Outreach engine · Apollo round 1 · built 2026-08-25
Twelve names, nine companies, twelve credits.
The 69 Form-5500 plan sponsors near the office, enriched through Apollo on 25 Aug. The loop ran end to end — list → domain → decision maker → verified address → sequence — for twelve credits of eighty-five and no dollars. This page is the plan that came out of it.
Report: docs/APOLLO-ROUND1-REPORT.md · campaign: APOLLO-CAMPAIGN-R1.md · warming: APOLLO-WARMING-PLAN.md · runbook: APOLLO-TEAM-RUNBOOK.md · contacts: apollo-round1-contacts.csv · rails: PROSPECTS-PIPELINE.md
Where this stands right now
List cleaned + scored Apollo account live — Professional trial 21 domains resolved 12 verified decision-makers Sequences + copy written Lead magnets built Sending domain chosen — mail.hipsoninvestments.com Sending mailbox not yet created Trial expires ~7 Sep — plan decision due FINRA review — SEQ-B held Ameritas gate — S3 held
Round 1 ran on 25 Aug. Twelve verified decision-maker emails across nine Houston employers, for twelve credits of eighty-five and no dollars. Nothing sends until michael@mail.hipsoninvestments.com exists and is connected to Apollo, and until a human has read every draft. Full write-up: docs/APOLLO-ROUND1-REPORT.md.
Round 1 results — 25 Aug
The loop ran end to end: list → domain → decision maker → verified address → sequence. Detail in docs/apollo-round1-contacts.csv.
| Account | Employees Apollo | Participants Form 5500 | Decision maker found | Track |
|---|---|---|---|---|
| BD Energy Systems | 47 | 48 | Sommer Scott — Director of HR + Andrea Rutherford, Cost Controller | SEQ-A |
| Upstream Engineering | 54 | 46 | Marisol Burk — VP Administration + Ronnie Smith, President | SEQ-A |
| Tribble & Stephens | 21 | 20 | Gary Kelley — President + Jim Roach, Sr VP Preconstruction | SEQ-A |
| American Tech Vision | 87 | 77 | Sumit Arigapudi — Founder & President | SEQ-A |
| H-Town Technologies | 78 | 43 | Pramila Tamminayana — Manager, HR TPA trap corrected — 5500 named Qian Liu | SEQ-A |
| AIG Technical Services | 43 | 46 | Priscilla Freeman — Office Operations Manager | SEQ-A |
| Nationwide Coins | 45 | 48 | Porsche Anderson — CEO verify identity by phone before sending | SEQ-A |
| Dynamic Engineers | 39 | 7 | Roland Teoh — Owner & CEO domain is dynamiceng.com, not dynamicengineers.com | SEQ-B |
| Compansol | 30 | 26 | Gunjan Seth — Vice President TPA trap corrected — 5500 named Marcus Bowie | SEQ-B |
Employee count re-graded the list — two of the biggest names came off it
BioUrja filed 131 participants; Apollo says 600 employees. Radix filed 48; Apollo says 1,900. Both are enterprise, neither is a small-group prospect, and both are removed from round 1. Cutting the other way: Dynamic Engineers filed 7 participants against 39 employees — 82% of the staff are outside the plan, and that gap is the strongest single story on the whole list.
Company and headcount data is free in Apollo. Finding this out cost nothing; finding it out after pitching them would have cost credibility.
Half this list should never go through Apollo at all
Thirteen of 34 shortlisted companies produced no reliable Apollo match — the dental practices, the single-location clinics, the four-person professional firms. Apollo's graph is built from B2B digital exhaust and these businesses leave almost none of it. They get called and visited, not emailed, which is how the 14 already-touched accounts were worked and is cheaper per contact than Apollo is.
Apollo's own AI qualifier is wrong for this business — override it
Apollo marked Priscilla Freeman, Office Operations Manager at a 43-person engineering firm — precisely our buyer — as Disqualified. It is trained on enterprise software sales, where an office manager is noise. At a 20–90 person Houston employer the office manager frequently is the benefits decision. Treat it as a hint, never a decision.
1 · Three findings that change the plan
These came out of reading the list rather than counting it. Each one would have cost money or reply rate if we had skipped it.
The new spreadsheet is the old spreadsheet
Near Hipson Office List (1).xlsx is content-identical to the 69 rows imported on 14 Aug and sitting in /admin → Prospects today. Same plans, same 14 field notes, no new drops, no new call outcomes — diffed row by row, zero new intel. Round 1 is therefore an enrichment, not an import. The value we add is the four columns the sheet does not have: domain, real decision maker, employee count, renewal month.
A quarter of the named contacts are not the customer
17 of 69 Admin Name values appear on multiple unrelated companies — outsourced plan administrators listed on the Form 5500, not the employer's benefits decision maker. Writing "Hi Marcus, I see you handle benefits at Moana Minerals" to five different companies is the fastest way to burn a new sending domain.
| Listed name | Plans | Companies |
|---|---|---|
| Marcus Bowie | 5 | Kovan Technology · Compansol · Moana Minerals · WTS Energy · Doehring & Doehring |
| Qian Liu | 4 | H-Town Technologies · Enclave Vision · VSMSQ Structural · UPPF |
| Kenesha Brown | 2 | Now Insurance · Greenhill — also share a phone number |
| Richard Tatum | 2 | Valka Rice · Josephine's Day Spa |
| Nick Rice | 2 | JBLREG Holdings · IHNI Group |
| Edward Rojas | 2 | Amaanah Refugee Services · Precision Golf Academy |
All 17 flagged SHARED-ADMIN in the CSV, scored −10, and marked ignore the listed admin — Apollo finds the real one.
This is mostly a 401(k) list, not a group-health list
Median participants: 12. Only 19 of 69 reach the 25–200 band the level-funded pitch is built for; 27 have fewer than ten. But the caveat cuts the other way — 5500 participant count is a floor on headcount, not headcount. A 40-person company with poor take-up files 12 participants. That is exactly why Apollo org enrichment (true employee count) is the first thing we buy, and why no segment below is final until it runs.
2 · The list, after cleaning
67 workable companies by segment
Provisional — industry is a keyword guess off the company name and is replaced by Apollo's SIC/NAICS on enrichment.
S1 · Small-group employers
Median 46 participants · $68.7M
The core group-health play. Energy, engineering, tech, construction. Angle: "you're not seeing all the options" — level-funded underwriting. The 50-life case study belongs here and nowhere else.
S3 · Micro-sponsors
Median 8 participants · $48.7M
Too small for level-funded today, but holding real plan assets with no active advisor. Angle: fiduciary file + fee benchmarking. Never a performance claim — FINRA 2210. Held until Ameritas clears.
S4 · Medical & dental
Median 7 participants
Dual play. Their own 4–21-life staff plan is secondary; the asset is their patient families, who need exactly what /aba, /health and /quote do. A practice that refers beats one that buys.
S5 · Referral partners
Attorneys, CPAs, appraisers
Plus Frank J Lucka & Associates, surfaced by the Cyphersix call. Never send these the S1 sequence. Their clients ask them benefits questions they cannot answer. Coffee, not automation.
The A-list — first wave
Ten A-grade accounts, scored on size, industry, plan assets, proximity and prior field contact. The A/B pool is 26 and roughly doubles the moment renewal months land.
| Company | Participants | Plan assets | Industry | Note |
|---|---|---|---|---|
| BioUrja Holdings | 131 | $12.5M | Energy | Largest headcount on the list |
| Camelot Integrated Solutions | 100 | $1.5M | Tech | Highest score |
| Enerfin Resources | 83 | $11.9M | Energy | Drop logged — no one in office |
| American Tech Vision Solutions | 77 | $0.5M | Tech | Drop logged — 401(k) + insurance with ADP |
| BD Energy Systems | 48 | $3.4M | Energy | — |
| Radix US | 48 | $2.7M | Tech | — |
| Upstream Engineering | 46 | $8.9M | Engineering | — |
| Sonangol Marine Services | 32 | $7.1M | Energy | Same building as Sonangol USA |
| JFT Construction | 27 | $2.5M | Construction | Address stale — office moved |
| Sonangol USA | 27 | $3.8M | Energy | 1177 Enclave Pkwy |
Excluded, and why
| Company | Reason |
|---|---|
| C-Mar Group Holdings | Listed plan admin is Michael Hipson. Existing relationship, not a prospect — must never enter a cold sequence. |
| Gush Education KG2 | Explicit "was not interested" on the call; plan and health both with their payroll company. Respect it — revisit on the renewal calendar in 12 months, not in a sequence. |
Two more data traps kept in the file rather than silently dropped: Valka Rice — the sponsor phone reaches a Chick-fil-A (scored −10). Cyphersix — the number reaches an accounting firm, which is not a dead end but a referral-partner lead, so it moves to S5.
3 · The three gaps — two closed, one open
These were the blockers on 25 Aug morning. Two closed the same day; the third is free, public, and still the most valuable field missing.
| Gap | Why it blocks sending | How it closes |
|---|---|---|
| Domain closed | Apollo keys on domain; we had company names derived from plan names. No domain, no enrichment, no sequence. | 21 of 34 resolved. Matched Apollo’s listed phone against the Form 5500 sponsor phone — an exact match happened 11 times and makes the identification certain rather than a guess. The remaining 13 aren’t in Apollo at all and go to the phone-and-door channel. |
| Real decision maker closed | 17 rows carried a TPA in the contact field; the rest carried an unverified name. | 12 verified emails across 9 accounts, capped at two per account. Both TPA rows that made the shortlist were corrected to the real person — Pramila Tamminayana at H-Town, Gunjan Seth at Compansol. |
| Renewal month still open | Worth +40 in the score, more than every other signal combined. It turns cold outreach into calendar-timed outreach. | DOL EFAST2 / Schedule A scoped pass. Public, free, costs no Apollo credits — and Apollo does not have it. The highest-value thing left to build. Until it exists, SEQ-A step 1 asks the prospect directly. |
4 · Apollo, and the rules that keep it cheap
Account is live: Hipson Benefits, MT is Admin, Professional trial expiring ~7 Sep, 85 credits. No API key created — not needed until the Cloud Function automation exists, and it doesn't yet.
What does and doesn't cost money
| Action | Cost |
|---|---|
| Searching companies · searching people | Free |
| Seeing a name, title, company, city, employer size | Free |
| Revealing an email address | 1 credit |
| Revealing a mobile number | More than 1 |
| Saving a contact / adding to a sequence | Reveals the email, so 1 credit |
So: search endlessly, reveal sparingly. An hour narrowing a search to exactly the right nine people costs nothing. The moment you click Access email you have spent money you can't get back.
Standing rules. Never more than 2 people per company. Never reveal someone you aren't contacting this month. Never reveal an address whose status is guessed or extrapolated — it will bounce, and a bounce costs the domain far more than a missed contact costs us. Round 1 spent 12 of 85.
Two trial constraints found on 25 Aug
Bulk select-all is capped. Contacts have to be ticked one at a time. Fine at 12; painful at 200 — factor it into the plan decision.
Apollo challenges rapid automated queries. Scripted enrichment tripped Cloudflare mid-run. Round-2 scale needs the official API on a paid tier, not browser automation.
Firestore stays the system of record; Apollo is a utility. Log append-only, no deletes. This is the BenefitZone lesson written down for the third time.
5 · Sending identity — decided, and the risk that came with it
MT chose mail.hipsoninvestments.com on 25 Aug over registering a separate go. subdomain. Full plan: docs/APOLLO-WARMING-PLAN.md.
Cold outreach now shares a reputation with client mail
It's a defensible call — the domain is already verified in Resend, so round 1 starts this week instead of in four. But it means: if a prospect marks a Hipson cold email as spam, the hit lands on the same domain that carries lead alerts, portal notices and the daily brief. There is no firewall between them.
The gap nobody should assume away: the DEFAULT_ALLOWLIST warming guard in functions/mailer.js protects the transactional path only. Apollo authenticates straight to a mailbox and never touches that code. No allowlist protects prospects. The volume ceiling below is the entire mitigation.
Before the first send: create michael@mail.hipsoninvestments.com and connect that mailbox to Apollo. Never connect the primary @hipson.com mailbox — that address is how existing clients reach Michael, and a complaint against it has no recovery path.
Standing recommendation: register go.hipsoninvestments.com before round 2 scales past ~50 contacts a month and move cold sending there, leaving mail. clean. One DNS zone, three records. Revisit at the 30-day review.
The volume ramp
Per mailbox, per business day. A new sending identity earns volume; it doesn't get given it.
| Week | Emails/day | What's running |
|---|---|---|
| Wk 1 · 25 Aug | 3 | SEQ-A step 1 only, hand-sent, 9 contacts spread over 3 days |
| Wk 2 · 1 Sep | 5 | SEQ-A steps 2–4 land. First replies expected. |
| Wk 3 · 8 Sep | 8 | SEQ-B opens if FINRA review cleared. Steps 5–6. |
| Wk 4 · 15 Sep | 12 | Full round-1 cadence. Step 7 — the 5500 Snapshot — sends. |
| Wk 5+ | 20 max | Only if bounce < 2% and complaints = 0. Otherwise hold. |
Stop conditions — any one halts sending the same day. Bounce rate ≥ 3% over a rolling 20 sends · a single spam complaint · any provider warning on the domain · Gmail Postmaster reputation below "High". Halting means: pause every sequence, send nothing for 72 hours, check /admin → Email health, resume at half the previous rate.
The five things that actually build reputation
Volume ramps matter least. These matter more, and they're free.
| Verified addresses only | All 12 round-1 emails came back Apollo-verified. One dead mailbox at 3 sends/day is a 33% bounce rate for that day. |
| Replies, not opens | The single strongest positive signal to Gmail and Outlook. SEQ-A step 1 asks a question answerable in four words because a short reply beats a long read. |
| Reply in thread | Step 4 replies to step 1 rather than starting a new subject. Threads read as correspondence, not campaign. |
| No links or attachments in email 1 | Both carry spam-filter weight. The Renewal Runway doesn't appear until step 4, after one message has landed clean. |
| Instant opt-out | Apollo does it automatically. Verify once by opting a test address out and confirming suppression. |
6 · The sequences, as written
Three, not one. The 25 Aug enrichment proved these are three different businesses hiding in one spreadsheet, and a single sequence would be wrong for two thirds of the list. Full copy: docs/APOLLO-CAMPAIGN-R1.md.
| Sequence | Who | Live now | Opening offer |
|---|---|---|---|
| SEQ-A · Renewal Runway | S1 — 20–90 employees, real group-health opportunity | 9 contacts / 7 accounts | Level-funded vs fully-insured, timed to renewal |
| SEQ-B · The Participation Gap | S3 — plan exists, most staff aren’t in it | 3 contacts / 2 accounts FINRA gate | Fiduciary + fee benchmark. No product pitch. |
| SEQ-C · Referral partners | S5 — CPAs, attorneys, payroll firms on the list | 7 identified, 0 sequenced | Mutual referral. Never automated. |
Hard cap for round 1: 12 contacts, 9 accounts. Deliberate. The sending domain cannot absorb more than this in month one without risking the domain that carries client mail.
SEQ-A · eight steps over 24 days
| # | Day | Type | What |
|---|---|---|---|
| 1 | 0 | The renewal-date question | |
| 2 | 2 | Call | Direct line. No voicemail on attempt 1. |
| 3 | 4 | Connect, no note. A note converts worse. | |
| 4 | 6 | Reply in thread + the Renewal Runway | |
| 5 | 9 | Call | Voicemail this time, referencing the email |
| 6 | 13 | Drop | Printed Runway to the front desk — everyone is inside 22 minutes |
| 7 | 17 | The 5500 Snapshot — the one they actually open | |
| 8 | 24 | Task | Breakup, then nurture. Not trash. |
SEQ-A · step 1, day 0
Subject: quick one about {company}’s renewal month
{first_name} —
Michael Hipson, benefits advisor here in west Houston — about {distance} from your office on {street}.
I work with Houston employers in the 20–100 range on group medical, and the single thing that decides whether a renewal goes well is how early it starts. Most brokers open the file 30 days out. That’s too late to do anything but accept the increase.
Two questions, and no pitch either way:
1. What month does your medical plan renew?
2. Has anyone shown you a level-funded comparison in the last two years?
If the answer to #2 is no and you’re between 20 and 100 lives, that’s usually worth an hour. If it’s yes, tell me who and I’ll leave you alone.
Michael Hipson · Hipson Investments · Houston, TX · [address] · [one-click unsubscribe]
No attachment, no link, no claim, one question a human answers in four words. Nothing in it triggers FINRA 2210. The distance line is the only personalisation and it is true for all nine — closest 13.2 mi, furthest 21.9 mi.
SEQ-A · step 4, day 6 — reply in thread
Subject: (reply — no new subject)
{first_name} — following up once.
Attached is the thing I’d actually want if I were in your seat: a one-page 90-Day Renewal Runway — what has to happen in each of the three months before a renewal date, and what it costs you if it doesn’t.
No form, no email capture. If it’s useful, use it — you don’t need me to run it.
If you’d rather I just tell you what {company}’s public filings suggest about your plan, say the word and I’ll send that instead.
SEQ-A · step 7, day 17 — the 5500 Snapshot
Subject: {company}’s Form 5500, in one page
{first_name} — last one from me.
{company}’s Form 5500 is public. I built a one-page read of yours — participant count against headcount, plan assets, and what the filing implies about take-up. It took me four minutes and it’s attached.
The line I’d look at first: {participants} participants against roughly {employees} employees.
That’s either fine or it’s an expensive problem, and the filing can’t tell you which. A conversation can.
FINRA-gated where the account has a retirement plan. Principal review before first use.
SEQ-B · step 1 FINRA gate — held
The premise is a real one: Dynamic Engineers has 39 employees and 7 plan participants. 82% of the staff aren’t in the plan. That isn’t a sales angle, it’s a fiduciary fact the owner may never have been shown.
Subject: 7 of 39 at {company}
{first_name} —
Michael Hipson, Houston. I look at Form 5500 filings for local employers, and {company}’s stopped me.
Your most recent filing shows {participants} plan participants. Public sources put your headcount around {employees}.
There are good explanations for a gap that size — eligibility waiting periods, a lot of recent hires, high turnover in one department. There are also expensive ones.
I’m not selling you a plan. I’m asking whether anyone has benchmarked the fees and the participation rate against comparable Houston plans in the last three years. If nobody has, you’re carrying a fiduciary obligation you can’t evidence.
Happy to send the benchmark scorecard I use. It’s free and it names its sources.
Does not send until principal review is recorded. Steps 2–6 mirror SEQ-A’s cadence, substituting the Fee & Participation Scorecard for the Renewal Runway.
SEQ-C · not a sequence
Seven referral partners identified — including Cyphersix, whose sponsor phone reaches Frank J Lucka and Associates, an accounting firm. That was logged as bad data on 14 Aug. It isn’t; it’s a referral lead. One email, one call, one coffee, run by Michael personally. Automating it would destroy the only thing that makes it work.
Per-account overrides that must be respected
| Account | Override |
|---|---|
| Nationwide Coins | Apollo matched “Nationwide Coins” to the filing’s “Nationwide Coin & Bullion Reserve”. Verify by phone before any send. |
| Dynamic Engineers | Email domain is dynamiceng.com, not the web domain. Andrea Teoh already had an in-person drop — reference it, don’t pretend it’s first contact. |
| Upstream Engineering | Lead with Marisol Burk (VP Administration), not the President. She scores higher and admin VPs own the renewal file. |
| AIG Technical Services | Apollo marked Priscilla Freeman “Disqualified”. Override. At a 43-person firm the office operations manager is the benefits decision. |
| H-Town Technologies · Compansol | The 5500 contact is a TPA. Write only to the Apollo-found contact. |
| BioUrja · Radix | Removed from round 1. 600 and 1,900 employees — re-graded out of small-group entirely. |
Your day, in five minutes
Open Tasks in Apollo each morning. Work top to bottom. That’s the whole job. Full walkthrough: docs/APOLLO-TEAM-RUNBOOK.md.
| You see | You do |
|---|---|
| An email task | Read the draft. Change at least one thing so it isn’t identical to the last nine. Send. |
| A call task | Ring the direct line. First attempt: no voicemail. Second: leave one, referencing the email. |
| A LinkedIn task | Connect. No note. |
| A drop task | Print the Runway, drive it over, hand it to the front desk. |
| A reply | Stop the sequence for that person immediately and answer as a human. |
When someone replies, the campaign is over for them. The worst thing Apollo can do is send step 5 to someone who answered step 3. It usually detects replies and pauses. Usually is not always. Check.
7 · Lead magnets — built 25 Aug
Three, to the brand guide, all print-clean at Letter size because half this list gets visited rather than emailed. Each carries its own disclosure block.
The 90-Day Renewal Runway live
A three-phase countdown from a group medical renewal date — what to do at 90, 60 and 30 days out, and what each omission costs. Every phase ends with a “cost of skipping” line, because a checklist without consequences doesn’t get used.
The strategic choice worth naming: it is genuinely useful without us. It tells an employer how to run their own renewal properly. That’s the point — it earns the right to the second email, and an employer who works through it and finds they can’t produce 24 months of claims data has diagnosed their own problem and reached our conclusion by themselves.
No form. No email capture. Gating a one-page worksheet from a nine-person campaign would cost more in reply rate than the addresses are worth. Lands at SEQ-A step 4.
The Form 5500 Snapshot live · noindex
The killer asset, and the reason this campaign is different from every other broker’s. One page per prospect, built entirely from their own public Form 5500 plus the employee count Apollo just gave us: participants against headcount, plan assets, average balance, a participation gauge, and a plain-English verdict that changes with the numbers. Pre-loaded with all nine round-1 accounts, printable per account.
Why it works: it required nothing from them. Every competitor’s first touch asks for census data, claims history, or “fifteen minutes to learn about your business”. This one arrives already knowing, and shows its work.
Dynamic Engineers’ snapshot renders the sentence the whole campaign turns on: 32 of 39 people are outside the plan.
Internal, noindex — it names specific companies. Leave-behind and SEQ-A step 7.
401(k) Fee & Participation Scorecard live FINRA gate
Twelve yes/no questions across fees, participation and the fiduciary evidence file, self-scoring in the browser with nothing sent anywhere. Score bands are written to be honest rather than alarming — the 6–9 band, where most decent small plans land, says plainly that the gaps are documentation rather than decisions.
It creates a gap the reader discovers themselves. A sponsor who can’t answer “have your fees been benchmarked in the last three years” has been told nothing by us and has learned something uncomfortable. That is a better opening than any claim we could make. Attaches to SEQ-B.
Already live and mapped elsewhere: /aca-review · /aba screener + Texas ABA Playbook EN/ES · /health quiz · /quote · /flyer · /group-intake.
Still unbuilt, ranked: Benefits Benchmark questionnaire (S1, 8–10 taps, same pattern as /aca-review) · co-marketing one-pager (S5) · “Houston Small-Group Market” quarterly report (S5/brand, built from public EFAST aggregates).
8 · Funnels and the other three rails
Where each segment lands
Every path terminates in a tool we already own. Only the Benefits Benchmark is new, and it is a schema swap on the existing tap-only pattern.
S1 cold email → 5500 Snapshot → /group-intake or Benefits Benchmark → discovery call S3 cold email → Fiduciary File Checklist → 15-min fee benchmark call [Ameritas gate] S4 practice call/visit → co-branded family screener → /aba or /health → /quote S5 intro coffee/call → co-marketing one-pager → quarterly briefing list C quarterly newsletter only — no sequence, no spend
Wire before the first send: UTM tags on every sequence link (utm_source=apollo, utm_medium=email, utm_campaign=s1-round1) so replies and page hits reconcile, and enrichOnLeadCreate so the team sees company size before the same-day callback.
Email is one rail of four — and not the strongest here
| Rail | Why it matters on this list |
|---|---|
| Door drops | Five already logged. The list clusters hard: 2000 S Dairy Ashford holds four sponsors, 1001 S Dairy Ashford holds three, and Enclave Parkway, Eldridge Parkway and Highway 6 hold most of the rest. One afternoon covers a dozen. A printed Snapshot hand-delivered outperforms any email in this plan. |
| Calls | Nine outcomes already logged. Touch 2 of every sequence, made by a person. |
| Connect and follow A/B decision makers before touch 1. No pitch in the request. | |
| Field QR → /flyer | Already live. Put it on the printed Snapshot. |
| Partner referrals | S4 and S5 compound over time. Email does not. |
9 · Guardrails — non-negotiable
| System of record | Firestore, never Apollo. Export terms in writing, append-only log, no deletes. |
| A human sends | Every recommendation comes from a licensed human. Drafts reviewed before enrollment. |
| TCPA | No automated SMS, no AI-voice, to any cold list. |
| CAN-SPAM | Identity, physical address and a working unsubscribe on every send. |
| FINRA 2210 / SEC | No 401(k) or investment performance claims in any outreach — ever, cleared or not. |
| Never fabricate | The 50-life case study is the only outcome figure we hold, used exactly as consented. |
| Ameritas gate | Gates the retirement pillar. Benefits, ACA and payroll outreach are unaffected and can run now. |
| Sending | Cold volume never leaves hipson.com (Exchange, all four staff), and the Resend key never touches the Apollo rail. |
| Credits & PII | Enrich A/B deep, C shallow, company search batched. BenefitZone Export/ stays gitignored; enrichment writes behind the admin gate only. |
10 · What we measure, and by when
Four numbers decide whether round 2 gets bigger or gets fixed. The 30-day review is 24 Sep.
| Metric | Target | Status 25 Aug |
|---|---|---|
| Domains resolved | — | 21 of 34 shortlisted · 13 aren’t in Apollo and never will be |
| Verified decision-maker email | — | 12 across 9 accounts, all verified |
| Credits spent | ≤ 44 budgeted | 12 of 85 — 73 remain |
| Renewal month known | ≥ 3 of 9 | 0 — needs the EFAST2 Schedule A pass |
| Bounce rate | < 2% | No sends yet |
| Spam complaints | 0. Anything else is a failure regardless of meetings | No sends yet |
| Reply rate | 4–8% benchmark · below 3% means the copy is wrong, not the list | No sends yet |
| Booked conversations | ≥ 1 in 30 days | — |
Round 1 fails if it produces a meeting and a complaint. The complaint costs more than the meeting is worth, because the domain also carries client mail.
Calendar
| When | What | Bar |
|---|---|---|
| Wk 1 25–29 Aug | Create michael@mail.hipsoninvestments.com, connect to Apollo. Build the three sequences from the spec. Verify Nationwide Coins by phone. SEQ-A step 1 to 9 contacts at 3/day. | Mailbox live · 9 sent clean |
| Wk 2 1–5 Sep | SEQ-A steps 2–4. First calls and the Runway drop. FINRA principal review of SEQ-B and SEQ-A step 7. | First replies · 0 complaints |
| Wk 3 8–12 Sep | Trial decision made by 7 Sep. SEQ-B opens if review cleared. EFAST2 renewal-month scraper started. | Plan bought or programme stopped |
| Wk 4 15–19 Sep | Full cadence, 12/day ceiling. 5500 Snapshots sent. Firestore mirror live in /admin. | ≥1 booked conversation |
| 30-day review 24 Sep | Read the four numbers. If complaints are 0 and replies clear 4%, register go.hipsoninvestments.com and scale round 2 there, leaving mail. clean. | Round-2 go / no-go |
11 · Cost, and what is still open
Round 1 cost $0. Everything ran inside the free Professional trial. The bill starts on 7 Sep or not at all.
| Option | Monthly | When it’s right |
|---|---|---|
| Drop to Free | $0 | Round 1 produces nothing. Free cannot search or enrich meaningfully — this ends the programme. |
| Professional, monthly, 1 seat | ~$79 | Recommended. Round 1 produces ≥1 booked conversation. Monthly, not annual, until 90 days prove volume. |
| Organization, monthly | ~$119/user, 3-seat min | Only if the Cloud Function automation in APOLLO-MAX-PLAN.md is actually being built. Full API access lives here — and it isn’t needed until enrichOnLeadCreate exists. |
Open items, with owners
| Item | Owner | By when |
|---|---|---|
| Trial decision — Professional monthly, or let it lapse | MT | 7 Sep — hard deadline |
Create michael@mail.hipsoninvestments.com, connect to Apollo | MT | Before first send |
| FINRA principal review — SEQ-B and SEQ-A step 7 | Justin / principal | Before retirement content ships |
| Verify Nationwide Coins identity by phone | Justin | Before that account is sequenced |
| Build the three sequences inside Apollo from the spec | Justin | Week 1 |
| Mirror contacts + outcomes into Firestore | Dev | Week 1 — see §12 |
| EFAST2 renewal-month scraper — free, public, scores +40, still the highest-value missing field | Dev | Round 2 |
Decide on go.hipsoninvestments.com | MT | 30-day review, 24 Sep |
| Whose name sends — Michael H, Justin, or both | Mike | Sequence build |
| Ameritas compliance review | Michael H | All S3 retirement outreach |
The Apollo MCP connector now shows Connected in the Claude directory, but connectors attach at session start — it should be usable next session, which would replace most of the manual browser work.
12 · Firestore mirror — Hipson stays the system of record
The BenefitZone rule, restated: Apollo is a data source, not the pipeline. Truth lives in Firestore and shows in /admin.
Field on prospects.{id}.apollo | What it holds |
|---|---|
domain · domainConfidence | Resolved domain and how certain we are: PHONE-EXACT · PHONE-BLOCK · SOLE-HIT · VERIFY |
employeesApollo | The field that re-grades everything. Overrides participants for all segment logic. |
contacts[] | name · title · email · emailStatus · apolloPersonId · revealedAt |
sequenceId · sequenceStep · lastTouchAt · replyAt | Where each contact is in the campaign |
optedOutAt | Set → permanently suppressed everywhere, Apollo and Hipson both. One-way flag; nothing clears it. |
creditsSpent | Running total against the budget |
Rules the mirror enforces at write time, not by convention: doNotSequence: true on C-Mar and Gush · a contact whose name matches the six-name TPA blacklist is rejected, not warned · employeesApollo wins over participants everywhere.