Cold DMs that actually get a reply — what the research says

You wrote one. You rewrote it three times. You hit send. Three days later you check and the message sits there, read, no reply. So you look at what the big accounts say to do. Use their name. Mention something from their profile. Keep it short. You do all of that. The next five messages get the same silence.

So you go looking for what actually works. Not tips. Evidence. Studies with sample sizes and methods you can check. They exist. Some of them tested millions of messages. What they found does not always match what gets repeated.

Here is what we read. We went looking for research on what makes someone reply to a message from a stranger. We read peer-reviewed papers and industry studies with disclosed methods. Every one is linked, dated and quoted below. One section tells you what we could not prove.

The baseline is low and that makes the details matter

Start with the floor. How often does a cold message get any reply at all?

Backlinko analyzed 12 million outreach emails in partnership with Pitchbox. They published the findings on 16 April 2019. The overall response rate was 8.5%. That means 91.5% got ignored or deleted.

Another study looked at 12,000 campaigns across 2,000 accounts. BuzzStream coded over 30,000 individual emails to measure what moved the number. The median reply rate was 12%. The bottom quarter got under 7%. The top quarter cleared 20%.

Those are emails. Instagram DMs may behave differently. No published study we could find gives a baseline reply rate for cold DMs on Instagram specifically. Meta reports plenty about the app, but the numbers it puts on the record stop well short of the inbox. So treat the email numbers as a reference, not a prediction. The constraint is the same. Most messages from strangers get no reply. That makes every choice count.

Shorter messages get more replies, to a point

Three studies looked at message length. All three point the same way.

BuzzStream ran a separate analysis on millions of emails across hundreds of campaigns. Messages under 300 words got reply rates that were more than double those of longer emails. Open rates stayed stable until crossing 1,000 words. Reply rates dropped hard past 300.

The Backlinko study broke it down further. Subject lines of 36 to 50 characters outperformed short ones by 32.7%. That is long enough to describe what the message is about. But the body still needs to stay short.

The strongest evidence comes from a 2019 paper presented at CSCW. Ashton Anderson and Kristen Altenburger published Causal Effects of Brevity on Style and Success in Social Media. They used crowdsourcing to shorten real social media messages by varying amounts. Then they measured success. Concise messages outperformed the originals when shortened by up to 30 to 40 percent. The optimal reduction sat between 10 and 20 percent.

That last one matters because it isolated brevity as a cause. Shorter messages were not just correlated with success. Shortening a message made it more successful. Up to a point. Cut more than 40 percent and the benefit disappeared.

Put a number on it. A message of 100 words should land around 80 to 90 words. A message of 200 words should land around 160 to 180. Shorter than that and you probably help. Much shorter and you may hurt.

Fake personalization is worse than none

Personalization is not one thing. It splits into two kinds, and only one of them works.

BuzzStream's campaign study split emails into three groups. Template-only emails with no personalization got a 16% reply rate. That is the baseline. Emails with fake personalization did worse. A generic reference like "Great post!" dropped performance by 1.3 percentage points. Below the baseline.

Real personalization did the opposite. Contextual relevance gave a 10.7 percentage point boost. That means a specific reference tied to the reason you are reaching out. Something that shows you read their work and know why you are contacting them. Add a reference to how they promoted similar things in the past and the boost climbs another 2.3 points.

Backlinko found the same split. Personalized subject lines raised response rates by 30.5%. Personalized email bodies raised them by 32.7%. Starting with a generic "Hi," signaled a mass message. People could tell.

The lesson is not "personalize more." It is "personalize correctly or do not personalize at all." A message that pretends to know you and clearly does not reads worse than a straightforward template. A message that references something specific and ties it to the ask reads like you did the work.

Similarity makes people listen

Why does a shared connection help? The answer is not obvious. You might guess that people trust their friends' friends. Or that a mutual connection signals you are not a spammer. Both sound right. Neither is the whole story.

Ali Faraji-Rad, Bendik M. Samuelsen and Luk Warlop published a paper in the Journal of Consumer Research in 2015. The title is On the Persuasiveness of Similar Others. They ran four studies. Their finding cuts deeper than trust. Similarity creates a feeling of certainty. When someone seems like you, you feel like you understand how they think. That feeling validates what they say. Even when the similarity is shallow.

The effect went away when people were already certain. It also went away when they were primed to feel certain for a different reason. So the mechanism is the feeling itself. Not the fact. Similarity makes advice feel right.

A second study tested a related idea. Gregory Maio and five co-authors published Social values as arguments in Frontiers in Psychology in 2014. They ran four experiments with sample sizes of 60, 50, 100 and 82. Participants read arguments that linked two values. Arguments linking similar values were more persuasive than arguments linking opposing ones. The effect held across all four studies.

Apply that to a message. Mentioning a shared interest is not just rapport. It is a persuasion device. Saying "I also work with small e-commerce brands" does more than show you have something in common. It makes your message feel more certain. More right. Whether or not it actually is.

People feel obligated to return a favor, even a tiny one

Reciprocity is older than marketing. Dennis Regan ran the defining experiment in 1971. He published it as Effects of a Favor and Liking on Compliance. Participants thought they were rating art. Another person in the room, Joe, was actually Regan's assistant. During a break, Joe either bought the participant a Coke or returned empty-handed. Later, Joe asked participants to buy raffle tickets. The ones who got a Coke bought about twice as many tickets.

The size of the favor did not match the size of the return. A Coke is cheap. Raffle tickets add up. And the effect held even when participants said they did not like Joe. The favor created obligation. Not liking. Obligation.

That mechanism shows up in outreach advice all the time. Lead with value. Give before you ask. The advice is correct. The research backs it. But the research also shows a limit. The favor has to come first. And it has to feel like a favor. Not a pitch with an ask attached.

A message that says "I loved your post on X" and immediately asks for something is not leading with value. It is flattery. A message that links to something useful with no ask attached is a favor. You can follow up later. Regan's subjects did not buy tickets because Joe was great. They bought them because Joe gave them something first and they felt they owed him.

None of this was tested on Instagram DMs

No published study we could find tested these principles on Instagram DMs specifically. The brevity research ran on social media posts and emails. The personalization studies ran on email campaigns. The similarity research ran in lab settings and surveys. The reciprocity finding is 50 years old and it tested raffle tickets.

So nobody can honestly put a number on it. Not a percentage lift. Not a target reply rate. There is no experiment proving what these principles do inside an Instagram inbox. Not one we ran. Not one from anyone selling templates. Not one from anyone running a course on cold outreach.

The principles are real. The effects are documented. The jump from email and lab studies to Instagram DMs is ours. It is a reading of related evidence, not a finding. If someone quotes you a number for how much a tactic lifts Instagram DM replies, ask them for the study. Then send us the study too, because we could not find it.

Templates built from the findings

Everything above describes a mechanism. Not a script. But scripts help. So here are four templates. They are not magic. They are illustrations of the principles. Use them as a start, not a formula.

Template 1: Brevity + specific relevance

Hey [name], saw your reel on [specific topic]. I work with [type of business] and we solved [specific problem they mentioned]. Happy to share what worked if useful.

Under 30 words. References something specific. Ties the reference to the offer. No fluff.

Template 2: Shared connection + clear ask

[Mutual connection] mentioned you. I am working on [specific thing relevant to them]. Would 10 minutes this week work to ask you two questions about [their area]?

Leads with the connection. States what you want. Puts a time limit on it. Makes the ask easy to evaluate.

Template 3: Reciprocity + no immediate ask

Hey [name], I put together a list of [specific resource relevant to them]. Thought it might help with [problem they have mentioned]. No ask, just sharing: [link]

Value first. No request attached. If they reply, you have opened the door. If they do not, you follow up later. That follow-up has reciprocity behind it.

Template 4: Similarity + context

I also run a [type of business similar to theirs]. Saw you dealt with [specific challenge]. We just worked through the same thing. I wrote up what worked. Want me to send it?

States the similarity in the first sentence. Ties it to something they experienced. Offers something concrete. Lets them choose.

All four stay under 100 words. All four reference something specific. None of them pretend to be more personal than they are. That last one matters. The BuzzStream data showed fake personalization hurt. A template that works is better than a personal message that does not.

Where the evidence thins out

Five things we looked for and could not pin down.

The Instagram baseline. We found overall reply rates for email outreach. We did not find a published reply rate for cold Instagram DMs. So we do not know if 8.5% or 12% applies, or if Instagram behaves differently. No primary source we could access reported it.

Effect sizes for DMs specifically. The studies above give numbers. Personalized emails get 32.7% higher response. Messages under 300 words get double the replies. Those are email findings. We do not know if Instagram DMs move the same amounts. No study we read tested them.

How long similarity lasts. The Faraji-Rad paper shows similarity drives persuasion in the moment. It does not say how long that lasts. If someone replies three days later, does the similarity effect still hold? We do not know.

Whether follow-ups work the same way. Backlinko found that one follow-up boosted email replies by 65.8%. Does that transfer to Instagram? Instagram DMs are more interruptive than email. A follow-up might help less or hurt more. We found no data either way.

Optimal message length for DMs. The brevity research gives percentages. Shorten by 10 to 20 percent. But it does not give an absolute floor. Is 10 words too short for a DM? Is 50 too long? The studies do not settle it.

How we did this

We searched for peer-reviewed research and disclosed industry studies on message response rates, brevity, personalization, similarity and reciprocity. We read the sources directly on 18 August 2026. Every quotation was copied from the page as a reader sees it. Two of the sources are vendor studies, BuzzStream and Backlinko, and both disclose their sample sizes and methods. Three are peer-reviewed papers in academic journals. One is a classic experiment from 1971.

Limits: most of the outreach studies measured email, not Instagram DMs. The similarity and reciprocity research ran in controlled settings, not live social platforms. The two vendor studies were published by companies that sell outreach software, and no one audited them. We report them because their methods are disclosed and their samples are large. We note the limits for the same reason.

Sources

1. Brian Dean, We Analyzed 12 Million Outreach Emails. Here's What We Learned, 16 April 2019. Backlinko, in partnership with Pitchbox. Sample: 12 million emails. Retrieved 18 August 2026.

1. BuzzStream data: The outperforming outreach tactics we learned from analyzing 12,000 campaigns, undated by publisher. BuzzStream. Sample: 12,000 campaigns across 2,000 accounts, over 30,000 individual emails coded. Retrieved 18 August 2026.

1. We analyzed over 5 million emails to find the ideal email length, undated by publisher. BuzzStream. Sample: over 5 million emails across hundreds of campaigns. Retrieved 18 August 2026.

1. Ashton Anderson and Kristen Altenburger, Causal Effects of Brevity on Style and Success in Social Media. Proceedings of the ACM on Human-Computer Interaction, Volume 3, Issue CSCW, Article 119, November 2019. Retrieved 18 August 2026.

1. Ali Faraji-Rad, Bendik M. Samuelsen and Luk Warlop, On the Persuasiveness of Similar Others: The Role of Mentalizing and the Feeling of Certainty. Journal of Consumer Research, Volume 42, Issue 3, October 2015, pages 458 to 471. Four studies, sample sizes not specified in abstract. Retrieved 18 August 2026.

1. Gregory R. Maio, Ulrike Hahn, John-Mark Frost, Toon Kuppens, Nadia Rehman and Shanmukh Kamble, Social values as arguments: similar is convincing. Frontiers in Psychology, 2014. Four experiments with sample sizes of 60, 50, 100 and 82 participants. Retrieved 18 August 2026.

1. Dennis T. Regan, Effects of a Favor and Liking on Compliance. Journal of Experimental Social Psychology, Volume 7, Issue 6, November 1971, pages 627 to 639. Retrieved 18 August 2026.

Changelog

• 2026-08-18: First version drafted.