Facebook quick filter
Find out the intent of the campaign and map that to the attribution model.
Filter by Facebook and that attribution model.
Look at the ROI, sales, and revenue columns.
Click on benchmark and see how campaigns for the attribution model normally perform for that company.
So-so ROI campaigns can keep running if you have high customer LTV over time, or your sales velocity gives you reason to think that the campaign ROI over time is going to trend upward.
You can get your customer LTV from the ROI grid, but also look at the customer LTV for all sources to see if you have reason for optimism that you will get more value from your acquired customers. Then look at sales velocity to see how long new leads take to buy, IF this campaign is to cold traffic for lead gen. if you find that new leads take longer to buy then the campaign run time and time of acquiring the leads, you again have reason for future LTV optimism and can keep running the ads.
Further analysis is identical to a negative ROI ad campaign, which is this: If similar intent campaigns had negative ROI in a week, at a similar ad spend, and similar clicks & leads, you can hang in there if the similar intent campaign eventually hit ROI positive in a time frame you are comfortable with. Otherwise it might be time to cut bait.
A key factor here is your sales cycle and your customer LTV. If you have high customer LTV, and your sales cycle takes longer than a week, you may find this campaign swing to ROI positive soon. Past performance is the guideline we use.
Assuming your new campaign is "new and improved" you'd like to see better performance or at least equal performance than the past. If your sales velocity is 2 weeks, you might need to give the ad a few weeks to know the true ROI. If the numbers or your budget don't allow for that, you can stop the campaign, then re-check the ROI after your typical sales cycle for the leads acquired from this campaign has completed.
Additionally, click on the benchmark icon for that row. Pick the attribution model / intent, and you vs. yourself benchmark. See how this row's performance compares against your past.