The twelve design decisions that separate high-converting Instagram DM automations from the rest, ranked by strength of evidence, from 798,041 real conversations.
Twelve design decisions separate the Instagram DM automations that convert from the ones that do not, and they come from real data: 798,041 conversations across 753 automations. The biggest wins are to put the link on the second message, not the first (42% versus 29%), keep the first message under two sentences, match one or two keywords on a specific post rather than six keywords on every post, offer one button rather than three (32% versus 2%), and build in no delays, because the median conversion happens 65 seconds after the trigger. Every rule here is a pattern across different businesses rather than a controlled test, so copy the strongest-evidenced ones and treat the rest as a hypothesis to run on your own account.
Treat every rule here as a strong hypothesis, not a law. They come from comparing 753 Instagram automations that handled 798,041 conversations over 90 days, and they are ordered by how much evidence sits behind each one, with every rule carrying the number of automations it is based on. What they are not is a controlled experiment. Different businesses built these automations for different audiences with different offers, and the biggest variable of all, whether the thing on the other side of the link is worth having, is invisible in the data. Where a rule rests on only a handful of automations, we say so, and you should weight it accordingly. The right way to use this guide is to copy the strongest-evidenced rules first, then test the rest on your own account.
These are ranked by the strength of the evidence behind them, not by how interesting they sound. The figure in bold on each rule is the gap between the automations that follow it and the ones that do not.
The dataset behind these rules. Inrō data.
42% versus 29%. 753 automations. Automations that deliver the link on the second or third step convert at 42%. Leading with the link straight away drops you to 29%, and waiting until the fourth step or later drops you to 26 to 28%. The shape that works is one short line confirming what the person asked for, then the link. Leading with the link skips the beat where the person recognises that you understood them; burying it costs you everyone who drifted away between messages. This is the rule that survives the toughest test in the dataset, comparing comment automations only against other comment automations, so the result is not just an artefact of one trigger type being better than another.
39% under 120 characters versus 22% over 320. 753 automations. The first message is the only one every triggered person is guaranteed to see. Under about 120 characters, length makes no difference at all. Between 120 and 320 you lose a few points. Past 320 characters conversion falls off a cliff, and those automations are five to six times more likely to end up in the worst-performing group. Measured in words rather than characters the same shape appears: under 10 words converts at 39%, over 60 words at 23%.
45% versus 28%. 260 and 35 automations. One keyword is a clean instruction. Six is defensive coverage of spellings, plurals and translations, and it fills your automation with people who typed something adjacent rather than the thing you asked for. If you genuinely need to cover two languages, use two automations rather than one automation with ten keywords.
43% versus 15%. 226 and 31 automations. Hand-picked posts plus a keyword is a campaign, and it converts at 43%. Every post plus every comment is a catch-all, and it converts at 15%, with almost six in ten of those automations landing in the worst-performing group. If you want something always-on, use caption matching instead of a fixed post list: automations that pick posts by matching words in the caption convert at 43% and follow your content automatically. The same logic applies to DM triggers. Automations set to fire on any incoming message convert at 12%, and seven in ten of them sit in the bottom group. Always match a keyword.
32% versus 2%. 1,429 and 343 individual messages. A button message with one option converts at 32%. Two options convert at 38% but get clicked half as often. Three options convert at 2%. This is measured at the level of individual messages rather than whole automations, which makes it one of the cleanest comparisons available: it measures the button, not the business that made it.
39% versus 32%. 611 and 142 automations. Email sequencing habits do not transfer. Automations with no waiting time anywhere convert at 39%; ones with a delay convert at 32%. Message count points the same way: two or three messages convert at 41%, more than eight convert at 30%. The reason is timing. Median conversion happens 65 seconds after the trigger, while the person is still holding their phone. A delay does not build anticipation, it returns them to the feed, and if they never reply, Instagram's 24-hour messaging window closes behind you. Run nurture sequences as separate DM campaigns to people you have already filed, not inside the automation that answers the comment.
46% versus 29%. 88 and 60 automations. Automations that post a public reply under the triggering comment convert markedly better than ones that only send the DM. A fixed reply performs best in this data at 46%; rotating between several variants converts at 38%, and an AI-written reply at 39%. Part of this is social proof, since everyone scrolling sees that the account responds, and part is simply that people check whether their comment worked.
39% versus 36%. 310 and 443 automations. Gating the link behind a follow costs you almost nothing in conversion and buys you followers. Automations that open with the follow request convert at 43%, better than ones that open with a button at 37% or a plain message at 32%. The caveat: the follow gate is also one of the more failure-prone steps in the product, failing about one time in ten. If your follow-gated automation looks like it is under-performing, check whether it is failing rather than being ignored.
40% with none or one versus 26% with more than seven. 409 and 13 automations. The general finding on copy decoration is that less is more. Automations with the fewest detectable copy techniques in the first message convert best. Bullet lists, all-caps shouting, urgency language, promo codes and named prices all go with lower conversion. The worst offender is a raw URL pasted into the message text rather than a tracked link: those automations convert 21 points lower, and the click is invisible to you anyway. The heaviest-emoji comparison rests on just 13 automations, so read the direction, not the exact figure.
A question adds about 3 points, a follow request about 7. 753 automations. Of all the things you can detect in a first message, asking for a follow and asking a question are the two that go with higher conversion. A greeting helps slightly. Almost everything else is neutral or negative. A question also has a mechanical benefit: a reply keeps Instagram's 24-hour messaging window open.
40% at four to seven actions. 487 automations. Fewer than four steps usually means something is missing: no public reply, no filing, sometimes no tracked link. More than twelve and the flow gets fragile, with more places to fail, more branches that never fire, and more chances for the person to drop out. The most common shape among high performers is six steps: a public comment reply, the ask, the tracked link, a closing message, filing the contact, and firing the pixel.
40% versus 31%. 520 and 230 automations. Adding the person to a folder does nothing for the conversion happening right now, so this is really a signal about the builder: people who organise their audience also tend to write better automations. Copy it anyway, for two reasons. It is the cheapest habit on this list, and the folder is what makes next week's campaign possible. Accounts using more than eleven tracked links convert at 50%, against 10% for accounts using one or none: measurement and organisation travel together.
Worth knowing, if only so you do not waste time on them.
The shape that wins is simple to build: one keyword, one post, one button, and the link on the second message, with a public reply under the comment and the contact filed at the end. If you only change one thing about an automation you already run, move the link to the second message and cut the first message to two sentences, the two largest and best-evidenced gaps in the data, both about a minute of work. The rest is worth testing on your own account, because the one variable that matters most, whether your offer is worth having, is the one no dataset can measure for you. When you want a builder that assembles the whole shape and tracks every step so you can see where people drop, the comment-to-DM builder does it on a free plan.
A public reply under the comment, a short question with one button, the tracked link, a one-line closing message, and the contact filed into a folder. Six steps, first message under 120 characters, and the link on step two. That is the shape the data points to again and again.
The guided comment-to-DM builder, unless you need something it cannot do. Automations built with it convert at 42% against 32% for custom-built ones, because it constrains you into a shape that works rather than letting you add the delays and branches that drag performance down.
Two or three, which converts at 41%. More than eight drops to 30%. Every extra message is another chance to be ignored, so send the fewest that get the person to the link.
Duplicating an automation you already have working is the strongest starting point in the data, at 43%. Building from scratch converts at 37%, and generating one from an AI prompt at 34%, where longer prompts do worse rather than better. A short, clear description beats a detailed one.
The copy and structure rules do, since they are about the message rather than the entry point. The trigger rules are specific to each entry point. There were too few Story-reply automations in this dataset to give reliable guidance on that entry point specifically.
Move the link to the second message and cut the first message to two sentences. Those are the two largest and best-evidenced gaps in the data, and both take about a minute to change.
Attract more leads, target them with DM marketing, and automate all your interactions on Instagram!
.png)
