Key finding
Across 196,696 LinkedIn connection requests given at least 30 days to mature, acceptance by day of week ranged from 26.38% to 28.24%, a spread of 1.86 points, and weekend requests were accepted within 0.27 points of weekday requests.
What we measured
"What is the best day to send LinkedIn connection requests" may be the most-answered question in outreach, and nearly every answer is a chart with no stated sample, no denominator, and no source. This study answers it with a first-party denominator: every connection request sent through Reachium between 8 January 2025 and 27 July 2026, bucketed by the day it was sent, with acceptance measured after every request had at least 30 days to mature.
The answer is that there is nothing to optimize. The days are the same.
Methodology
Dataset window. All figures cover connection requests sent between 8 January 2025 and 27 July 2026, pulled on 27 August 2026. The cohort is restricted to requests sent at least 30 days before the pull, so late acceptances cannot flatter recent days. That leaves 196,696 requests, of which 53,050 were accepted (27.0%).
Denominators. Acceptance rate per day is accepted requests divided by requests sent on that day. Reply rate per day is replies divided by accepted requests from that day. Days are ISO days of the UTC send timestamp.
Robustness cut. The dataset is concentrated: the largest customer workspace accounts for 22.88% of matured requests. The findings table is repeated with that workspace excluded.
The findings
| Day sent | Requests | Accepted | Acceptance | Reply rate (of accepted) |
|---|---|---|---|---|
| Monday | 30,139 | 8,064 | 26.76% | 26.39% |
| Tuesday | 29,248 | 7,860 | 26.87% | 27.06% |
| Wednesday | 29,192 | 7,701 | 26.38% | 26.80% |
| Thursday | 29,278 | 8,267 | 28.24% | 28.38% |
| Friday | 27,367 | 7,379 | 26.96% | 27.02% |
| Saturday | 26,297 | 7,040 | 26.77% | 27.22% |
| Sunday | 25,175 | 6,739 | 26.77% | 27.39% |
The spread from the worst day (Wednesday, 26.38%) to the best (Thursday, 28.24%) is 1.86 points. Reply rates spread by 1.99 points. Weekdays as a group accepted at 27.04% across 145,224 requests; weekends accepted at 26.77% across 51,472. The weekend penalty everyone assumes is 0.27 points.
Excluding the largest workspace, the range moves to 27.30% (Monday) through 30.09% (Thursday), a spread of 2.79 points. Thursday edges highest in both cuts, and it would be easy to headline "send on Thursday". We are not doing that, for a reason the same dataset provides: in our study of month-to-month variance, the median swing in a single workspace's acceptance rate between two consecutive months is 5.69 points, three times the entire day-of-week spread. An effect smaller than the ambient noise is not an effect you can bank.
Why the hourly charts you have seen are noise
We also bucketed the same 196,696 requests by send hour. The table is unpublishable as advice, and it is worth showing why. The 09:00 UTC bucket accepted at 46.90% and the 12:00 UTC bucket at 14.21%, which would be a spectacular finding if it meant anything. It does not: 28.08% of all sends carry a 00:00 UTC timestamp because that is when sending windows open, and each hour's acceptance rate is dominated by which accounts happen to send at that hour. Hour-of-day charts measure who sends, not when works. Any timing chart that does not address this is composition dressed up as strategy.
What this means
The folk model says decision-makers clear their invites on weekday mornings, so timing your sends matters. The data says timing your sends is close to irrelevant, and there is a good mechanical reason: as our acceptance-timing study measured, only 46.7% of acceptances arrive within 24 hours of the send, and a quarter take longer than six days. A request sent on Saturday is mostly answered during the week anyway. The send day decides when the request enters the queue, not when the recipient looks at it.
What actually moved acceptance in this dataset is who you are and who you target: company size of the target spans 7 points, industry spans 11 points, and daily sending volume spans 9 points. Day of week spans less than 2. Optimizing send day is rearranging the smallest lever on the board.
In practice
An SDR who pauses campaigns on weekends is protecting themselves from a 0.27-point difference while giving up two sending days out of seven. In this data, the 51,472 weekend requests accepted at 26.77%, within noise of the weekday 27.04%. The data does not say weekends are better; it says they are not meaningfully worse.
A RevOps lead who sees Thursday at 28.24% and mandates Thursday sends is acting on a 1.9-point edge in a metric whose normal month-to-month movement within one workspace is 5.69 points. If the edge is real, it is too small to detect in any single team's numbers.
Limitations
Observational, one platform. All requests were sent through one platform's scheduling, which spreads sends across working windows. A person hand-sending at 23:00 on a Sunday is not represented here.
UTC days. Days are UTC, and recipients span time zones. A Friday evening send in California is Saturday in UTC. This blurs day boundaries, and the flatness could partly reflect that blurring. It cannot manufacture flatness across the whole week, but exact per-day values should be read loosely.
Composition across days is not controlled. Different accounts send on different schedules. The robustness cut excluding the largest workspace (22.88% of requests) shifts the whole range upward (27.30% to 30.09%) and widens the spread to 2.79 points, which is the scale of composition effects in this data.
The dataset is concentrated. The top five workspaces account for 59.70% of matured requests, the top ten for 78.50%. These are the sending patterns of committed outbound teams.
A null result has limits. This study supports "the day does not matter much", not "timing can never matter anywhere". A dataset with different geography or a different mix of senders could find more. It would still need to clear its own noise floor, and ours does not.