The cold DM experiment: 1,319 messages, 9 cities, 4 countries — and why you can’t copy my results.

I ran an actual tracked experiment cold-DMing girls on Instagram. My real profile — you can see it here. Nine cities, four countries: Brazil, Thailand, Colombia, the Philippines. 1,319 messages sent. 273 replies. 67 conversations moved to WhatsApp. Here’s the breakdown, the exact method, and — more useful than the numbers themselves — why those numbers won’t transfer to you even though the method will.

If you’re looking for how to slide into DMs and actually get a reply, the method matters more than the line — this is what I tracked, city by city, to find out what that method really is.

A cold DM, if the term isn’t obvious, is a message sent to a stranger who has never talked to you and doesn’t follow you. No context, no shared friend, no prior interaction. That’s the entire test: can a message to someone with zero reason to reply get a reply?

Table of contents.

• The profile picture does the filtering before you write a word.
• Don’t overbuild the profile.
• The opener’s only job is curiosity.
• I push to WhatsApp fast — it’s a filter, not a convenience.
• I say what this actually is.
• The numbers.
• What actually jumps out.
• The benchmark.
• I’m not married to the channel. I’ll use whatever channel creates the outcome.

The profile picture does the filtering before you write a word.

When a stranger messages a girl on Instagram, it doesn’t land in her main inbox. It lands in requests — a separate folder she checks on her own schedule (you’ll be surprised how fast some girls do this) — and the only thing she sees before deciding to open it is your profile picture. If the picture doesn’t make her curious enough to click, the message underneath never gets read. It doesn’t matter how good the opener is. It never gets tested.

That’s the actual reason the picture outranks the opener — not a vague “it matters more,” but a sequence. It’s mechanical. Picture gets the click. Opener gets the reply. The bio matters least of the three, because by the time she’s reading it, she’s already decided whether to engage.

Don’t overbuild the profile.

The profile doesn’t need to be heavily curated. Actually, this is more likely than not to work against you. Heavily curated reads as optics management, or as I call it, “fake” — and most T1 girls are going to forgive a lot of things, but they won’t forgive fakeness.

Quick definition, since I use this same split elsewhere: T2 is a regular hot girl — attractive, clearly above average, but not T1. T1 is a genuinely super hot girl, Instagram-influencer level. Below both is everyone else. Also note that the Asian data below was T2-heavy, with a few T1 girls, while the South American data skewed heavily toward T1. Simply put, there are very hot girls everywhere. The question is just how many, and South America seems to be top-heavy while Asia seems to be top-thin for my particular classification.

Normal pictures. The important thing is that nothing reads as unattractive. This is the bar — not a styled grid, not an influencer feed. Building that out before you’ve run a single message is effort spent on a problem you don’t know you have yet. Fine-tune the picture or the opener only after you’re actually running numbers and the reply rate comes back bad. Test first. Adjust second.

The opener’s only job is curiosity.

The line I use — some version of “was it you I saw yesterday” — isn’t mine. It’s John Anthony Lifestyle’s technique; credit where it’s due. John is a friend of mine and happens to be world-class at this. And I’m always looking for expert advice in any field of interest. The opener works because it’s vague enough to trigger curiosity and specific enough to sound like it could be true. “Hey” gets ignored more. A claim that might be about her doesn’t.

I match the language to the country if I speak the language. And since I speak Spanish and Portuguese fluently, I DMed in the original language in these areas: Spanish in Spanish-speaking countries, Portuguese in Portuguese-speaking ones, and English across Asia. That’s the whole rule. Also, I didn’t properly track cold DMs when I lived in Europe and Eastern Europe, so I won’t invent numbers out of memory.

By the way, one thing worth testing, if the girl’s Instagram account is open, would be to find a picture of her you like and send it back to her with a comment like “I think I saw you yesterday.” I expect this to work well. Yet I haven’t tested it, so I can’t speak on it.

This is really the answer to how to slide into a girl’s DMs without it reading as generic — the opener’s only job is to make her curious enough to answer, not to be clever.

I push to WhatsApp fast — it’s a filter, not a convenience.

I push every conversation that gets a reply to WhatsApp quickly. Several reasons. I personally prefer WhatsApp over DMs. This is just me. Also, on WhatsApp you don’t have to compete with 500 guys DMing the girl and with reels sucking her in.

Also, for me it filters: a girl who won’t move off Instagram is telling you directly she isn’t that interested — no argument required, no reading her tone. WhatsApp isn’t “more legitimate” or somehow less online than Instagram. It’s exactly as online. The point was never the platform. It’s whether she’s willing to move to it at all. I wouldn’t do this if I didn’t have more incoming replies than I can possibly meet or handle. So if you want to replicate this, be sure that it’s okay for you if the girl stonewalls the conversation. In my case, I never bluff. If something is a boundary, it’s a boundary, and it gets enforced with no exceptions. So the WhatsApp test is a real-time saver for me — it separates girls who have high interest from girls who don’t.

I say what this actually is.

Before I actually ask for the switch to WhatsApp, I tell the girl that she’s very pretty, I’m interested in getting to know her, and I want to see how the chemistry is between us. Every message states real interest, plainly. Not a friendly hello. Not a social interaction. There’s no ambiguity about what kind of message she’s getting — and that’s what makes the WhatsApp filter work in the first place. You can’t filter for real interest if she doesn’t know what she’s being asked to be interested in.

The numbers.

Same profile, same opener pattern per country, tracked by hand: did she reply (yes/no), and did the conversation move to WhatsApp (closed)? Here’s every city I tracked.

Total: 1,319 messages. 273 replies. 67 closed.

This is a self-tracked field experiment: one man, one profile, one method, run by hand across nine cities. It’s still the largest first-party read I have on this.

What actually jumps out.

Two really big things.

First, there’s a huge difference in results for me between Asia and South America. As you can see, the gap between Manila, Philippines, and Goiânia, Brazil, is close to 17x — 3.32% versus 57.73%. That’s not a rounding difference. That’s a different world.

I won’t jump to the false conclusion that South America is easier to get a reply in. I think this is very specific to the man.

Second, the uncanny part is that it matches what happens to me in person on the ground, so to speak. In the Philippines, when I approach a girl, it’s very likely that she gets visibly anxious, looks around, and blows out — leaves the interaction — almost immediately after I say hi. In Brazil, girls in completely ordinary situations — a cashier, a receptionist, a girl at the gym — will stop whatever they’re doing to talk back to me. The DM data lines up with the street data almost too well. Same man, same looks, same everything, and yet different outcomes due to the environment.

Here’s the part that matters more than the numbers: this is specific to me. Friends who’ve tried similar things tell me the opposite pattern — most guys, in their experience, do better in the Philippines than in Brazil. My results are a function of my own look, my profile, and how I read culturally in each place, not a claim about which countries are “better” for men in general. I also don’t have the data to say this is city-specific rather than just country-specific. Take the ranking as mine, not as a rule.

The benchmark.

The realistic average reply rate for a cold DM to a total stranger sits at 1% to 5%, going by outside data on cold outreach generally. Manila — my worst city — comes in at 3.32%, right inside that range. That’s actually useful: it means the tracking is honest, not cherry-picked. My best cities, Goiânia and São Paulo, run 10 to 17 times that baseline.

I’m not married to the channel. I’ll use whatever channel creates the outcome.

If online paid off better for me, I’d run everything online. If a different offline method worked better, I’d run that instead. I do cold approach in person because I enjoy it more — it’s faster, and online is more of a patience game than I like. That’s a preference, not a claim that it performs better. Also, if you don’t care about Instagram’s TOS too much, you could have an assistant run this for you if you give them a script. This would be an efficient way to do it.

Across the roughly 22 countries I’ve lived in, Tinder works well almost everywhere for me except Asia. Asia is really not the place for me in this area. Bumble is strong in a handful of countries. Instagram cold DM swings hard by country, which the table above shows. I can’t speak to how any of this performs in Western Europe or the US right now — I left those areas around when Hinge came out, and I haven’t run this there since.

The method is universal. You don’t need to track every message, but if you want to know precisely what works for you, then yes, track them. For me, I like to rank the results by source, same as I would do in business, and put my effort where the signal is highest. That’s the real answer to how to slide into DMs that holds up — not a line to copy, but a method to run yourself. The specific numbers — which city, which country, which channel wins for you — aren’t universal. You won’t get my results by copying my line. You’ll get your own results by running the same test, then adapting to the feedback you get.