Where do the assets come from?
Break reference videos into frameworks, characters, scenarios, stories and connections. People collect links and review results; AI performs structured analysis.
Product Marketing Methodology
PRODUCT CONTENT / Methodology
From successful-content analysis to product matching and script assembly.
Using the AI video efficiency project for Niu Qian's team at Kuangben, explore how assets enter a library, product benefits match audiences and scenarios, and frameworks expand into scripts across product lines.
FDE / Product Marketing Methodology
General AI video production workflow before improvement
General AI video production workflow before improvement Start with three concrete questions
How do benefits match the assets?
Identify the product's primary benefits and moments of need, extract visual evidence, then index the corresponding character, scenario and story assets.
How do you keep producing scripts?
Retain frameworks and reasons to buy, replace products, character relationships and scenarios, and explain each segment's purpose and connections.
3–4 → Dozens / day
Video analysis: from manual creative analysis to AI analysis with human review.
≤10 → Dozens / day
Script assembly: from manual production to batch assembly using a single Skill.
Before: repeated handoffs among creative producers, generation specialists and editors
Creative producers manually analyzed characters, visuals, actions and dialogue to create brief scripts. Generation specialists wrote prompts from reference images and shot timing, then passed generated videos to editors. Information moved through spreadsheets and verbal communication, often requiring repeated revisions.
Content system: turn assets into a searchable resource
Collect proven references first, let AI analyze them by field, and have people review and correct the results. Save decision rules, prompts and approved cases for reuse in product analysis and script production.
| Asset layer | What to retain | How to use it |
|---|---|---|
| Frameworks and connections | Content segments' roles, order and transitions | Reuse structures and retain reasons to buy |
| Characters, scenarios and stories | Relationships, actions, environments, triggers and causal chains | Search by the product's audience and moments of need |
| Products and evidence | Primary and secondary benefits, conditions of use, supporting visuals and restricted claims | Keep product facts in scripts within supported bounds |
| Results and versions | Source records, element IDs, test results and revision rules | Trace sources and retain effective combinations |
Detailed analysis: turn a video into reusable fields
Record visible and audible facts first, then organize frameworks and character relationships. Separate facts, reasonable inferences, product claims and creative extensions so reference content is not treated as evidence for your own product.
| Analysis layer | Key records | Delivery purpose |
|---|---|---|
| Timing and shots | Timecodes, original copy, shot size, angle, camera movement, subject actions, cuts and audio continuity | Reconstruct segment boundaries and shot transitions |
| Characters and needs | On-screen characters, target audiences, users, buyers, relationships, triggers, specific tasks and expected outcomes | Distinguish who appears from who buys |
| Scenarios and stories | Location zones, foreground/midground/background, props, lighting, lived-in details, triggers, actions, conflict and product entry | Support scenario recreation and story adaptation |
| Benefits and evidence | Pain points, benefits, claims to prove, evidence levels, conditions of use and prohibited wording | Link needs to supporting evidence |
| Adaptation and cataloging | Elements to retain or replace, unique element IDs, source records and tags | Make assets searchable, traceable and reusable |
Data dimensions: connect content judgment to business results
Process metrics help locate problems, while advertising and conversions check business outcomes. Align snapshot dates, collection times and measurement periods when combining tables. Keep different business lines and production stages distinct.
| Data source | Tracked dimensions | Decision supported |
|---|---|---|
| Qianchuan advertising dashboard | Ad spend, asset spend ranking, advertising ROI and conversions | Compare the top 10 and bottom 10 asset groups by spend to select references worth analyzing |
| Advertising process metrics | First-three-second retention, completion rate, likes and comment content | Assess opening appeal and use comments to identify further customer needs |
| Reference content performance table | Likes, comments, shares, saves, on-screen comments and changes; engagement rate, duration, publishing time and collection time | Retain performance context and discover topics and presentation methods |
| Product analysis master table | Counts of validated benefits, candidate benefits, supporting visuals, assets, formulas/qualifications and pain points/value propositions; analysis status | Track accumulated product knowledge and analysis completion |
| Product analysis details | Applicable audiences, scenarios, evidence descriptions, shot sizes, actions, sound/touch, recommended and prohibited wording, compliance levels and priorities | Provide reusable fields for script retrieval and storyboard production |
| Production operations | Average daily output per person, generation retry rate, weekly template count and asset cost (CNY/second) | Measure efficiency and cost separately for analysis, scripting and video production |
The document mentions 66 collected works. This page shows the tracked dimensions and decision process. ROI, retention and other metrics are left unspecified where complete-period data is unavailable.
Script assembly: explain what was selected, why and how it connects
- Confirm the product and objective
Read product materials, confirmed benefits, applicable audiences, moments of need and available evidence.
- Read existing assets
Respect actual table headers and verify source videos and records. A sentence pattern alone does not justify inventing a complete original framework.
- Assemble content segments
Choose story, sentence-pattern or direct-advertising segments. Explain their roles, order and transitions, then match people, settings and audio.
- Review and feed back
Label retained, adapted and newly generated parts. Review product facts and production conditions, then feed results back into the rules library.
A complete video consists of content segments that can be retrieved, replaced or adapted independently. The framework defines their order, roles and connections. Segments may use a customer, merchant, creator or narrator perspective without forcing them into one continuous story.
Based on “Kuangben AI Video Generation Efficiency Case: Key Outcomes of Niu Qian's Team.” The materials comprise meeting notes, process screenshots and additional client feedback. Some execution rules, including “moments of need first,” were added later and are not presented as verified meeting conclusions.