Connection limits and SSI — the rate-limit topology most operators misread.

The 100-per-week connection-request cap

LinkedIn announced the current cap on outbound connection invitations in 2021, replacing what had previously been a soft ceiling north of 200 per week. The official communication, surfaced through the LinkedIn Help Center and through in-product warning modals once an account approaches the threshold, is that members are limited to roughly 100 invitations per week, with the platform reserving the right to apply additional restrictions based on prior acceptance patterns.

Three details about that number are operationally load-bearing and almost never read correctly. First, the measurement window is a sliding 7-day window, not a calendar week. An account that sends 100 invitations on Monday cannot send any further invitations until the following Monday — the counter does not reset on Sunday at midnight. Operators planning around a calendar-week cadence routinely under-utilize the ceiling for the first three days of every week and then hit the cap by Thursday.

Second, the cap is applied per-account, independently. A three-account architecture has a 300-per-week aggregate ceiling, not because the platform aggregates across accounts but because each account is independently bounded. This is the entire premise of the multi-account pyramid pattern (Chapter 1).

Third, the 100 number is a ceiling, not a target. An account whose acceptance rate is below the empirical normal range will hit a behavioral throttle well before reaching 100, and the throttle will not arrive as a clear notification — it will arrive as a degraded acceptance rate, then as a search-result limitation, then as a connect-blocked state. Operating at the ceiling without monitoring the leading indicators is the single most common cause of a first-tier soft-block.

The 200-per-week historical cap

Prior to 2021, the effective cap was approximately 200 invitations per week, applied loosely and with frequent informal allowances for accounts with high acceptance rates. A meaningful population of accounts that existed continuously through the policy change still observes an effective ceiling somewhat above 100 — typically in the 120 to 140 range — before the soft-block heuristic engages.

This is not a documented allowance, it is not portable to new accounts, and it is not reliable. The empirical observation is that long-tenured accounts with multi-year clean behavioral histories occasionally tolerate volumes the published policy disallows. Treating this as a feature is the wrong posture. The 100-per-week cap should be the operational ceiling regardless of what an individual account appears to tolerate, because the day the legacy allowance is withdrawn is not announced in advance.

Account tier and rate-limit topology

LinkedIn's pricing tiers are not, strictly, pricing tiers. They are rate-limit packages with messaging features attached. The relevant tiers for outbound operation, in ascending order:

The 100-per-week connection-request cap is constant across all tiers. Premium does not buy more invitations; Sales Navigator does not buy more invitations; Recruiter does not buy more invitations. The InMail allocation, the search ceiling, and the messaging feature surface differ — the connection-request cap does not.

This is the detail operators paying $99 per month for Sales Navigator are most surprised by. The upgrade purchases search visibility and InMail credits. It does not purchase a higher cold-connect throughput.

The Social Selling Index

The SSI is LinkedIn's published behavioral-quality score, accessible at linkedin.com/sales/ssi, and it is the score most operators read as a vanity metric. It is not a vanity metric. It is one of the inputs to the rate-limit enforcement layer.

The score is a 0-100 composite across four dimensions, each weighted 0-25:

• Establish your brand — profile completeness, content publishing, recommendations, endorsements
• Find the right people — search usage, saved leads, advanced filter engagement
• Engage with insights — feed interaction, content reactions, comment activity
• Build relationships — connection growth, message exchanges, senior-title connection density

The empirical observation across multi-account agency operations is that high-SSI accounts — typically the 70-and-above range, particularly with high marks on "Establish your brand" and "Build relationships" — tolerate slightly higher action volume before triggering soft-blocks than equivalent low-SSI accounts. A 25-point SSI account running at 80 connects per week will hit restrictions sooner than an 80-point SSI account running at the same volume. The differential is not enormous and it is not officially documented, but it is consistently observable across hundreds of account-weeks of behavioral data.

The operational implication: an account being warmed up (Chapter 5) should target an SSI in the 60-75 range before cold-connect volume ramps. This is achievable through profile completion, three to five posted content artifacts, ten to twenty inbound engagement actions per day, and steady connection growth. It is not achievable through automated engagement, which produces a recognizable signal pattern (Chapter 7).

The withdrawal-acceptance ratio

Every outbound connection request sits in one of three terminal states: accepted, ignored, or actively declined. The fraction of pending invitations relative to accepted invitations is, in our observation, the single most predictive leading indicator of soft-block escalation, and it is documented in LinkedIn's help center as a factor the platform considers when applying invitation restrictions.

The empirically observed threshold is approximately 30%. When pending invitations exceed 30% of the rolling outbound volume — meaning the account is sending faster than acceptances are coming back, with stale invitations accumulating in the Sent folder — the soft-block heuristic engages within a few days. The signal LinkedIn reads is straightforward: the account is approaching strangers who do not recognize the inviter, the recipients are ignoring the requests, and the account is continuing to send anyway.

The operational hygiene is to withdraw stale pending invitations on a rolling basis — typically anything pending for more than two to three weeks. The withdraw action is in the My Network > Manage invitations > Sent surface. Withdrawing 30 to 50 stale invitations per week is normal and does not produce a negative signal. Withdrawing 200 in a single session does.

Connection-acceptance rate

The acceptance rate — accepted invitations divided by total invitations sent — is the parallel metric the platform uses to evaluate whether outbound is producing recognizable, wanted contact or pattern-matched spam. The empirical thresholds observed across multi-account operations:

• Under 20% acceptance — the targeting is wrong, or the profile is producing no recognition signal. This is a flag and produces throttling within 2-3 weeks.
• 20-40% acceptance — normal cold range. Targeting is functional, profile is presentable, copy is acceptable.
• 40-60% acceptance — warm cold range. Either the targeting is excellent, the brand has independent recognition, or the connection note is doing real work.
• Above 60% acceptance — almost exclusively achievable on 2nd-degree-and-warmer audiences, on accounts with significant external brand recognition, or on rosters with strong company-affinity overlap (alumni, ex-employees, customers).

An operator targeting under 20% acceptance is operating an account that the platform is, correctly, classifying as low-quality outbound. The fix is upstream of volume — targeting, profile, copy — not downstream.

Search rate limits

The commercial use limit is the soft-block on free-account search that surfaces as the message "You've reached the commercial use limit on search." The threshold is opaque and adaptive — the platform does not publish a specific number — but the empirical pattern is that free accounts performing on the order of 30-50 unique searches per month against profiles outside the immediate network will hit the limit by mid-month, with reset on the first of the following calendar month.

Sales Navigator lifts this limit. It does not, however, lift it to infinity. Even on Sales Navigator, sustained high-volume search activity — typically several thousand profile views per day, or saved-search runs that pull thousands of leads per hour — produces a separate throttle that surfaces as degraded search responsiveness and eventually as a temporary search block. The published guidance from LinkedIn does not specify the ceiling. The empirical operational ceiling is in the low thousands of unique profile-loads per day per account.

Profile-view rate limits

Profile views are tracked per-day, are visible to the viewed profile by default (a privacy setting the operator can adjust), and have an undocumented soft ceiling that engages in the high hundreds to low thousands per day per account.

The visibility detail is operationally significant in two directions. Outbound: a recipient who sees the inviter's profile show up in their "Who viewed your profile" panel is more likely to accept a subsequent invitation, because the request appears warmer. Inbound: a recipient who sees a sequence of fifty viewers from the same account over a single day will recognize the pattern and may report the profile. The healthy operating pattern is distributed views across business hours, not bursts.

High-volume profile-view sessions — several hundred profiles viewed within an hour — produce restriction risk independently of any other behavior. Operators using profile views as a warmup signal should distribute the volume across the working day and cap daily volume well below the empirical ceiling.

InMail credit economics

InMail credits are the paid-tier currency for messaging outside the connection graph. The published allocations:

• Premium: 5 InMail credits per month
• Sales Navigator Core: 50 credits per month
• Sales Navigator Advanced: 50 credits per month
• Recruiter Lite: 30 credits per month
• Recruiter: 150 credits per month

The mechanic worth understanding is the response-rate refund. An InMail that receives any response — including a decline — within 90 days returns the credit to the account. An InMail that is ignored consumes the credit permanently. The economics consequently favor copy that produces a response, including a polite "not interested" response, over copy that produces silence.

The operational math: at 50 credits per month, with a typical 15-25% response rate on competently written InMails, the effective monthly outbound is approximately 60-70 sends, of which 10-15 produce some response and 1-3 produce a booked meeting. The cost-per-meeting on Sales Navigator at $99/month is in the $35-100 range, depending on response rate and meeting-conversion rate. Exhausting InMail credits in week one of the month — which most new operators do — produces a three-week dry period that the rest of the rate-limit topology cannot make up for.

The 300-character connection-note limit

The connection-request note was historically 300 characters on free accounts and 1000 characters on paid tiers. In late 2022 the platform restricted the 1000-character allowance to a subset of paid users on a usage-based rollback, and the practical posture for most operators today is to assume a 300-character ceiling regardless of tier.

The operational implication for cold outreach copy is substantial. 300 characters is approximately 50 words. It is enough for a salutation, a single specific observation about why the inviter is reaching out, and a request — and almost nothing else. The decision tree on the connect-note is binary: either write a note that produces a meaningfully higher acceptance rate than no note, or omit the note entirely.

The empirical observation is that connection requests with no note achieve approximately the same acceptance rate as connection requests with generic notes ("Hi {name}, would love to connect!") — both in the 20-30% range on cold cohorts. Connection requests with specific, recipient-relevant notes — referencing the recipient's role, recent post, shared connection, or shared affiliation — achieve acceptance rates 1.5 to 2x higher. Generic notes are a tax on the character budget and produce no acceptance lift.

The soft-block escalation ladder

Restrictions on LinkedIn are not binary. They escalate through a graduated ladder of increasingly invasive throttles, each with its own time-to-restoration and each with its own threshold for triggering the next tier:

The durations above are first-occurrence. Repeats extend roughly geometrically — a second connect-block lasts a week, a third can last a month, a fourth typically escalates to a tier-4 lock. The ladder is stateful across tiers: the platform tracks restriction history and applies increasing severity to repeats.

Sales Navigator data export rate limits

Sales Navigator's saved-lead and lead-list features have their own rate-limit envelope, distinct from the search and connect ceilings. Saved leads are capped at 10,000 per account. Lead lists are capped at 25 lists per account, with 1,000 leads per list on Core and 2,500 on Advanced. List exports — to the extent they are available, which is partial and varies by tier — are limited to dozens of rows at a time through the UI surface.

The operational implication for sequencing is that the segmentation of audiences into addressable cohorts must happen inside Sales Navigator's surface, not through bulk export to an external system. Operators attempting to extract the full lead list to an external sequencer in a single session will hit both the export ceiling and the automation-detection layer (Chapter 7).

Common operator failures observed in production

• Running at the 100/week cap without monitoring withdrawal rate. The operator hits the cap consistently, the pending pile accumulates, the 30% threshold engages, and the connect-block surfaces three weeks later as a complete surprise.
• Treating SSI as a vanity metric. The score is published, the operator dismisses it as a sales gimmick, and the account operates at low SSI with no investment in profile or content. The rate-limit elasticity is consequently lower, and the soft-block engages earlier than the published cap would suggest.
• Exhausting InMail credits in week one. The new Sales Navigator subscriber sees 50 credits and spends 40 in the first four days against a generic cold list, recovers a 5% response rate, and operates the remaining three weeks of the month with no credits and no playbook.
• Compressing outbound into a single daily window. The operator sends all 20 daily invitations between 9:00 and 9:30 AM. The platform reads this as automated. The behavioral throttle engages without any single threshold being crossed.
• Operating a Premium account expecting higher connect throughput. The 100-per-week cap is identical on Premium. The $30/month is purchasing search and InMail, not invitation volume.
• Ignoring the pending-withdrawal hygiene. The Sent invitation list is allowed to accumulate indefinitely. Six months in, the operator has a thousand-pending stack that the platform reads as evidence of low-quality outbound and that takes weeks to drain safely.

Pre-deployment checklist

• Account SSI measured and recorded as a baseline before cold outbound begins; target 60+ before ramping volume
• Withdrawal hygiene procedure documented and scheduled — typically weekly, withdrawing invitations pending more than 2-3 weeks
• Pending-to-accepted ratio monitored on a rolling weekly basis, with a 30% soft ceiling and an automatic volume reduction below that threshold
• Connection-acceptance rate tracked per cohort, with the 20% floor as the trigger to revise targeting or copy before pushing more volume
• InMail credit budget distributed across the calendar month, not front-loaded into week one
• Daily invitation volume distributed across business hours, not compressed into a single window
• Search and profile-view volume capped well below the empirical platform ceiling, regardless of tier
• Connect-note copy A/B tested against no-note baseline before any assumption that the note is producing acceptance lift
• A documented response playbook for tier-1 and tier-2 soft-blocks — including the requirement to pause outbound entirely during the restriction window, not work around it

Where this fits in the broader infrastructure

The rate-limit topology is the operational ceiling that account architecture (Chapter 1), the detection model (Chapter 2), and the IP infrastructure (Chapter 3) all sit underneath. An account architecture with three accounts has a 300-per-week ceiling; an account architecture with seven accounts has a 700-per-week ceiling. The detection model determines how close to the ceiling the account can safely operate. The IP infrastructure determines whether the behavioral signal the platform reads matches the published policy or triggers the parallel anomaly classifier.

Warmup (Chapter 5) is the deliberate construction of the SSI score, the connection density, and the engagement history that buys the operational headroom within the rate-limit envelope. Messaging architecture (Chapter 6) is the conversion layer that turns the rate-limited outbound into pipeline. Automation (Chapter 7) is the category of tooling that, depending on selection, either operates inside the published rate limits or produces the behavioral signal that triggers the lower tiers of the soft-block ladder.

The 100-per-week number is the headline. The topology underneath it is the operational reality. An account run with attention only to the headline will be restricted within a quarter. An account run with attention to the topology will run indefinitely.