You don't need a model with 40 inputs to forecast social content. Take your past posts, group them by format and theme, find the median result of each group, and multiply by what you plan to make. Then divide the cost of producing each group by that result, and fund the cheapest results first.
TLDR: Tag past posts by format and theme, take the median result per group, give each group a production cost, and rank groups by cost per expected result. Fund the top groups, cap the bottom ones, and re-run the numbers each quarter.
Start with a table of 90 days of posts
Export the last 90 days and add two columns: format and theme. Pick one result that matches your goal, such as engagement rate, saves or link clicks. Just one, because mixing metrics makes the ranking meaningless.
Here is the structure, with illustrative numbers:
| Group | Posts | Median result | Cost per post | Cost per expected result |
|---|---|---|---|---|
| Carousel, education | 22 | 4.1% | $90 | lowest |
| Short video, behind the scenes | 18 | 3.0% | $150 | middle |
| Photo, offer | 30 | 2.5% | $40 | middle |
| Studio video, brand story | 8 | 2.2% | $600 | highest |
The numbers are examples. Yours will differ, and that is the point.
Use the median and a range, never one number
A single viral post bends an average. Use the median, then also note the middle half of results. In Mydrop's data, the middle half of Instagram posts ran between 1.1% and 6.7% engagement, with a median of 3.1%. That spread is normal, so your forecast should be a range with a most-likely value, not a promise.
If a group has fewer than 20 posts, say so and label the forecast "low confidence". Low confidence is a reason to test with a small budget, not to ignore the group.
Adjust for what history cannot see
Past results assume the future looks like the past. Make three adjustments, each in plain words:
- Seasonality. If a theme only works in one quarter, forecast it for that quarter only.
- Fatigue. A format you have posted 30 times in a row will drift down. Cut the forecast by a stated percentage.
- Posting time. Midnight to 6 am ran 17% below the median in Mydrop's data, so do not forecast a result for a post planned there.
Write the adjustment next to the number, so the next person can see why the forecast moved.
Rank by cost per expected result
Divide the production cost of each group by its median result. A cheap photo with a 2.5% median can beat a $600 video with a 2.2% median by a wide margin. The ranking shows where more money does nothing.
Then set three buckets:
- Fund fully: the top third by cost per expected result.
- Fund as proven: the middle, at current volume.
- Cap or retire: the bottom third, except where the post has a non-social job such as a sales video.
For the clean-up side of this decision, see reducing creative debt.
Keep a small test budget
Put 10 to 15% of creative spend into things your history cannot rank, such as a new format or theme. Without that bucket, the forecast just confirms the past. Treat each test as 8 to 10 posts, then add them to the table.
Tag every post to keep the forecast alive
The forecast is only as good as your tags. A tagging schema with format and theme as required fields keeps the table current. Mydrop's Performances analytics shows results per post and per platform, so the export takes minutes, and the calendar shows what is already planned in each group.
Show the forecast as three scenarios
Budget owners trust a forecast that admits uncertainty. For each funded group, write a low, likely and high case from the middle half of past results. Then show what each case means in plain terms: how many extra saves, clicks or leads, and what the group costs. If the low case still beats the cost of the next best use of the money, the choice is easy. If only the high case justifies the spend, move it into the test bucket. This framing turns a debate about opinions into a short check against numbers everybody can see.
Build the table before your next budget meeting
Export 90 days, add the two columns, work out the median per group and divide cost by result. Bring one slide with the ranking and one request: move the money from the bottom third to the top third for one quarter.
Google review
Trustpilot review
Product Hunt comment
