The published ceilings, the refusals we actually collected across 1,636 invitation attempts, and the pacing that took our refusal rate from 6.5% of attempts to 0.4% in eight days. Measured 30 August 2026.
| Limit | Figure | Where it comes from |
|---|---|---|
| Plain invitations per week, basic account | about 150 | Our LinkedIn provider's documented limits |
| Personalised invitations per month, basic account | about 5 | Same documentation, and it is the line most senders miss |
| Personalised invitations, premium account | a much larger allowance | Same documentation |
| Our own rolling ceiling | 100 invitations per 7 days per profile | We count it ourselves rather than waiting for the platform to say no |
| Cooldown after a refusal | 30 to 60 minutes on that profile | Our own rule, set 20 August 2026 |
| Hold on a person after a withdrawal | about 3 weeks | Provider documentation on the refusal code |
Read the provider documentation before you set a number. Our first version of the refusal rule was reasoned from first principles, sounded careful, and cost a sender a week of silence. Ten minutes in the documentation answered the same question with a figure.
Between 27 July and 28 August 2026 our profiles made 1,636 invitation attempts. The platform refused 74 of them. Split at the day we changed how we pace:
| Period | Invitations through | Refused | Share of attempts refused |
|---|---|---|---|
| Through 20 August 2026 | 1,030 | 72 | 6.5% |
| From 21 August 2026 | 532 | 2 | 0.4% |
Same profiles, same audience, roughly the same daily volume. Two changes explain the whole gap.
On a basic account the note attached to an invitation draws on an allowance of about five a month. Our own 21-day measurement across every sending profile, which set the rule on 20 August 2026:
| Account type | Invitation carried | Went through | Refused | Share refused |
|---|---|---|---|---|
| Basic | a note | 31 | 61 | 66% |
| Basic | no note | 627 | 2 | 0.3% |
| Premium | a note | 309 | 0 | 0% |
Two hundred to one. So a note now goes out only from a premium profile, and on a basic profile the personalisation moves to the first message after the connection is accepted, where nobody is counting it. The old budget counter had been counting successful sends, which meant a refused attempt left the counter untouched and the profile attached the note again the next day.
When LinkedIn declines an invitation, the answer is a temporary refusal with a cooldown attached. It is the cheapest warning available, and it is easy to throw away, because to whoever is watching the queue it looks like one invitation that went missing.
Our first rule stopped the profile for the rest of the day and paused it for a week after three refused days. Then we read the journal: across 23 days containing at least one refusal, invitations from the same profile went out successfully later the same day on 18 of them. The rule had been converting a short cooldown into a lost day, and a profile with a lost day looked identical, on our own counters, to a healthy profile with an empty queue.
The rule that runs now: a refusal pauses that profile for 30 to 60 minutes and the next attempt goes to a different person. A profile stops for the day only after three consecutive refusals with zero successful sends that day. An individual who refuses three times is closed out, because the provider documentation says the code covers an invitation to that person already pending or recently pending, and the platform holds them for about three weeks after a withdrawal.
Keep the body of the provider's error. A bare status code costs a manual investigation; the same error with its message answers the question in ten seconds.
A profile that fires its daily quota inside the first hour reads as automation to the platform and as a burst of notifications to the person receiving it. Our 1,562 sent invitations are spread across 207 distinct hours of the day, one cold touch per profile per minute at most, with the interval recomputed on every tick as the remaining day divided by the remaining plan.
Two more habits belong here. A new profile climbs from 1 invitation a day to 10 over ten working days before it carries a real batch. And stale pending invitations get cleared: we withdrew 2,501 of them between 13 and 29 August 2026, which keeps the pending queue honest and frees the people in it to be invited again once the platform's hold expires.
Provider documentation puts a basic account at roughly 150 plain invitations a week. We hold our own profiles to a rolling ceiling of 100 per seven days. Our busiest profile sent 80 in a calendar week and the fleet averaged 32.7.
Most often the personalised note. On a basic account the note draws on an allowance of about five a month, and in our 21-day measurement 66% of noted invitations from basic accounts came back refused, against 0.3% with no note.
Pause that profile for 30 to 60 minutes and send the next invitation to a different person. Stop the profile for the day only after three consecutive refusals with no successful send that day. Keep the error body; it names the reason.
From a premium profile they cost nothing: 309 noted invitations, zero refusals. From a basic profile they cost most of your invitations. Move the personalisation into the first message after the connection is accepted instead.
About three weeks, per the provider documentation on that refusal code. So withdrawing a stale invitation frees your pending queue immediately and frees the person after the hold expires. We withdrew 2,501 between 13 and 29 August 2026.
Climb from 1 invitation a day to 10 across ten working days, and hold the profile's first twenty touches out of every verdict. In our data those first twenty return 15.4% accepts against 21.0% from invitation twenty-one onward.
Book a 30-minute call. Tell me who you need to reach and I will say plainly whether this is a fit.