What the data for this category looks like
The core data for this category comes from landing page access logs, heatmaps, form submission data, and advertising source parameters. Associated daily documents include landing page plans, point burying requirement documents, A/B test reports, and form field lists. Data update frequency is weekly or per single campaign cycle (approximately 7–14 days). The most time-consuming processing steps are page path analysis, form blockage troubleshooting, and experiment result judgment. Abnormalities can lead to lead loss, inability to attribute data, or increased advertising costs.
What constraints these characteristics impose on the knowledge base retrieval and recall link
The data sources for landing page conversion optimization are mostly user behavior information. Associated documents include structured point burying fields, A/B test conclusions, and campaign rules. Precise recall for business fields must be supported. The data update frequency is weekly or per single campaign cycle. Retrieval must adapt to temporarily updated campaign documents to avoid recalling expired content. Time-consuming steps such as page path analysis rely on fine-grained access data, so retrieval must recall precise documents matching the user's access node, while controlling retrieval latency to support the real-time nature of pre-sales Q&A. In addition, the structured characteristics of documents such as form field lists require retrieval to distinguish between business fields and general descriptions, preventing irrelevant content from mixing into results and affecting the accuracy of Q&A.
How to set the configurations
The following configurations are adapted for FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 semantic matching threshold | Matches professional terminology and business fields related to landing pages, avoiding recall of irrelevant general documents |
limit | Top 3–5 retrieval results | Adapts to the concise Q&A needs of landing page conversion optimization, controlling context length |
usingReRank | Enable re-ranking function | Distinguishes between different types of business documents such as A/B test reports and form fields, improving recall accuracy |
datasetSearchUsingExtensionQuery | Enable question extension retrieval | Automatically expands user pre-sales questions to match hidden rules in landing page plans |
collectionIds | Specify exclusive knowledge base ID | Isolates business data, avoiding recall of documents from other categories |
searchMode | hybrid retrieval mode | Combines keyword and semantic retrieval, adapting to both structured and unstructured documents related to landing pages |
Three common mistakes
- Symptom: Unable to repeat the content of the uploaded landing page Excel form field list during conversation, with a prompt that the file cannot be read. Cause: The extended retrieval configuration for structured files is not enabled, or the knowledge base did not correctly parse the cell fields of the Excel file.
- Symptom: Retrieval results include expired A/B test reports, and cannot match the current campaign rules. Cause: No retrieval condition for filtering documents by campaign cycle is set, or data updates were not synchronized to the knowledge base.
- Symptom: Retrieval returns too many irrelevant general technical documents, and cannot locate landing page-specific point burying requirements. Cause: The
similaritythreshold is set too low, or the exclusive knowledge basecollectionIdswas not specified.
How to confirm the configuration is correct
- Upload a latest landing page plan document, initiate a query containing "landing page form fields", and check whether the retrieval results include the content of this document.
- Import an expired A/B test report, initiate a query containing "current campaign test results", and check whether the retrieval results prioritize recalling the latest documents.
- Upload an Excel-formatted form field list, initiate a query containing "extract the content of the tenth row", and check whether the corresponding field information can be returned correctly.
- Initiate a query containing "page path analysis method", and check whether the retrieval results only include documents exclusive to landing page conversion optimization, with no irrelevant content from other categories.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.