RFP: VidLab Video Optimization Tool
cait.k...@moveon.org
VidLab Video Optimization Tool
overview
The goal of this tool is for campaigners (and in our case MoveOn Video Lab staff) to optimize videos through A/B testing. This testing happens through controlled use of the Facebook ads platform, even if the resulting optimization is then applied to a free, organic post.
The objective here is to create a tool that would be used by staff to test the optimal combination of videos, share texts, and thumbnails to maximize video views or actions. It could be used on every video they create and could become part of their typical workflow from video pitching to posting.
details
Facebook ads are organized into: Ad Campaigns › Ad Sets › Ads, which the tool is feeding into from the frontend UI via API.
The tool would do the following with user choice points called out:
① Create a new test
User Input:
Prompt user for: test name
Prompt user for: objective (video views or clicks)
Prompt user for start time: [Default = now + 1 hour]
Prompt user for test length: [Default = 4 hours. Can be 1-12 hours]
if clicks, Prompt user for: landing page.
Backend:
Create new Ad Campaign with:
ad campaign name = test name
ad campaign objective = objective user selected
② Enter Variants
User Input:
Prompt user for: video versions (v)
Prompt user for: thumbnails (th)
Prompt user for: share texts (tx)
[Maximum of v*th*tx = 18 allowed]
if clicks, Prompt user for: description
if clicks, Prompt user for headlines (h)
if clicks, Prompt user for: button label
Backend:
create one ad set for each v + th + x. Each ad set has one ad
set each ad set budget = $4, total budget = v*th*tx*$4
③ Confirm Setup
Show users each of the variants that will be tested.
[Tier 2: Allow specific combinations to be nixed or added]
Show proposed budget. Allow overall budget to be changed. If changed, prorate each ad budget to keep them equal while still totaling overall budget.
User hits confirm.
④ Set up Ads
Backend:
Details for all ad sets
Set custom audience: Lookalike audience (1%) to people who like MoveOn.org page
Set locations: United States
Set age: 18+ - 65 (all ages)
Set gender: all
No detailed targeting
Set edit placement: All devices, Facebook feeds only
Set optimization: Video Views -or- Clicks
Set bid amount: automatic
Set when you get charged: 10-second views -or- Clicks
Set lifetime budget (see above)
Set start time and end time
Set ad set name = testname_v#th#tx#
Details for ads
Ad names are same as ad set names. (And one ad per ad set)
If clicks, links have ?source=[ad name]
All ad set/ad combos should be submitted for review together.
⑤ Report results
[It’s likely that all of this will be done in a web application created by MoveOn staff, in which case nothing would be required from the developer other than to pull the data from Facebook and load it to Redshift]
Topline: Show aggregated key metric (10-sec views or clicks) for each component.
And then show a table of each combination tested with:
Ad name (e.g. ACAvideo_v2th3st1)
File name of uploaded video for ad
File name of uploaded thumbnail for ad
Share text of ad
Link to ad post
Amount Spent
Reach
3-Second Video Views
10-Second Video Views
Video Watches at 75%
Video Average Watch Time
Post Shares
Post Comments
Post Reactions
10V / Reach
75% / 3V
Shares / 3V
Reactions, Shares, Comments / Reach
Example of report:
“The ad with the most 10 second views was video #2, thumbnail #4 and share text #2”
Each numeric column is color coded red to green corresponding to the range of values from low to high for easy visual identification of the best performing videos based on each value. Ad sets can be re-ordered by column value by clicking on the column name. (Example table omits video filename, thumbnail filename, share text, and link to ad for simplicity.)
[Tier 2: For click campaigns, join data with Action Kit data to determine the number of signups or actions generated from each ad.]