
Postback
2,133
Revenue
1,863
Redirect
930
Attribution
694
Click Script
587
Duplicate
466
Offer/Partner/Tracking dominate every call. The agent needs to master these concepts first. Transaction ID + Postback (4,379 combined mentions) confirms these are the #1 setup topic.

Haley Betts & Chris Chimento — Deep Dive

Full Fathom call history analysis · Dec 2025 – Mar 2026
HB
Haley Betts
Primary TSM · Implementation Lead
58Calls
30.2hTotal Time
31 minAvg Duration

Call Type Breakdown

Implementation
24
Onboarding
13
Scoping
10
Other
7
Internal
2
61 unique client domains · Top: callsbybrand.com (4), paulstreet.co (3), capradigitals.com (3), rein.ai (2)
Key pattern: Haley is implementation-focused (41% + 22% onboarding = 63% hands-on setup). She's the go-to for Shopify, Ringba, and standard S2S. Her calls are shorter (31 min avg) because she follows repeatable playbooks.
CC
Christopher Chimento
Senior TSM · Complex Integrations
60Calls
43.3hTotal Time
44 minAvg Duration

Call Type Breakdown

Scoping
12
Implementation
10
Onboarding
9
Technical
2
Troubleshooting
2
62 unique client domains · Top: prefdata.net (4), simplertrading.com (3), quotesdirect.com (3), nxtwaveresearch.com (3)
Key pattern: Chris averages 44 min (13 min longer). Scoping is his #1 call type. He handles Leadspedia, attribution redesigns, multi-platform integrations. His calls are complex but patterns exist across repeat clients.

Haley vs. Chris — At a Glance

MetricHaley BettsChris Chimento
Total Calls (Dec–Mar)5860
Total Hours30.2h43.3h
Avg Call Duration31 min44 min
Top Call TypeImplementation (24)Scoping (12)
Unique Clients6162
SpecializationShopify, Ringba, S2S, standard setupsLeadspedia, attribution, complex architecture
Ticket Volume (8mo)235 tickets (48% of all TCS)98 tickets
Automation OpportunityHIGH — repeatable playbooksMEDIUM — complex but patterns exist
Combined: Haley + Chris = 118 calls, 73.5 hours, 123 unique clients in just 3 months. An agent that handles even 30% of Haley's repeatable setups frees up ~9 hours/month of TSM time.

Full Team Call Stats (29 Recorders)

NameCallsHoursAvg (min)Top Call Types
Laurie Cutts7429.924.3Strategy/Review, Onboarding
Denis Urnikis6529.327.1Onboarding, Implementation
Christopher Chimento6043.344.1Scoping, Implementation
Lexi Norman5930.430.9Onboarding, Implementation
Haley Betts5830.231.3Implementation, Onboarding
Irina Sergushina5220.523.6Strategy/Review, Onboarding
Isaac Maier4623.330.3Strategy/Review, Implementation
Peter Bloy4119.228.1Strategy/Review
Francisco Bervis3917.927.5Onboarding
Jake Adee3417.230.4Strategy/Review
Sharbel Sader3113.826.7Onboarding
Peter Kurjanowicz3014.629.2Onboarding
Cindy Shen3013.827.6Strategy/Review
Franciny Estrella2312.131.6Onboarding
Whitney Bastian188.729.0Onboarding
Derek Mohl167.327.4Strategy/Review
Eddy Erdogan157.831.2Onboarding
Elaura Johnson105.231.2Strategy/Review
+ 11 more recorders with <10 calls each (Genny, Kristin, Mari, Sabrina, Dasha, Marwan, Jason, Joaquin, Sooah, Elizabeth)

Key Insights from Full Call Analysis

What Ticket Data Missed

  • Impact.com is everywhere — 170 calls mention it vs. only 23 tickets. Migration discussions happen on calls but don't generate TCS tickets
  • Cake/TUNE migrations — 81 combined call mentions. Clients migrating from legacy networks is a huge pattern not fully captured in tickets
  • HasOffers still active — 14 call mentions. Thought this was dead but clients still reference it
  • Salesforce integration — 19 call mentions but barely shows in TCS tickets. These might go through different channels

Volume Patterns

  • Steady state: ~70-80 calls/week across the whole team
  • Chris has the longest calls (44 min avg) — complex scoping takes time
  • Laurie has the most calls (74) but shortest avg (24 min) — high-volume quick touchpoints
  • Onboarding dominates — 165 of 736 (22%) are onboarding calls
  • Only 7 calls flagged as troubleshooting — most debugging happens inside "implementation" or "other" calls
Bottom line: The 11 deep-analyzed scoping calls (top section) identified specific automation patterns. This bulk analysis of 735 transcripts confirms those patterns at scale and reveals new ones — especially around platform migrations (Impact, Cake, TUNE) that the agent should be trained on.
488 TCS tickets (Jul 2025 – Mar 2026) + 11 deep-analyzed scoping calls + 736 Fathom calls (735 transcripts, 293K lines) via API • March 3, 2026 • Everflow