SQL model and methodology.
Where the S/M/L expectation numbers in the sales script come from, every step of the math, and where the number is a measured floor versus a modeled assumption. Built 14 Jul 2026.
Confidential, internal only. The underlying funnel data is from an outbound campaign Sofia ran through her prior agency, before Spice GTM existed as a company. The client is not named anywhere on this page and the raw counts below must never be attributed to a specific company outside the team. This is the same operator-track-record framing used in Results & timeline on the sales script: proof of what the operators can run, not a Spice GTM client result.
1. Source data
Sofia's original funnel breakdown, screenshots pending upload here.
PendingEmail funnel screenshot (SmartLead). Drop the file at assets/sofia-email-funnel.png and it replaces this block.
PendingLinkedIn / GetSales funnel screenshot. Drop the file at assets/sofia-linkedin-funnel.png and it replaces this block.
2. Raw funnel counts
Email (SmartLead)
| Stage | Count | Conversion |
|---|---|---|
| Sent | 60,734 | 100% |
| Delivered, estimated | 59,842 | 98.5% of sent |
| Human replies | 335 | 0.55% of sent |
| Warm / positive replies | 96 | 0.16% of sent |
| SmartLead meeting requests | 34 | 0.056% of sent |
| Warm reply → meeting request | 34 / 96 | 35.4% |
| Booked meetings, per booking list | 43 pending | 0.071% of sent, final |
Booking count marked pending at capture time, not yet confirmed final by Sofia. Treat 43 and the 0.071% derived from it as provisional until she locks it.
LinkedIn / GetSales (OnSocial flows only, 2026-01-01 to 2026-07-13)
| Stage | Count | Conversion |
|---|---|---|
| Connection requests sent | 4,667 | 100% |
| Accepted connections | 1,097 | 23.5% of requests |
| Messages sent | 1,004 | 91.5% of accepted |
| Replies | 180 | 17.9% of messages |
| End-to-end reply | 180 / 4,667 | 3.9% of requests |
| Booked meetings, per booking list | 14 pending | 0.30% of requests, final |
Filtered to OnSocial flows only. If Sofia ran other LinkedIn flows in the same window, this understates true LinkedIn output, worth checking before treating 0.30% as the ceiling.
3. Statistical margin
Both final rates are built on small counts. Treat the headline percentages as midpoints of a real range, not exact figures.
| Metric | n | Rate | Approx. 95% range |
|---|---|---|---|
| Email, sent → booked | 43 | 0.071% | ~0.05% to 0.09% |
| LinkedIn, sent → booked | 14 | 0.30% | ~0.15% to 0.45% |
Relative error scales roughly with 1/√n. At n=14 a single extra or missing booking moves the rate by around 7%, which is why the LinkedIn range is wide. Do not read either final percentage as more precise than it is.
4. Unit economics, full run-rate
Applying the rates above to one unit of infrastructure at full sending volume, no ramp. 35/day is the safe rate per mailbox, not per domain. Each domain runs 3 mailboxes, corrected 14 Jul 2026, this was previously modeled as 35/day for the whole domain, a 3x understatement of email volume.
| Unit | Full volume | SQL / month |
|---|---|---|
| 1 mailbox | 35/day × 22 sending days = 770/month | ~0.5-0.6 |
| 1 domain (3 mailboxes), rounded | 100/day × 22 sending days = 2,200/month | ~1.5-1.6 |
| 1 LinkedIn account | 150/week × ~4.3 weeks = ~650/month | ~1.8-2.0 |
5. Ramp: why month 1 is lower than month 3
Two things are stacked here, kept separate on purpose.
- Real timing. Retainer billing starts month 3 (Pricing, sales script). Months 1-2 are the fixed-price setup, during which domains and LinkedIn accounts are built and warmed. By the time a client is paying for SQL delivery, the infrastructure is not starting from zero.
- Deliberate buffer. Even with that head start, month 1 of billed growth is priced at roughly 50% of the month 3 ceiling, month 2 at roughly 75-80%. This is not a measured warmup curve, it is a chosen margin for what only shows up in live production: a specific ICP's real list quality, spam-filter behavior on that client's exact domains, anything the pooled historical data does not capture. The goal is a number the team clears, not one it chases.
6. Scale efficiency: why L is not 3x S
Domains scale linearly 5 → 10 → 15, straight 3x. SQL output does not, on purpose.
| Tier | Linear full run-rate | Month 3 ceiling, set | Efficiency vs. linear |
|---|---|---|---|
| S | ~14 | 10-12 | ~79% |
| M | ~25 | 13-17 | ~60% |
| L | ~39 | 15-20 | ~46% |
Be honest about what this is. The month 3 ceilings were set directly by Pavel on 14 Jul 2026, not derived independently from data, then the efficiency percentages were back-calculated to explain them. The mechanism behind the taper is real and defensible: more domains and LinkedIn seats aimed at the same ICP hit the same finite pool of accounts, so each added unit of infrastructure converts at a lower rate than the first units, an outbound program does not scale linearly forever. But the exact curve, 79/60/46%, is a fit to a target, not a measurement. Do not quote these percentages to a client as if they were observed.
7. Final output: S / M / L
| Tier | Infrastructure | Month 1 | Month 2 | Month 3+ |
|---|---|---|---|---|
| S | 5 domains (15 mailboxes), 3 LinkedIn | 5-6 | 8-9 | 10-12 |
| M | 10 domains (30 mailboxes), 5 LinkedIn | 7-8 | 11-12 | 13-17 |
| L | 15 domains (45 mailboxes), 8 LinkedIn | 8-9 | 13-14 | 15-20 |
These are the exact numbers on the sales script's Expectations section. If either changes, change both in the same sitting, they are meant to stay identical.
8. Open items
- Booking counts pending. 43 (email) and 14 (LinkedIn) were not final at capture. Re-run this whole page once Sofia locks the booking list.
- Domain aggregate volume, unverified. ~100/day per domain (3 mailboxes × ~35/day, rounded) is arithmetic, not a tested ceiling. Some providers rate-limit at the domain level, not just per mailbox. Confirm with Sofia that a domain has actually run at 100/day sustained before this number goes into a client-facing commitment.
- Tier sizing, revised three times same day. Started at 8/10/14 domains, moved to 10/20/30, then 5/10/15, LinkedIn fixed at 3/5/8 throughout. Month 3 SQL then reset from straight unit-economics scaling to a flatter, Pavel-set curve (section 6). If any of this changes again, update this page and the sales script together, they must never show different numbers.
- Email rate versus the objection script. The sales script's "already 80% done" objection calls a prospect's 0.1% conversion spam. Our own floor, 0.071%, is below that. The objection line was reworded 14 Jul 2026 to argue system-vs-static-list instead of a percentage comparison we cannot currently win. Revisit once the new engine (signal-based targeting, Yaroslav) has live data and the floor actually clears 0.1%.
- LinkedIn scope. OnSocial flows only. Confirm with Sofia whether other flows ran in the same window; if so, 0.30% and the LinkedIn unit economics are understated.
- Time window mismatch. LinkedIn window is stated (2026-01-01 to 2026-07-13, ~6.4 months). Email window was not, so the two channels' "per month" figures are not guaranteed to cover the same period. Confirm before leaning on any month-over-month comparison between channels.
- New engine, not yet measured. Everything above reflects the old, static-list engine. Once Yaroslav's signal-based targeting and segmentation are live with a real client, replace the floor rates with fresh data rather than assuming an uplift.
Built 14 Jul 2026, revised three times same day · source: funnel data from Sofia (SmartLead + GetSales/OnSocial), client withheld · revision 1: 35/day is per mailbox not per domain, 3 mailboxes per domain, email volume corrected 3x up · revision 2: tier grid resized to 5/10/15 domains (from 10/20/30), domain rate rounded to 100/day, LinkedIn fixed at 3/5/8 · revision 3: added scale-efficiency taper (section 6), month 3 ceilings reset to Pavel-set 10-12 / 13-17 / 15-20 instead of straight-line unit-economics scaling · feeds the Expectations section of the sales script, keep both in sync