Original data · H2 2026 edition · a production-database study, not a survey
Twitter Cold DM Benchmarks
43,672,794
initial cold DMs on X/Twitter, analyzed in this edition.
Aggregated, anonymized benchmarks from 43.7 million real X/Twitter cold DMs sent through Xreacher over six months: what reply rates typical campaigns actually get, how hard mass multi-account blasting underperforms targeted solo outreach, when to send, which lead profiles reply, and how much daily volume accounts sustain. Recomputed each edition from the full campaign dataset with the methodology published below.
Data window: the 180 days ending July 3, 2026 · Published 2026-07-03 · Free to cite with attribution and a link
The dataset
The dataset
Every figure on this page is computed across all campaigns sent through Xreacher in the stated window, aggregated and anonymized. No individual customer data is shown.
85.3M
Xreacher has recorded 85.3M initial cold DMs on X/Twitter since launch (85,311,813 as of July 3, 2026).
43.7M
This edition of the Xreacher benchmarks covers 43.7M cold DMs (43,672,794 initial messages; follow-up messages are counted and analyzed separately) sent on X/Twitter in the 180 days ending July 3, 2026.
15,045
The data spans 15,045 distinct sending accounts across 1,124 campaigns in the window.
Finding 03 · Pre-DM behavior
Following a lead before DMing
The X-platform follow action (Campaign.followSettings), not a follow-up message — see the Follow-ups section below for that. Comparison is scoped to campaigns that had this setting enabled, to avoid comparing against campaigns that never used it at all.
Within the 378 follow-enabled campaigns only
~4x
Within campaigns that follow leads before DMing them, leads who were followed first replied at roughly ~4x the rate of leads who were only DMed — 1.42% versus 0.35%.
Methodology note: Correlational, not causal: leads followed before a DM and leads that were not may differ systematically within the same campaign (follows can fail or be skipped, e.g. protected accounts, daily follow limit reached). Scoped to the 378 campaigns with follow enabled; sample of 5,506 followed threads out of 200,000.
Finding 04 · Follow-ups
Follow-ups
How often follow-ups are used, and what they actually do. Low usage is partly rational economics rather than a verdict on follow-ups: a follow-up consumes the same scarce daily DM allowance as a fresh prospect, so senders tend to spend the quota on new leads instead of second touches. We still refuse to publish a naive follow-up 'lift': threads with and without follow-ups are not comparable, because follow-ups only go to leads who have not replied yet, and the naive full-pool comparison (0.94% vs 0.43%) reverses inside cohorts (0.92% vs 1.35%) — a Simpson's paradox we would rather disclose than publish. Instead, the rescue-rate figure below measures follow-ups the honest way, by reply timing: did a silent lead reply after the follow-up went out, at a rate above same-age silent leads who never got one?
Reply-timing measurement · 186K follow-up threads, non-bulk cohort
99.3%
99.3% of all cold DM threads consist of a single message — and even among smaller, non-bulk senders, 96.5% of threads never receive a follow-up.
Methodology note: Follow-ups count against the same daily DM limit as new outreach, so preferring fresh prospects over second touches can be the economically rational choice — this is a usage stat, not evidence that follow-ups underperform.
1.5-2.3x
Leads still silent 3+ days after a cold DM who then received a follow-up replied at 0.88-0.96% afterwards — 1.5-2.3x the 0.42-0.58% rate of same-age silent threads that never got a follow-up — while follow-ups sent within the first 3 days showed no benefit over simply waiting (0.44% vs a 0.58-0.79% baseline).
Methodology note: Correlational reply-timing analysis on the non-bulk cohort: 185,903 follow-up threads, each reply placed before or after its first follow-up via message timestamps, with a 21-day observation buffer on both arms so late replies count equally. Which silent threads get a follow-up is not randomized, and the no-follow-up baseline includes campaigns that never use follow-ups. Early-follow-up cell (sent at 1-3 days): 12,890 threads.
Finding 05 · Send timing
Send timing
How send time relates to reply rate, computed excluding bulk adult-industry operations so the pattern is not dominated by one audience type. Hours are UTC.
Reply rate by send hour (UTC) · non-bulk cohort, 5.4M threads
19:00-02:00 UTC
DMs sent between 19:00-02:00 UTC — US afternoon and evening — replied at roughly 1.4%, about 60% above the 09:00-12:00 UTC trough (0.86%).
Methodology note: Correlation across a global sender base, not a causal test; sender and audience timezones vary. Computed on the non-bulk cohort (5.4M threads).
Reply rate by send hour — full table
| UTC | New York (ET) | Berlin (CET) | DMs sent | Reply rate |
|---|---|---|---|---|
| 00:00 | 20:00 | 02:00 | 226,646 | 1.44% |
| 01:00 | 21:00 | 03:00 | 221,327 | 1.45% |
| 02:00 | 22:00 | 04:00 | 200,056 | 1.55% |
| 03:00 | 23:00 | 05:00 | 110,792 | 1.45% |
| 04:00 | 00:00 | 06:00 | 85,173 | 1.62% |
| 05:00 | 01:00 | 07:00 | 79,018 | 1.66% |
| 06:00 | 02:00 | 08:00 | 64,596 | 1.52% |
| 07:00 | 03:00 | 09:00 | 44,163 | 1.50% |
| 08:00 | 04:00 | 10:00 | 70,326 | 1.69%highest |
| 09:00 | 05:00 | 11:00 | 77,362 | 0.91% |
| 10:00 | 06:00 | 12:00 | 97,094 | 0.86% |
| 11:00 | 07:00 | 13:00 | 135,499 | 0.80%lowest |
| 12:00 | 08:00 | 14:00 | 165,956 | 0.88% |
| 13:00 | 09:00 | 15:00 | 321,587 | 1.30% |
| 14:00 | 10:00 | 16:00 | 327,876 | 1.13% |
| 15:00 | 11:00 | 17:00 | 349,539 | 1.31% |
| 16:00 | 12:00 | 18:00 | 369,705 | 1.28% |
| 17:00 | 13:00 | 19:00 | 381,083 | 1.41% |
| 18:00 | 14:00 | 20:00 | 370,446 | 1.42% |
| 19:00 | 15:00 | 21:00 | 401,471 | 1.52% |
| 20:00 | 16:00 | 22:00 | 376,953 | 1.28% |
| 21:00 | 17:00 | 23:00 | 354,557 | 1.32% |
| 22:00 | 18:00 | 00:00 | 310,189 | 1.42% |
| 23:00 | 19:00 | 01:00 | 246,577 | 1.37% |
Non-bulk cohort (5.4M threads), 180 days ending July 3, 2026. Early-UTC-morning hours (04:00-08:00) also show high rates but on much smaller volume. Local-time columns use current offsets (New York UTC-4, Berlin UTC+2); the window spans daylight-saving changes, which blurs local hours by about one hour.
Finding 06 · Targeting
Who replies
Reply rate by the contacted lead's profile, from a large random sample of threads joined against scraped lead attributes (non-bulk cohort).
X Premium vs non-Premium leads — consistent across follower bands
1.8x
Leads with X Premium (verified) replied at about 2.3% versus 1.25% for non-Premium leads — a 1.8x difference that held across every follower band.
Methodology note: Weighted across follower bands in the non-bulk cohort sample (~200,000 joined threads). 'Verified' = active X Premium subscription.
Finding 07 · Historical volume
Daily volume, then and now
How much daily volume accounts actually sent during the window analyzed. Important: X has since tightened DM limits substantially (roughly 35 DMs/day for X Premium accounts as of mid-2026), so treat these historical volumes as context about the era the data comes from — not as pacing guidance for today.
50-99
During the window analyzed, the median active sending account sent 50-99 DMs per day; smaller non-bulk senders were closer to 25-49.
Methodology note: Median of ~290,000 account-days, January-July 2026. X's DM caps have since been reduced (~35/day for Premium accounts as of mid-2026); volumes in this dataset predate those limits and should not be read as recommended pacing.
Summary
Practices associated with higher reply rates
Each guideline restates a finding above, with its qualifier. Correlational findings describe associations in observational data, not causal effects.
Choose an account strategy deliberately. Running 1-3 accounts produced the highest reply rate per DM (2.24-2.51%); running 11+ accounts produced the most total reach at a lower per-DM rate (0.39%). Both patterns appear throughout the dataset.
2.51% vs 0.39%Prioritize X Premium leads with 10,000-100,000 followers — the strongest profile measured. Premium leads replied at roughly 1.8x the non-Premium rate in every follower band.
3.19%The strongest correlation in the dataset favors follow-before-DM: within follow-enabled campaigns, followed leads replied at roughly 4x the rate of leads who were only DMed. Correlational, not a controlled test — but it is a one-setting change worth testing in your own campaigns.
1.42% vs 0.35%Sends in the 19:00-02:00 UTC window showed the highest reply rates and 09:00-12:00 UTC the lowest — a correlation across a global sender base, not a controlled test. Worth preferring the evening window where scheduling allows.
+60%Keep threads under observation for at least a week: 22.6% of replies arrived more than 7 days after the initial DM.
22.6%If following up, wait at least 3 days. Follow-ups sent within 3 days showed no measured benefit over waiting; follow-ups after 3+ days of silence were associated with 1.5-2.3x the spontaneous late-reply rate. Correlational.
1.5-2.3xBenchmark against the campaign-level median (0.57%), not the pooled average (0.43%). Top-decile campaigns exceeded 3.7%.
>3.7%Methodology
How these figures are computed
Source
Full production dataset of Xreacher, an X/Twitter outreach platform: 43,672,794 initial cold DMs, 15,045 sending accounts, 1,124 campaigns, 180 days ending July 3, 2026. Aggregated and anonymized; no individual customer, account, or campaign is identifiable.
What counts as a reply
Any response from the contacted lead after the initial DM. Attribution is last-touch — replies on threads with follow-ups credit the final stage. AI-agent conversation after the first reply is not counted again.
Sampling
Headline volume stats use the full dataset. Stats joining lead attributes (targeting, follow-before-DM) use large random samples (~200,000 threads); cell sizes are disclosed inline with each stat.
Timing precision
Reply detection is polled by the sending infrastructure, so time-to-reply has polling-cadence granularity. We publish buckets, not minute-level medians, and buckets are slightly conservative.
Cohorts
“Non-bulk” excludes campaigns of users who selected the OnlyFans/OFM business type at onboarding (self-reported; older accounts without onboarding data remain in the non-bulk cohort). Account-count cohorts reflect accounts actually used in the window, not subscription plan.
What is not published
Cross-cohort ratios computed from pooled data (the naive follow-up comparison reverses inside cohorts — a Simpson's paradox disclosed in the follow-ups section), and any stat without its caveat attached. Correlational findings are labeled correlational. Every figure is recomputed from scratch each edition.
FAQs
Where does this data come from?
From aggregated, anonymized campaign data across all X/Twitter outreach sent through Xreacher in the stated window — 43.7 million DMs from 15,045 sending accounts. Figures are computed with database-level aggregations over the full dataset (large random samples for stats that join lead attributes), and no individual customer, account, or campaign is identifiable.
Why report a median campaign rate instead of one overall average?
Because a pooled average is dominated by a handful of industrial-scale senders: operations running 11+ accounts produced 97.5% of raw volume at a 0.39% reply rate, while typical small senders earned 2.2-2.5%. A single pooled number (0.43%) describes the volume extremes, not a typical campaign, so we publish the campaign-level median and the cohort split instead — and show all three numbers.
Can I cite these benchmarks?
Yes. Every stat has an anchor link you can reference directly, and you may quote any figure with attribution to 'Xreacher Cold DM Benchmarks' and a link to this page. The edition label states the data window so citations stay accurate as the page is refreshed.
What counts as a reply?
A thread counts as replied when the contacted lead sends any response after the initial DM. Attribution is last-touch: if follow-ups were sent, the reply is credited to the thread's final stage. Automated or AI-agent conversations after the first reply are not counted as additional replies.
How often is this updated?
Each edition is recomputed from the full dataset and the edition label and publication date are updated accordingly. Previous editions' figures are not silently changed; a refresh replaces the edition label so any external citation remains traceable to its window.
Last updated 2026-07-03 · Cite as “Xreacher Cold DM Benchmarks (H2 2026)” with a link to this page; every stat above has its own anchor.