azcomputerguru/claudetools
1
0
Fork 0
Code Issues 58 Pull Requests Actions Packages Projects Releases Wiki Activity

sync: auto-sync from GURU-5070 at 2026-07-23 07:40:10

Browse Source 
Author: Mike Swanson
Machine: GURU-5070
Timestamp: 2026-07-23 07:40:10
This commit is contained in:
Mike Swanson
2026-07-23 07:40:53 -07:00
parent ecef1085e3
commit aa43bb7d1a
1 changed files with 219 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

219
session-logs/2026-07/2026-07-23-mike-prospect-list-vetting.md Normal file
View File

@@ -0,0 +1,219 @@
# Prospect Call-List Vetting — "Arizona Computer Guru New List.csv"
## User
- **User:** Mike Swanson (mike)
- **Machine:** GURU-5070
- **Role:** admin
---
## Session Summary
Mike supplied a purchased 238-row Tucson-area lead list intended as a call list for an outside
sales / appointment-setting firm working on ACG's behalf. The task was to vet it: dedupe against
existing and former clients, and strip prospect categories ACG does not want to onboard. Mike
noted this had been done before, but no record of prior criteria existed anywhere in the repo
(grep over `wiki/`, `session-logs/`, `.claude/memory/` for "call list", "lead list", "ICP",
"do not onboard" returned nothing), so the criteria were re-established from scratch this session.
The file was located at `C:\Users\guru\Documents\Arizona Computer Guru New List.csv`, not
Downloads as stated. First finding: the CSV is a **vendor preview export with no usable contact
data**. Columns `work_email`, `personal_email`, `work_phone`, `work_mobile`, `mobile`,
`alt_mobile`, `branch_phone`, `hq_phone`, `street1`, `street2`, `company_revenue`,
`company_employee_count`, `linkedin_url`, `primary_industry`, and all department-size and `hq_*`
fields contain the literal strings `Yes`/`No` (field-availability flags), not values. Verified
across all 238 rows. Only `company`, `domain`, `first_name`, `last_name`, `job_title`,
`job_level`, `job_department`, `city`, `state`, `zip`, `location_type`, and `matchId` hold real
data. A setting firm cannot dial this file as-is.
Dedupe was done against Syncro rather than `wiki/clients/` — the wiki has ~75 client articles
while Syncro holds 3000 customers and 768 contacts. Both were paginated out to local TSVs and
matched three ways: email domain (customer `.email` plus all contact emails, free-mail domains
excluded), normalized business name, and difflib fuzzy. Domain matching produced every real hit;
fuzzy below 0.86 produced only false positives (TMM Family Services -> AMP Facility Services,
NTH Consulting -> SRK Consulting, Stark Electric -> Fiesta Electric). Sixteen leads matched; 13
were genuine. Ticket and invoice counts plus max dates were then pulled per matched customer to
separate active from dormant: 4 active (Dataforth, SunDanzer, Patient Care Advocates, Ralph's
Transfer — all invoiced in 2026), 9 dormant former clients dating 2015–2022.
Mike confirmed all four proposed exclusion categories (IT/MSP competitors, orgs with in-house IT
staff, micro/consumer-facing shops, healthcare/HIPAA-heavy), plus keep-and-flag handling for
former clients and for remote decision-makers. All 238 rows were then classified by hand into
call tiers A/B/C or an exclusion bucket, with 14 ambiguous companies researched via WebFetch
before classification. Output: two CSVs in `~/Documents` — 168 callable companies tier-sorted,
70 excluded with a reason per row. Criteria were saved as a repo memory so the next list runs the
same way without re-asking.
## Key Decisions
- **Deduped against Syncro, not the wiki.** `wiki/clients/` covers ~75 clients; Syncro has ~3000
customers. Using the wiki would have missed every dormant/one-off record.
- **Included Syncro *contacts* (768) in the domain map, not just customer records.** This caught
three leads that exist only as contacts under a differently-named customer: SunDanzer (cust
139935), Aviva Children's Services (cust 2122768), Dataforth's older record (cust 578095).
- **Treated a Syncro record alone as meaningless.** Several matched customers have zero tickets
and zero invoices (Ironwood Tree Experience, Marana Chamber). Activity counts + max dates were
required to call anything "active."
- **Did not cut on the CSV's city/state column.** `location_type` = `Remote` (77 rows) means the
*contact* works remotely; the company is usually still Tucson. Verified example: The Carlson
Company reads "Lakewood CA" but is a 60-year Tucson plumbing/HVAC shop. Cutting on the state
column would have dropped ~34 valid AZ prospects. Instead: verify HQ via the domain, keep, and
flag `remote_contact`.
- **Kept medical device manufacturers and distributors** despite the healthcare exclusion —
Wells Johnson, NeuroVRD, Medfix International, Avery Therapeutics are not covered entities.
Flagged to Mike as a judgment call open to reversal.
- **Excluded Bluespan as a competitor** — Tucson ISP with 15,000+ subscribers and its own network
staff, not obvious from the name.
- **Excluded 6 orgs on the contact's own job title** (Network Administrator, Sr Systems Engineer,
IT Coordinator, IT Administrator, Sr Systems Administrator) — self-identifying in-house IT.
- **Three tiers instead of a flat keep/drop**, so the setting firm has a dial order rather than
168 undifferentiated rows.
- **Left 5 unverifiable companies in at Tier B/C rather than dropping them**, marked `UNVERIFIED`
for the setter to qualify — dropping on a failed HTTP fetch would discard real prospects.
## Problems Encountered
- **CSV not in Downloads.** Mike said Downloads; the file was in `~/Documents`. Resolved with a
`find` over `/c/Users/guru` maxdepth 4.
- **Syncro `/contacts` pagination caps below the requested `per_page`.** `per_page=100` over 8
pages returned only 400 unique rows against `total_entries: 768` (effective cap ~50/page).
Resolved by looping 20 pages and `sort -u`; confirmed 768 unique.
- **Six WebFetch verifications failed** — TLS cert mismatches (ziprtechnologies.com and
openenclose.net both point at shared bluehost/wpengine certs), HTTP 403 anti-scraping
(gvfan.com, swstucson.com), TLS internal error (palletrecyclers.com), connection refused
(repicon.com). Not logged to errorlog — these are remote-site conditions, not harness failures.
Resolved by marking the 5 affected leads `UNVERIFIED` at Tier B/C rather than guessing.
ZIPR Technologies was still excluded as a competitor because Mike explicitly confirmed it in
that category when answering the criteria question.
- **`os.environ['SP']` KeyError** in the first extraction script — assumed a scratchpad env var
existed. Resolved by using the literal absolute scratchpad path.
## Configuration Changes
**Created (repo):**
- `.claude/memory/feedback_prospect_list_vetting.md` — durable vetting criteria (type: feedback)
- Appended index line to `.claude/memory/MEMORY.md`
**Created (Mike's Documents, not in repo):**
- `C:\Users\guru\Documents\ACG Call List - VETTED.csv` — 168 rows, tier-sorted
- `C:\Users\guru\Documents\ACG Call List - EXCLUDED.csv` — 70 rows, reason per row
**Scratchpad (session-local, not committed):**
- `scratchpad/leads.tsv` — extracted lead fields
- `scratchpad/syncro_customers.txt` — 3000 customers (id, business_name, contact name, email)
- `scratchpad/syncro_contacts.txt` — 768 contacts (customer_id, name, email)
- `scratchpad/matches.txt` — dedupe match report
- `scratchpad/classify.py` — full 238-row classification + CSV generator (rerunnable)
## Credentials & Secrets
None discovered, created, or rotated. Syncro API key was read via the standard path
(`source .claude/scripts/syncro-env.sh`, vault `msp-tools/syncro`, user `mike`, key length 51).
All Syncro calls were read-only (GET) — no writes, no bot alerts required.
## Infrastructure & Servers
- Syncro PSA: `https://computerguru.syncromsp.com/api/v1` — 3000 customers, 768 contacts
- No other infrastructure touched.
## Commands & Outputs
```bash
# Locate the file (was NOT in Downloads)
find /c/Users/guru -maxdepth 4 -iname "*New List*"
# -> /c/Users/guru/Documents/Arizona Computer Guru New List.csv
# Prove the contact columns are availability flags, not data
python -c "import csv,io; r=list(csv.DictReader(io.open('Arizona Computer Guru New List.csv',encoding='utf-8-sig'))); print(sorted(set(x['work_phone'] for x in r)))"
# -> ['No', 'Yes'] (same for work_email, mobile, hq_phone, street1, company_revenue,
# company_employee_count, linkedin_url, primary_industry)
# Syncro customer pull (30-page guard, 100/page)
source .claude/scripts/syncro-env.sh
curl -s "${SYNCRO_BASE}/customers?per_page=100&page=${P}&api_key=${SYNCRO_API_KEY}" | tr -d '\000-\037' \
| jq -r '.customers[] | [.id, (.business_name // .fullname // ""), ...] | @tsv'
# -> 3000 rows
# Contacts pull — per_page is capped ~50 despite per_page=100; loop 20 pages
for P in $(seq 1 20); do curl -s ".../contacts?per_page=100&page=${P}&api_key=..." ; done
# -> 768 unique (matches meta.total_entries)
# Activity check per matched customer
curl -s "${B}/tickets?customer_id=${CID}&per_page=100&api_key=${K}" # .meta.total_entries + max(.tickets[].created_at)
curl -s "${B}/invoices?customer_id=${CID}&per_page=100&api_key=${K}" # .meta.total_entries + max(.invoices[].date)
curl -s "${B}/customer_assets?customer_id=${CID}&per_page=100&..." # .assets | length
```
**Activity results for the 13 genuine matches:**
| Lead | Company | Syncro cust | Tickets (last) | Invoices (last) | Assets | Verdict |
|---|---|---|---|---|---|---|
| 209 | Dataforth Corporation | 578095 | 207 (2026-07-17) | 232 (2026-07-17) | 50 | ACTIVE |
| 9 | SunDanzer Refrigeration | 139935 | 153 (2026-03-31) | 286 (2026-07-19) | 0 | ACTIVE |
| 107 | Patient Care Advocates | 29907453 | 15 (2025-09-19) | 33 (2026-07-01) | 0 | ACTIVE |
| 135 | Ralph's Transfer | 28715601 | 10 (2025-10-16) | 9 (2026-01-25) | 0 | ACTIVE |
| 62 | Termagon | 26038848 | 1 (2022-02-14) | 1 (2022-02-18) | 1 | FORMER |
| 120 | Earth Savers Energy | 16378562 | 4 (2022-12-27) | 3 (2019-06-04) | 0 | FORMER |
| 23 | Serrato Corporation | 24410171 | 1 (2021-07-19) | 0 | 0 | FORMER |
| 78 | Saguaro Solar | 21754858 | 1 (2020-10-19) | 1 (2020-10-20) | 0 | FORMER |
| 230 | Swaim Associates | 18118621 | 1 (2019-10-16) | 0 | 0 | FORMER |
| 37 | Mission Mgmt & Trust | 9778876 | 1 (2017-07-31) | 0 | 0 | FORMER |
| 35 | Aviva Children's Svcs | 2122768 | 0 | 1 (2015-04-10) | 0 | FORMER |
| 38 | Ironwood Tree Experience | 19057730 | 0 | 0 | 0 | FORMER (record only) |
| 61 | Marana Chamber | 17767946 | 0 | 0 | 0 | FORMER (record only) |
**False-positive fuzzy matches (rejected):** TMM Family Services ~0.80 AMP Facility Services;
NTH Consulting ~0.79 SRK Consulting; Stark Electric ~0.83 Fiesta Electric.
**Final classification:**
```
CALL LIST: 168 EXCLUDED: 70
Tier A: 87 HEALTHCARE 18
Tier B: 57 IN-HOUSE-IT 18
Tier C: 24 MICRO 13
former client: 9 NOT-AZ 10
remote contact: 48 COMPETITOR 7
unverified: 5 ACTIVE-CLIENT 4
```
**WebFetch verifications that changed a classification:**
| Company | Finding | Effect |
|---|---|---|
| Artec Industries | Off-road parts mfr, HQ North Salt Lake UT | -> EXCLUDE NOT-AZ |
| Urban Moment | Multifamily developer, Austin TX, no AZ ops | -> EXCLUDE NOT-AZ |
| Bluespan | Tucson ISP, fiber/wireless, 15k+ subs, own staff | -> EXCLUDE COMPETITOR |
| TracTru | SaaS/marketing platform, own devs + QA lead | -> EXCLUDE IN-HOUSE-IT |
| NTH Consulting | Tucson mgmt/enterprise-risk consulting, NOT IT | -> KEEP Tier A |
| The Carlson Company | Tucson plumbing/HVAC/sheet metal, 60+ yrs | -> KEEP Tier A |
| SWTDT | Southwest Truck Driver Training, Tucson + Phoenix | -> KEEP Tier A |
| Termagon | Tucson family pest control since 1995, multi-county | -> KEEP Tier A (former) |
| NeuroVRD | Tucson surgical-AI medical DEVICE maker, not provider | -> KEEP Tier B |
| Quick & Associates | Tucson CPA, 1331 N Wilmot Rd | -> KEEP Tier A |
| Oremus Material | Concrete, Chicago HQ w/ Tucson branch | -> Tier C GEO |
## Pending / Incomplete Tasks
- **BLOCKER for the setting firm: the source CSV has no phone numbers or emails.** Mike needs the
vendor to release the unlocked export. The vetted CSV carries `matchId` and `domain` per row so
the unlocked data can be joined back without redoing the vetting.
- **Open judgment call:** medical device makers/distributors (Wells Johnson #197, NeuroVRD #65,
Medfix International #29, Avery Therapeutics #15) were kept as non-covered-entities. Mike may
want them moved to EXCLUDED if he wants all healthcare-adjacent gone.
- **5 UNVERIFIED leads** need qualification before dialing: Third Act LLC / gvfan.com (#68),
Pallet Recyclers (#115), REPICON (#151), Open Enclose (#175), Southwest Solutions (#159).
- No wiki article implied (root-level session log) — no `/wiki-compile` needed.
## Reference Information
- Source file: `C:\Users\guru\Documents\Arizona Computer Guru New List.csv` (238 rows, 56 cols)
- Deliverables: `C:\Users\guru\Documents\ACG Call List - VETTED.csv`,
`C:\Users\guru\Documents\ACG Call List - EXCLUDED.csv`
- Rerunnable classifier: `scratchpad/classify.py` — edit the `C = {...}` dict (idx -> tier/reason/note)
and rerun to regenerate both CSVs
- Memory: `.claude/memory/feedback_prospect_list_vetting.md`
- Syncro API: `https://computerguru.syncromsp.com/api/v1` (read-only this session)
- Tier legend: `A` = priority call (right size/industry, owner-exec contact, local);
`B` = fits, lower priority; `C` = qualify first (geography, size, or unverified)
- Exclusion reasons: `ACTIVE-CLIENT`, `COMPETITOR`, `IN-HOUSE-IT`, `MICRO`, `HEALTHCARE`, `NOT-AZ`