The short answer: Instagram's head, Adam Mosseri, has said the three signals that matter most for ranking are watch time, likes and sends, and that you should track them as average watch time, likes per reach and sends per reach. Likes matter a little more for reaching your followers. Sends matter a little more for reaching people who do not follow you. So if growth is the goal, the question for every Reel is not "will people like this?" but "will anyone forward this?"
This guide turns that into a table, one formula, a test you can run on any idea and a monthly report.
The signals in one table
Mosseri shared the three signals in January 2025. Instagram's own explainer on ranking adds that Feed, Stories, Explore and Reels are each ranked separately.
| Signal | What it measures | Where it matters most | What to make for it |
|---|---|---|---|
| Watch time | How long people stay on the Reel | Everywhere. It is the first of the three | A first line that states the point, then a reason to stay to the end |
| Sends per reach | Of the people who saw it, how many shared it in a DM | Reaching non-followers | Something a person would forward to one specific friend |
| Likes per reach | Of the people who saw it, how many liked it | Reaching your existing followers | Posts your regulars recognise themselves in |
| Closeness (Stories) | How often someone views and replies to your Stories | Stories only | Polls, question boxes, replies you actually answer |
| Originality | Whether you made it | Recommendations | Your own footage, your own voice |
Notice that sends and likes are counted per reach. They are ratios, not totals. A big number on a post that reached a lot of people can still be a weak post.
How to calculate sends per reach
Sends per reach = sends ÷ accounts reached × 100
Both numbers are in the insights for each post. A worked example with round, made-up numbers:
| Reel A | Reel B | |
|---|---|---|
| Accounts reached | 2,000 | 50,000 |
| Sends | 60 | 200 |
| Sends per reach | 3% | 0.4% |
| Likes | 150 | 2,500 |
| Likes per reach | 7.5% | 5% |
Reel B looks like the winner in a screenshot: more reach, more likes, more sends. But a far smaller share of the people who saw it thought it was worth passing on. If B had a paid push or a lucky trending audio, it tells you little. Reel A is the idea to make again.
We are deliberately not giving you a "good" benchmark. Published ones vary widely by account size and category. Use your own median instead: work out sends per reach for your last 20 Reels, find the middle value, and treat anything at double that as a winner.
Rule of thumb: judge a Reel against your own median, not against someone else's screenshot.
The forward test
Before anything is shot, write one sentence: "[This person] would send this to [that person] because [reason]." If you cannot fill in all three blanks, the idea is not ready.
Here is the test run on Reel ideas for Kadak, a made-up chai brand that sells online.
| Reel idea | Who sends it to whom, and why | Verdict |
|---|---|---|
| "Three mistakes that make your chai bitter" | A daughter to her father, who boils the leaves for ten minutes | Pass |
| "Types of chai drinkers in every office" | A colleague to the team group: "this is you" | Pass |
| "What ₹10 of chai costs to make at a tapri", filmed at a real stall | A friend to the friend who is always saying he will open a cafe | Pass |
| "Monsoon gift box, 15% off this week" | Nobody. People do not forward discounts unless they are remarkable | Fail. Run it as an ad or a Story instead |
| A slow pan of the new packaging with music | Nobody | Fail |
| "We asked our packing team how they take their chai" | A team member to their family, and regular customers to each other | Pass, and it helps likes per reach too |
The reasons people forward are few: it is useful to someone they know, it is funny about someone they know, it says something they wanted to say, or it is news in their circle. A product shot with a price is none of these. That does not mean never posting offers. It means offers are for people already close to buying, so they belong in Stories, ads and WhatsApp, not in the Reel you hope will travel.
Originality rules
Instagram has been specific about originality:
- When it finds identical copies of a piece of content, it recommends only the original.
- Accounts that post other people's content ten or more times in 30 days, without adding to it, are removed from recommendations until they drop below that number.
- Content with visible watermarks is not eligible for these recommendations.
- Content that is materially edited is treated differently: a new voiceover, a reaction, a parody.
For a brand, this rules out the easy calendar of reposted memes and re-uploaded TikTok files. It also means a trend is fine if you add to it. How we plan original footage is covered in our guide to shooting raw Reels.
A monthly report template
Most social reports lead with followers and total likes. Swap that page for this one. The numbers in the example row are made up.
| Reel | Reach | Avg watch time | Sends | Sends per reach | Likes per reach | Non-follower reach | Make again? |
|---|---|---|---|---|---|---|---|
| Bitter chai mistakes | 12,000 | 11 sec of 24 | 300 | 2.5% | 6% | 70% | Yes, as a series |
| Your median, last 20 Reels |
Under the table, three lines only:
- Top two by sends per reach, and what they had in common.
- Bottom two, and what you will stop making.
- Next month's test: one variable, such as two different opening lines for the same Reel.
Followers and total reach can go at the bottom. They are results of the numbers above, not levers you can pull.
Four mistakes we see
Asking for the send. "Share this with a friend who needs it" at the end of a dull Reel changes nothing. The idea has to earn it.
Reading totals. A boosted post will always have the biggest numbers. Ratios show which ideas worked.
One post for every surface. Stories reward closeness and replies. Reels reward watch time and sends. Cutting one asset to fit all of them usually suits none.
Ignoring the replies. A send often starts a conversation that comes back to you as a DM. Answer those quickly. This is where sales begin.
A 30-day plan
- Week 1: Pull the last 20 Reels. Calculate sends per reach and likes per reach for each. Find your median.
- Week 2: Run the forward test on next month's ideas. Drop or move the ones that fail.
- Week 3: Shoot the survivors, with two opening lines each.
- Week 4: Report on the template above. Double down on the top two formats.
Where this fits
This is how we plan and report social media marketing: ideas chosen for the send, judged on ratios. The Reels themselves come from our content creation team, and the same forward test is a good filter when choosing creators for influencer marketing.
