Account architecture — and why the founder's account never runs cold outbound.

The central premise — account-as-asset, not account-as-channel

In email, the unit of reputation is the domain, and the domain is a fungible $12 line item at a registrar. A burned domain is replaceable in fifteen minutes. The asymmetry between provisioning and recovery effort is real but bounded.

On LinkedIn, the unit of reputation is the account, and the account is anchored to a legal identity that the platform deduplicates against name, photo, and graph proximity. It is also the substrate of accumulated 1st-degree connections, recommendations, endorsements, posted content, group memberships, and the implicit graph-neighborhood signal that flows from years of normal use. None of that transfers. A restricted account cannot be recovered through a registrar transaction; appeal succeeds in under 20% of automation-flag cases by empirical observation, and a successful appeal does not always restore the messaging surface.

The architectural implication is unsubtle. The account is treated as a long-lived asset, not as a fungible piece of infrastructure. Operators who internalize the asymmetry build a multi-account pyramid in which the most valuable account never carries automation risk, and they accept the per-account warmup runway cost as the price of preserving the assets at the top of the pyramid.

The founder's personal account — do not run cold outbound from it

The single most damaging mistake observed across early-stage LinkedIn outbound deployments is running cold connect-request volume from the founder's personal account. The pattern repeats with enough regularity that it is worth naming: a growth lead, working under a deadline, points an automation tool at the founder's session because it is the highest-credibility account on the graph and produces the highest connect-acceptance in seed testing. Two weeks later, the account is in soft-ban escalation; three weeks later it carries a permanent restriction badge that every existing 1st-degree connection sees on visit.

The empirical restriction rate against accounts that receive a first-tier automation flag is approximately 60%, measured across multi-account agency operators with several years of platform-side telemetry. That is the conditional probability — given a flag, the chance of permanent restriction — not the unconditional probability of a flag, which is lower but compounds across outreach weeks.

The collateral damage on a founder's account is not symmetric to the damage on a throwaway. The founder's account typically carries seven to fifteen years of graph reputation, recommendations from named operators, board-level connections, and a content history that is itself a fundraising and recruiting asset. Replacing it is impossible under the same legal identity — any second account under the same name and photo is flagged within days, often hours.

The operator who deploys cold outbound from the founder's account is mortgaging the company's most durable distribution surface against the marginal connect-acceptance lift of a higher-credibility profile. The trade is bad on expectation. The founder's account runs warm conversation, content publishing, and selective hand-curated outreach. It does not run sequenced volume.

Profile completeness — the All-Star threshold and the 500-connection mark

Profile completeness is the most legible single signal LinkedIn exposes. The platform's earlier internal grading — Beginner through All-Star — is now mostly subsumed into the Social Selling Index (Chapter 4), but the mechanics remain stable enough to plan against. An All-Star profile broadly requires a profile photo, a substantive headline, industry and location, a current position with description, two prior positions, education, at least five skills, and 50 or more 1st-degree connections.

The 500-connection threshold is a distinct and more important signal. At 500+ connections, the platform displays the count as "500+" rather than the exact number — a social-proof signal to anyone evaluating the profile in a connect-request decision. Acceptance rates on outbound are observably higher for accounts above the threshold, with a typical lift in the 15-30% range — large enough to matter, small enough that operators routinely under-invest in clearing it before launching.

The detection-side implication matters more. Incomplete profiles — under 50 connections, no photo, no posted content, default headline — are flagged faster and at lower behavioral thresholds. The classifier treats profile depth as a prior on legitimacy, and an account issuing 80 connect requests per day from a profile with no posts and 47 connections is, from the classifier's perspective, indistinguishable from a synthetic identity warming up for spam. A complete profile is not a guarantee against detection; an incomplete profile is close to a guarantee of it.

The multi-account pyramid pattern

The architecture deployed by every agency operator with non-trivial LinkedIn experience, and the architecture that early-stage teams arrive at by trial and error, is a three-tier pyramid:

• One primary account at the top. Typically the founder, an executive, or a public-facing operator with years of graph reputation. This account carries the long-lived asset value. It runs no cold outbound, no sequencing tool, no third-party browser-extension automation. It is the source of credibility for warm intros, content distribution, and selective hand-written outreach to high-value targets.
• Three to five secondary accounts in the middle. The actual operating tier — SDR team members, growth hires, dedicated outreach personas under real legal identities. These accounts run the volume. They carry the warmup runway cost, the residential proxy cost (Chapter 3), and the per-week connection-request ceiling under the platform's published limit (Chapter 4). They are operated under the legal identity of the team member who hosts them, not under synthetic or fabricated identities.
• Throwaway burner accounts at the bottom. Optional tier used for high-risk pattern testing — automation-tool evaluation, edge-case behavioral probes, the kind of experimentation that would carry unacceptable risk on a secondary account. These accounts are expected to burn. The operator does not invest in their reputation, and the typical operational lifetime is measured in weeks rather than years.

The pyramid is not a marketing construct. It is the equilibrium architecture observed in operators who have been running LinkedIn outbound for long enough to have lost accounts and learned from it. The cost discipline is straightforward: the higher in the pyramid an account sits, the lower the automation risk it carries.

The legal-identity constraint

LinkedIn deduplicates by legal identity. The detection systems combine name-matching, photo-matching, declared employer, declared education, and graph proximity to identify duplicate or successor identities. An account restricted under the legal identity of a specific person cannot trivially be replaced under that same identity — a second account under the same name, photo, employer, or education history is flagged within days, frequently within hours.

The implication for team composition is direct. Each secondary account in the pyramid is hosted by a distinct team member, under that member's actual legal identity, with actual employment and education history. Synthetic personas — fabricated names with stock photos and invented employment — are detectable at high confidence within the first outreach week and additionally constitute a Terms of Service violation under §3.1 of the LinkedIn User Agreement, which requires accurate identity information.

LinkedIn outbound at any meaningful volume is bounded by the number of team members willing to host an outbound account under their own identity. A four-person growth team can support four operating accounts at the secondary tier; a two-person team can support two. The architecture does not scale by adding accounts under the same name; it scales by adding team members.

Profile depth — recommendations, endorsements, content, groups

Beyond the completeness checklist, the classifier appears to weight what can be summarized as profile depth — the volume and recency of structured social signal on the account. Recommendations from named 1st-degree connections, skill endorsements, a posting history with engagement, group memberships with participation, and a profile-view distribution consistent with normal use all contribute to a prior on legitimacy that the behavioral classifier integrates against.

Across multi-account operators, accounts with substantive content history and at least three named recommendations are flagged 40-60% less often than otherwise-identical accounts without those signals, holding outbound volume constant. The mechanism is not documented; the effect is consistent enough to plan against. The implication for warmup (Chapter 5) is that pre-outreach engagement runway is not just a behavioral pattern — it accrues structural signal on the profile that persists past the runway and lowers the steady-state detection probability during outreach.

The connection graph as reputation substrate

The 1st-degree connection graph is the substrate of every other reputation signal the platform produces. Connect-acceptance depends on graph proximity — a request to a target sharing 12 mutual connections is accepted at a meaningfully higher rate than one with zero. Search visibility depends on graph distance — 2nd- and 3rd-degree connections are surfaced; 4th-degree and beyond are largely invisible without LinkedIn Sales Navigator. Content distribution depends on graph engagement.

Accumulated 1st-degree connections from years of normal use are not replaceable. A new account starts at zero, builds to 50 through warmup, and reaches the 500+ display threshold somewhere in months two through four under disciplined operation. The graph at any point bounds the targeting surface — requests to targets with zero mutual connections are accepted at 6-12% in seed testing; requests to targets with 3-5 mutual connections are accepted at 25-40%. The operator with an established graph is in a fundamentally different distribution from the operator starting fresh.

Photo and headline consistency across the pyramid

Within the pyramid, the requirement is that no two accounts share the same person. Same photo across two accounts is detectable by photo-hash matching at very high confidence. Same name across two accounts at the same declared employer is detectable by name-matching at very high confidence. The detection systems are not subtle on this dimension.

The operational rule that follows: each account is operated by a distinct human, with that human's photo, name, declared employer, and education. The photo on each account is a unique image — a professional headshot specific to that person, not a stock image, not a reused image from another account. Operators who reuse the same headshot across two accounts under different names produce a high-confidence flag within the first warmup week and lose both accounts simultaneously.

The same-employer dimension has nuance. Two real team members at the same company can both run outbound accounts — that is the normal case. But the headlines should differ substantively, the connection graphs should diverge over time, and the outreach targeting should not produce the pattern of both accounts requesting the same target within 24 hours, which the classifier reads as a coordinated-account signal.

Email association and account recovery

The email address associated with the account is the recovery channel, the inbox where restriction notifications arrive, and the identity signal the classifier checks against the declared employer domain. The operational pattern:

• Primary account. Associated with the founder's corporate email — the long-lived address that already aligns with the declared employer and has years of in-platform history.
• Secondary accounts. Each associated with a dedicated email under the company's primary domain, not under an alias. Email-domain alignment with declared employer is read as a legitimacy signal; operators who associate secondary accounts with personal Gmail or with the cold-sending alt-domain produce a measurable drop in steady-state SSI score.
• Burner accounts. Dedicated personal email addresses, isolated from corporate identity. Expected lifetime is short enough that the recovery surface does not matter.

An account restricted under an email address the operator no longer controls is unrecoverable even in cases where appeal would have succeeded.

LinkedIn Sales Navigator — paid layer, same underlying identity

LinkedIn Sales Navigator is a paid product layered on top of an existing account, not a distinct account. The seat unlocks expanded search, lead lists, the InMail credit pool (Chapter 6), and the saved-search and alert infrastructure. The underlying identity, graph, and restriction risk remain on the base account. Operators who treat Sales Navigator as a separate logical account misread the architecture — losing the seat means losing $80 to $130 per month; losing the underlying account means losing the graph.

The cost is justified where the targeting surface requires the expanded filters — title, seniority, function, geography, company size, technology stack — that the base product does not expose. It is also where the InMail credit pool lives. The data-export surface, with respect to LinkedIn User Agreement §8 and §8.2, is constrained — the product does not expose a bulk export of search results, and any tooling that programmatically harvests lead lists is operating against the User Agreement's scraping prohibition. The post-remand reversal in hiQ Labs v. LinkedIn at the Ninth Circuit shifted the legal terrain on that question; full treatment lives in Chapter 08.

Account lifecycle — the 12-to-18-month operational runway

The typical operational lifetime of a secondary outbound account is 12 to 18 months before accumulated flags warrant retirement. The pattern is not that the account is restricted on month 13; it is that flag density rises, soft-ban frequency increases, connect-acceptance drifts down as platform-side reputation deteriorates, and at some point the marginal cost of operating the account exceeds the cost of standing up a fresh one.

The pyramid is therefore a rotating cast. A new secondary account enters every six to nine months under disciplined operation, completes its 2-4 week warmup runway (Chapter 5), and is brought into production while an older account is rotated out. The team-composition constraint binds — new accounts require new humans. An outbound motion that depends on continuous account turnover is constrained by hiring, not by infrastructure.

Common deployment failures observed in production

• Running cold outbound from the founder's personal account. The catastrophic failure mode. Discussed at length above. The marginal connect-acceptance lift is not worth the asset.
• Photo reuse across the multi-account pyramid.Same headshot on two accounts under different names. Detected within days. Both accounts flagged. The operator typically discovers this after a stock-photo "persona" account triggers a high-confidence match against an old archived photo of the operator themselves.
• Incomplete profiles entering outreach. Account at 47 connections, no profile photo, no posted content, default headline, begins issuing connect requests on day one. Detection threshold reached within the first week. The fix is the warmup runway; the failure is treating warmup as optional.
• Premature volume. A correctly architected account ramps connect-request volume from zero to the platform ceiling over 2-4 weeks. An account that goes from zero to 80 connect requests on day one is flagged regardless of profile quality. The classifier reads the velocity, not just the volume.
• Name and employer matching across accounts. Two accounts at the same company, with similar headlines, requesting the same targets within the same 24-hour window. The coordinated-account signal is high-confidence and produces simultaneous flags.
• Synthetic personas. Fabricated names, stock photos, invented employment. Detected with high confidence within the first outreach week. Additionally violates §3.1 of the User Agreement. The architecture does not require synthetic personas; the pyramid scales with real team members.
• Email association under a domain the operator does not control. Account associated with a personal Gmail or a one-off alias. When the restriction notification arrives, the appeal channel is in an unmonitored inbox. The account is functionally unrecoverable.

Pre-deployment checklist

• One primary account identified and explicitly excluded from cold outbound; rules of engagement documented in the team's SDR playbook
• Three to five secondary accounts identified, each hosted by a real team member under that member's legal identity
• Each account at or above the 500+ connection display threshold before outreach begins, or scheduled to reach it during warmup
• Each account at All-Star or equivalent completeness — photo, headline, two prior positions, education, five or more skills, three or more recommendations
• Each account associated with an email under the company's primary corporate domain, with recovery access verified
• Distinct headshots, distinct headlines, distinct connection graphs documented across the pyramid; no two accounts share a photo
• LinkedIn Sales Navigator seat provisioned per secondary account where the targeting surface requires it
• Account lifecycle plan in place — rotation cadence of one new secondary account every six to nine months, with team-composition planning to support it
• Warmup runway scheduled per account (Chapter 5) before any outbound volume is committed

Where account architecture fits in the broader infrastructure

The account is the substrate every other decision sits on top of. The detection model (Chapter 2) operates against the behavioral signal produced by the account; the residential IP infrastructure (Chapter 3) attaches to the account's session; the connection-request ceiling and SSI score (Chapter 4) are per-account; warmup (Chapter 5) builds the account's prior on legitimacy; messaging (Chapter 6) consumes the account's InMail credits and 1st-degree graph; automation category selection (Chapter 7) varies the detection risk loaded onto the account; and the compliance posture (Chapter 8) defines the legal terrain, including the User Agreement §8 scraping prohibition and the §8.2 automated-systems prohibition.

Every subsequent chapter assumes the architecture in this one. An operator who skips the pyramid and points an automation tool at the founder's account is operating outside the assumed substrate, and the rest of the reference does not apply to their case.