Cross-Platform Customer Feedback: Build a Source-Linked Evidence Brief
A cross-platform customer feedback workflow can organize selected Reddit, YouTube and TikTok observations into one evidence brief. The goal is a traceable list of questions, contradictions and tests to run. A common table makes sources easier to review; it does not make different platforms, collection methods or audience samples statistically equivalent.
Inspect the inputs, access requirements and current credits before running a request.
Quick answer: Keep source facts, interpretation and delivery state separate. The downloadable helper works locally with saved inputs and makes no API or AI calls.
On this page: Mapping · Dated verification · Downloads
Reviewed November 5, 2026. Vendor-authored guide based on public contracts, official platform documentation and the explicitly labeled checks below. A real API response does not establish a successful n8n/Make platform import, live Sheets write, AI draft or performance uplift.
What research question can the platforms share?
Define a buyer situation before combining any rows. “Can this mug be carried in a backpack without leaking?” is narrow enough to examine across selected discussions. A general coffee video, a restaurant review and a mug complaint may all be about coffee but do not automatically answer that same question.
Record product identity or version, audience situation, time window, language and collection method for every source. The paid examples elsewhere in this cluster demonstrate API behavior; they were not collected as a matched cross-platform study. Do not pool their counts into a customer-prevalence statistic or present them as evidence that one problem dominates a market.
What belongs in a common evidence record?
Use a local evidence ID plus the actual source context. The companion schema uses evidence_id, platform, source_record_id, source_url, source_context, observed_at, text, theme and sample_type. Preserve parent or reply context where available. A local evidence ID makes a brief traceable even when a trimmed source has no stable comment ID.
Keep source_record_id empty when unavailable. In the TikTok trimmed sample, a video ID did not uniquely identify individual comments. Do not fabricate a comment ID by relabeling the video ID. If only a source video URL is available, label the reference at video level and preserve the observation-local row label.
How do you combine evidence without another API request?
Export already reviewed rows and use the local helper. The feedback command validates the common record fields and produces a CSV plus a small theme inventory. It has no network client, API credential or LLM connection. The supplied demonstration input is constructed data, explicitly labeled synthetic, so the command can be tried without collecting or publishing people’s comments.
Themes are supplied by a reviewer; the helper does not pretend to understand sentiment or classify buyer intent. It can count retained rows by a provided label, but those counts describe the file, not a population. It leaves source fields intact so a reviewer can inspect every observation behind a proposed finding.
How should you handle duplicates and contradictions?
Keep identity duplicates separate from repeated ideas. A duplicated source ID can be the same observation copied twice. Similar text from different people can be different observations. An edited comment or a screenshot repost can require manual review. Do not merge them merely because an AI summary calls them equivalent.
Record counterevidence beside a theme. A complaint that a cup retains heat poorly and a request for a cup that cools faster can reflect different use situations. A useful product brief preserves that distinction instead of averaging them into a meaningless sentiment score. Look for the conditions that explain the disagreement.
What should a reviewed evidence brief contain?
Separate source statements, interpretation and proposed experiments. Each finding should include supporting evidence IDs, source URLs, contradictions, missing context and the next test or documentation check. If a question is unsupported by the selected sample, leave it open.
An optional AI prompt can organize reviewed rows, but it must treat retrieved text as data and cite evidence IDs. Do not include credentials or individual contact information. The included prompt does not execute a model. A human decides which findings deserve an original article, product change or user interview.
What claims should the report avoid?
Do not compare platform engagement counts as one shared unit. A Reddit score, a YouTube like and a TikTok view measure different activities. They cannot be added into a valid popularity score without a separately justified model and comparable inputs. Selected comments do not establish how common a need is among all buyers.
Use the resulting brief to guide further research, then validate consequential claims with appropriate product evidence. Keep the underlying collection limits, missing fields and sample definitions visible. For implementation details, follow the separate Reddit comments, YouTube comments and TikTok comments guides.
What did the dated verification establish?
The APIs in this cluster have real dated examples, but they are not a matched cross-platform buyer study. In particular, the TikTok reference-video sample and the YouTube AeroPress review should not be pooled as if both described the same mug. The download below uses four constructed rows to test common-schema handling without claiming actual customer findings. Its themes are supplied labels, not AI-inferred results.
See the sanitized batch evidence record for scope and billing metadata. Full comments, author profiles, keys and account identifiers are not published. API tests do not establish authenticated n8n/Make execution, a successful Sheets write or an AI-generated report.
Field mapping and implementation checkpoint
| Evidence | Field or state | Handling |
|---|---|---|
| Evidence ID | Local unique evidence_id |
Required for every brief finding |
| Source identity | Nullable source_record_id |
Never substitute a video ID for a comment ID |
| Scope | source_context, sample_type |
Synthetic examples stay explicitly synthetic |
| Theme | Reviewer-supplied theme |
No semantic classifier is executed |
Run the constructed example
Download the explicitly synthetic input, then run:
python social-workflow-evidence-ops.py feedback --input cross-platform-feedback-synthetic-demo.json --output reviewed-evidence.csv
The helper rejects duplicate local evidence IDs, repeated non-null source identities within the batch, invalid source URLs and timezone-free observation dates. It does not merge repeated ideas or compute population-level prevalence.
Downloads and verification scope
The helper runs without API credentials and makes no network or model requests. n8n Code logic and local state behavior are checked with saved or constructed inputs as labeled. Make recipes are manual configuration notes, not authenticated scenario exports. Configure and test a scratch destination before enabling writes.
Read the related research guide and workflow recovery guide for context.
Sources
- Nexscope: matching API contract — current operation and source limits.
- n8n error workflows and Make error handling — recovery mechanisms to inspect before replay.
Frequently asked questions
Can platform counts be added into one popularity score?
Not without a separately justified model and comparable inputs. Scores, likes, views and selected comment counts are different measures.
Does the helper classify sentiment or call an LLM?
No. It validates supplied evidence and reviewer-provided theme labels locally, without network or model calls.
Are the demonstration feedback rows real customer comments?
No. The download labels them as synthetic and illustrative; they are not customer testimonials or research findings.
Last reviewed: 2026-11-05 · Maintained by the official Nexscope team · Editorial & evidence policy.