Sort by what’s on screen
Campaign names describe intentions. What’s on screen describes what people saw. Watch every post and give it one label for the thing that happens: a trick with the product, a joke about friends, the product as the character, a scene where nothing happens. If a post doesn’t fit, put it in “other.” You’ll come back to that bucket.
A few rules make the labels hold up. Label the thing that happens, not the mood or the message. One label per post. If two people would label a post differently, write down why and pick one. Watch with the sound off once: if the joke still lands, note it. In the report below, the most shared posts read with the sound off.
In a real Arcus content report for a hard seltzer brand, every Instagram and Facebook post from one summer campaign was watched and sorted this way: 39 Instagram posts and 29 Facebook posts. See the content report.
Use medians, compared with the account average
The median is the middle post in a group. It keeps one viral hit from hiding a weak pattern. Divide it by the account’s average per post and you get a number anyone can read: above 1 beats the account, below 1 doesn’t. Here is the Instagram scorecard from that report:
| What’s on screen | Posts | Median views | vs. average | Median shares |
|---|---|---|---|---|
| Hosting hacks: a trick with the can | 4 | 19,188 | 2.36x | 26 |
| A moment people were in* | 3 | 8,000 | 0.98x | 46 |
| Other | 7 | 6,186 | 0.76x | 14 |
| Can tricks and gags | 11 | 6,107 | 0.75x | 13 |
| Workday escape | 7 | 3,962 | 0.49x | 10 |
| Unusual hangouts | 5 | 2,415 | 0.30x | 12 |
| 3D animation* | 2 | 1,139 | 0.14x | 8 |
Account average: 8,148 views and 36 shares per post. *Fewer than four posts, so directional.
Check for one-post groups
Medians help, but also ask how much of a group’s total comes from its best post. In the same report, one post, a sign held up from a balcony, carried 45% of its group’s views and 77% of its shares. Without it, the group’s median fell from 3,962 to 3,119 views, and the group was the second weakest on the account. The hit was a permission gag with a sight joke. Two anti-work posts in the group, played straight, stayed under 2,300 views each. One post had been carrying an idea that didn’t work.
Now check the other direction: posts in “other” that land together. In the report, three posts about the end of a big summer tournament weren’t in any named group. Together they were number one by shares and number two by views on Instagram. The brief said to name a group when two or more posts land together, so the report named it a candidate, marked it directional, and set a test: does the next big shared moment do the same?
Watched and shared are different jobs
Hosting hacks led on views but sat below average on shares: a median of 26 against the account’s 36. The most shared post on the account wasn’t in any named group. It was a joke about the viewer’s friends going back to work after a big tournament: 409 shares on 8,000 views, 11 times the Instagram average for shares. Facebook was different again: it watched, and barely shared, at 1.2 shares per post. Our guides to watched versus shared and what to post on Facebook go deeper.
So build two scorecards, one for views and one for shares, per platform. A group can win one and lose the other, and that’s useful. It tells you which job each kind of post does, so you can plan a mix instead of chasing one number.
A checklist for your next review
- Export every post from the period, with views and shares, from your own accounts.
- Watch each one. Give it one label for what happens on screen.
- Work out the account average per post for views and shares, per platform. Never add platforms together.
- Find each group’s median and divide by the average.
- Mark groups of fewer than four posts as directional.
- For each group, check the best post’s share of the total. If one post carries most of it, read the group without it.
- Look in “other” for two or more posts that land together. Name them as a candidate group.
- Write three lines: keep, test, stop.
Here are the three lines from the report, as an example of how specific they should be. Keep hosting hacks: more physical tricks with the can that someone could copy tonight. Test a friend-group frame on top of the formats that already get watched, and watch the share rate. Retire 3D animation, and the workday idea as written: keep the permission gag, drop “quit your job.”
Questions worth asking your team: Do we group posts by brief or by what’s on screen? Which group would we stop? Which hit is carrying a weak group?
Common mistakes, and what this doesn’t prove
- Ranking by totals. A group with more posts wins by volume.
- Reading share rate alone. Low-view posts can post high rates on a handful of shares. The 3D animation group had the highest median share rate on Instagram, 0.9%, and the lowest views.
- Calling a group on two posts. Mark it directional and test again.
- Sorting by campaign name. Two posts from the same brief can do opposite things on screen. Label what happens.
- Ignoring “other.” In the report, “other” had higher median views than three of the four named groups, and the most shared post on the account came out of it.
- Forgetting test posts. In the report, one post was a test cut at 212 views. It stayed in the data, labeled, and wasn’t read as a creative failure.
This tells you which kinds of posts get watched and shared. It doesn’t tell you which ones sold. Lining posts up with your own sales, tickets or sign-ups is the next step, and a test is the step after that. Arcus does both once your social accounts and your sales sit in one place you own. To start, try it free with a question like “Which of our posts moved sales?”