BusinessNotes/wiki/queries/2026-07-26-eugene-90-day-plan.md

# Eugene's Concrete 90-Day Plan

#query #sales #planning

## Question asked

> Turn the corpus into [[eugene]]'s concrete 90-day channel/niche plan — the dated
> version [[2026-07-17-best-method-first-client]] stops short of — and resolve the live
> tension inside it: work the in-person first-client route, or build a
> [[marketing-system]] early?

## Assumptions this plan rests on

Stated up front so the plan is falsifiable rather than merely plausible. Each is a vault
inference, not a sourced fact:

| # | Assumption | Status |
|---|---|---|
| A1 | [[eugene]] is the vault owner | `tentative` — circumstantial (insin, the harness) |
| A2 | His target buyers are industrial-equipment vendors / machine-builder integrators needing CV & embedded work | `tentative` — vault inference, never stated by him |
| A3 | His natural deal shape is few × $10K+, not 100 × $1K/mo | `tentative` — inferred from domain |
| A4 | He has *some* prior paid work to mine for a result-pattern | **unverified** — if false, Week 1 step 2 drops and Week 2 carries more load |
| A5 | Enterprise-lite industrial buyers behave closer to [[sebastian]]'s enterprise than to [[dan-martell]]'s SMB | `tentative` — this is the single assumption the channel choice hangs on |

If A2 or A5 turn out wrong, the channel mix below is wrong and the online tier reopens.

## Answer — the short form

**One-to-one, in-person-weighted acquisition is the engine for these 90 days. The
[[marketing-system]] gets a small fixed time deposit — not because it produces leads in
this window (it won't), but because its clock is ~6 months long and because publishing
*is* the articulation practice that may be his actual blocker.**

That is the resolution of the live tension, and it is not a compromise — it follows from
the vault's own stage reading.

### Why the tension resolves this way

The vault had already half-inferred it; [[oskar-hartmann]] nearly states it outright
(*one repeatable channel builds the company; multichannel is a resilience property of
scale*). Three supports:

1. **The three-lever map is addressed to the wrong stage.** [[marketing-system]]'s
   pick-one-lever rule comes from a clip aimed at a founder with a stalled $1.5M. It
   omits outbound, [[referrals]], and in-person events entirely — i.e. the whole
   first-client toolkit — because those aren't *scalable* levers, not because they don't
   work ([[client-acquisition-channels]]).
2. **The same author runs one-to-one at $0.** His own $0→$100K blueprint prescribes
   phone-mining and cold calls ([[2026-07-23-make-my-first-100k-in-month]]). The
   pick-one-lever rule is stage-scoped by his own practice.
3. **The two poles agree about *this* buyer.** On enterprise specifically,
   [[sebastian]] and [[dan-martell]] — who agree on nothing else about channels —
   both say cold outreach doesn't open the door and trust does; they differ only on
   built vs. borrowed trust ([[relationships-as-moat]], [[partnerships]]).

**But the system isn't deferred to zero**, for one reason the stage reading alone misses:
[[marketing-system]]'s six-month lag means "start it when you need it" starts it six
months late. And publishing doubles as reps at explaining — the fix prescribed by the
*rival* diagnosis of Eugene's blocker ([[technical-founder-trap]]). A small deposit buys
both. It is budgeted below as ~2 h/week and explicitly **not** measured on leads.

### The channel count, settled for this case

[[ab-analytics]] says 3 channels × 90 days; [[dan-martell]] says 1 lever × 90 days
(`the vault's standing conflict`). Under the units reading the vault infers — a *lever*
is a discipline, a *channel* is a venue — both are satisfiable at once: **one lever
(relationship-mediated one-to-one), three venues.**

| Venue | Tier ([[client-acquisition-channels]]) | Why this one |
|---|---|---|
| **1. Warm mining of existing contacts** | Common / [[referrals]] | The only referral-shaped motion that needs no past client; cheapest, fastest |
| **2. One recurring industry room** — machine-builder / automation association, regional trade fair, not a generic Chamber | Low-key (in-person) | The tier both [[ab-analytics]] and [[sebastian]] rate highest; "3× = regular, 6× = trusted" needs *recurring*, so pick by calendar not prestige |
| **3. Partner-archetype recruiting** — integrators & equipment vendors who already hold the buyer | Out-of-the-box | [[partnerships]]: borrowed credibility. Per [[oskar-hartmann]]'s warning, court **many small partners, never one anchor** |

Deliberately excluded: cold email/LinkedIn campaigns at volume (A5 says this buyer
ignores them, and it is the effort that already failed him for ~6 months); the 100 × $1K/mo
money map (wrong deal shape, A3); building anything before payment.

## The plan — 2026-07-26 → 2026-10-24

### Week 1 (Jul 26 – Aug 1) — Name the target, fix the words

1. **Write the SOM in one sentence** and list **10 named prospects** — companies, not a
   category ([[tam-sam-som]], [[sell-before-build]]). If the list can't reach 10, the
   niche is wrong or too narrow; that failure is the week's most valuable output.
2. **Old-client mining** (per A4): call 3–5 past clients/employers — *"what result were
   you paying me to get?"*, never *"why did you choose us?"* ([[pain-discovery]]).
3. **Offer formula:** *"we do X for [ICP] to close Y pain and get Z result."*
4. **Commodity test:** strip the words "development", "computer vision", and every
   technology name. What's left? If nothing, there is no offer yet
   ([[productized-service]]).
5. **The articulation rep:** say the offer to one non-engineer and have them repeat it
   back. A [[referrals|referral]] only travels if the referrer can restate it in a
   sentence — which is why this sits in Week 1 and not in the content bucket
   ([[technical-founder-trap]]).

### Week 2 (Aug 2 – Aug 8) — The diagnostic that costs an afternoon

Run the **phone mine**: go through existing contacts and *ask past the person* — "do you
know anyone dealing with [the pain]?" Often lands on "yeah, me"; otherwise yields
warm, name-carrying intros ([[referrals]], [[2026-07-23-make-my-first-100k-in-month]]).

**This is also the experiment that discriminates between the vault's two rival diagnoses
of his 6-month blocker** — a test the vault has flagged as needed but never specified:

| What happens | Diagnosis | What to fix in Weeks 3+ |
|---|---|---|
| Contacts engage warmly but **can't restate** what he does | [[technical-founder-trap]] — articulation | Reps at explaining; the content deposit earns its keep |
| Contacts restate it fine but **know nobody** to send | [[relationships-as-moat]] — network reach | Weight venue 2 & 3 harder; the rooms are the work |
| Neither — low response | Relationship decay, not either diagnosis | Rebuild contact-by-contact before scaling any channel |

Vault synthesis; no source proposes this test.

### Weeks 3–4 (Aug 9 – Aug 22) — Open the three venues

- Book the **recurring** industry room through December. Recurrence is the mechanic, not
  attendance ([[relationships-as-moat]]).
- Identify 5–10 candidate **partner** organizations; approach as peers, not as a vendor
  ([[partnerships]]).
- First touches on the 10 named prospects, **each customized** — the template is a
  starting point, never a mass send ([[2026-06-15-linkedin-mail-template]]).
- **Follow-up cadence: Day 0 → 3 → 7 → 14 → 30**, 5 touches minimum, on every lead
  ([[sales-discipline]]).
- Start the **~2 h/week content deposit**. Measured in reps, never views. Scrambled order
  if published as know-how ([[information-vs-implementation]]).
- In every conversation, surface budget *before* presenting anything — obstacles, not
  objections ([[sales-discipline]]).

### Month 2 (Aug 23 – Sep 19) — Sell the smallest paid thing

This answers the vault's own open question — *what is the smallest paid test for a
dev-services offer?* ([[sell-before-build]], flagged there as arguably this plan's
missing first step):

> **A fixed-price, fixed-scope paid diagnostic** — a feasibility audit or discovery
> sprint on **one** inspection line / one device / one process.

It satisfies four constraints at once: money lands **before** the build (the services
analogue of the paid waitlist — unpaid interest doesn't count); it is small enough to
clear without procurement (dodging the whale-client trap [[oskar-hartmann]] documents);
it manufactures the **Before/After case study** the corpus has flagged as missing for
him; and it is a legitimate entry rung of an [[offer-ladder]].

Also this month:

- **Wizard-of-Oz the delivery.** First results produced semi-manually; automate only what
  repeats ([[sell-before-build]]).
- **Pricing-power probe** on the second quote — raise it and watch. Cheapest PMF test in
  the vault ([[pricing-from-value]]).
- **Ask for referrals immediately on delivery**, while the enthusiasm window is open —
  91% would, 11% are asked ([[referrals]]).

### Month 3 (Sep 20 – Oct 24) — Standardize only what repeated

- After **2–3 similar paid audits**, fix scope/price/name → a productized package
  ([[productized-service]]). Not before: the offer-first vs delivery-first dispute is
  unresolved, and this side of it costs nothing to obey.
- Write the case study as outcome + accountability, not scope
  ([[outcome-based-selling]]).
- Try the Zendesk motion on the warmest prospect: *"how much do you spend on X?"* →
  *"and if it were 10× less, would we talk?"* ([[2026-06-15-rodenko-selling-development-expensively]]).
- **Do not add a fourth venue.** Channel-hopping is the failure mode both rules agree on.

## The Day-90 gate (2026-10-24)

Review on **inputs**, decide on **outputs** — the reps-not-views substitution applied to
the plan itself ([[sales-discipline]], [[marketing-system]]).

**Inputs — did the plan actually run?** (target: all four yes)

- ≥ 10 named prospects worked to 5 touches each
- ≥ 6 appearances in the recurring room (the "trusted" threshold)
- ≥ 5 partner conversations opened
- ≥ 3 paid diagnostics *offered* (offered, not necessarily sold)

**Outputs — what they mean for the next 90 days:**

| Outcome | Read | Next |
|---|---|---|
| ≥ 1 paid diagnostic sold | Channel works | Repeat, productize, keep the system deposit small |
| 0 sold, but inputs all hit | The channel or A2/A5 is wrong | Change **one** variable — niche before channel |
| Inputs missed | No evidence either way | Re-run; nothing has been tested yet |
| Sold, but only via venue 1 | Warm network worked, cold rooms didn't | This is the [[referrals]] ceiling arriving early — *now* the [[marketing-system]] decision becomes live |

Thresholds are vault-inferred, not sourced. Their function is to be pre-committed, not
to be correct.

## Evidence trail

- Stage reading (one-to-one starts, systems scale); three-lever map's omissions — [[client-acquisition-channels]], [[marketing-system]]
- One channel then multichannel; SOM > TAM; whale-client trap; partner-hope warning — [[2026-07-26-how-to-build-a-billion-dollar-company-2027]] ([[oskar-hartmann]]), [[tam-sam-som]], [[partnerships]]
- Signal hierarchy, Wizard-of-Oz, paid-waitlist analogue, duration-without-revenue red flag — [[sell-before-build]], [[2026-07-26-main-principle-of-successful-business]]
- 3-tier channel taxonomy, 3×90, 3×/6× recognition, Day 0→3→7→14→30, 91%/11% — [[2026-06-15-17-ways-first-client]] ([[ab-analytics]]), [[sales-discipline]]
- In-person trust mechanics; "Big zero" on online outreach — [[relationships-as-moat]], [[2026-07-06-sebastian-interview-ai-and-software-engineering]] ([[sebastian]])
- Phone mining / ask past the person; obstacles-not-objections; offer → pre-sell → build — [[2026-07-23-make-my-first-100k-in-month]] ([[dan-martell]])
- Six-month lag, reps-not-views, referral dependency as ceiling — [[2026-07-20-referrals-will-sink-your-business]]
- Borrowed credibility, partner-archetype recruiting — [[2026-07-22-stop-cold-calling-do-this-instead]], [[partnerships]]
- Week-1 sequence (old-client mining, offer formula, commodity test) — [[2026-06-15-meta-analysis-selling-dev-in-ai-era]]'s 90-day roadmap, [[pain-discovery]], [[niche-selection]]
- Pricing-power probe, guarantee test — [[pricing-from-value]]
- Predecessor query (method, undated) — [[2026-07-17-best-method-first-client]]
- Eugene's blocker, deal shape, prior LinkedIn/articles effort — [[eugene]]

**Caveat on evidence quality.** Every channel and sales claim underneath this plan traces
to promotional YouTube videos and stage talks — *attributable, not verified*
([[overview]] Contradictions). The vault still holds **no failure-case source for the
productized / sell-dear model itself**, so nothing here has been tested against a
documented loss. The one adversarially-verified source in the vault
([[ai-productivity-evidence]]) touches the AI-capability premise, not these tactics.
Treat this as a pre-committed experiment design, not a playbook known to work.

## Follow-up questions

- **Which specific association / trade fair** serves machine-builders and industrial
  integrators in his region, and does it meet often enough for the 3×/6× mechanic? The
  plan's venue 2 is unnamed and that is its weakest joint.
- **Will an industrial buyer pay for a diagnostic at all**, or is a free feasibility
  assessment the market norm? If free is the norm, the smallest paid test must move
  downstream and Month 2 needs redesign.
- Does the net-60/90 enterprise payment reality break the pre-payment rule for services
  buyers of this size? (Open on [[sell-before-build]].)
- Is industrial CV/embedded a **beachhead or a fortress** niche? Changes what winning
  the SOM should lead to (open on [[tam-sam-som]]).
- Still outstanding and cheap: confirm/refute the two suspected [[dan-martell]]
  attributions — several rungs of this plan rest on his corpus, and it matters whether
  they are 5 or 7 of 20 sources. *(Updated by lint 2026-07-29: a third anonymous
  source — the Claude-Code short — was owner-attributed to him that day, moving the
  confirmed count 4 → 5 and making this plan's reliance on his corpus heavier, not
  lighter.)*

## Whether this output changed existing pages

Yes — five pages updated, no source claim altered. This page is synthesis over existing
pages plus four vault-original contributions: the **stage-based resolution of the
one-to-one vs. system tension** (with the six-month-lag exception that keeps the deposit
non-zero), the **phone mine as the discriminating test between the two rival diagnoses**
of Eugene's blocker, the **paid diagnostic as the smallest paid test for dev services**
(an open question on [[sell-before-build]]), and the **pre-committed Day-90 gate**. All
four are inference, marked as such.

Updated: [[eugene]] (Next Questions closed; Related Pages), [[sell-before-build]] (its
smallest-paid-test question now has a proposed answer), [[client-acquisition-channels]]
(its Eugene channel question now answered), [[2026-07-17-best-method-first-client]]
(forward link to its dated successor), [[overview]] ("Application is unstarted" thread).
Cataloged in [[index]], logged in [[log]].