DTC Support Stack Benchmark 2026

Source materials for Thunderbit Research's DTC customer-support tooling report:

Report: DTC Support Runs on Commerce, Not SaaS

Summary

This repository contains the public data outputs, scripts, report drafts, and visual assets behind a May 2026 static scan of visible customer-support widgets and vendor fingerprints on DTC brand home pages.

Headline sample:

• 1,148 DTC brand home pages in the effective static scan.
• Snapshot date: 2026-05-12 UTC.
• 397 home pages exposed at least one support widget in static HTML.
• Gorgias appeared on 267 scanned home pages, or 23.3% of the sample, with sample-echo caveats.
• Zendesk appeared on 49 home pages, or 4.3%.
• Intercom appeared on only 7 home pages, or 0.6%.
• 13 brands, or 1.1%, publicly surfaced AI customer-support language on the home page.

Repository Contents

Path	Description
process_files/out/	Parsed widget CSV, aggregate JSON, and final report check.
process_files/scripts/	Reproducible scripts for widget detection, stats, figures, report build, and final checks.
process_files/_shared/dtc_brand_pool_source/out/	DTC brand pool and shared website-detection outputs used by the support report.
figures/	Original chart outputs from the local report pipeline.
article_assets/	Web images used in the published Thunderbit article.
reports/	Published article Markdown plus earlier English, Chinese, and HTML report drafts.
social_assets/	LinkedIn editorial explainer image for the support-stack post.
source_package/	Original public process-files ZIP package uploaded with the article.

Key Files

• process_files/out/widgets.csv: one row per scanned home page, with detected support-widget and AI-chat signals.
• process_files/out/analysis_stats.json: aggregate metrics used in the report.
• process_files/out/widget_stats.json: detailed widget coverage and vendor summary.
• process_files/scripts/01_detect_widgets.py: rule-based visible-widget detector.
• process_files/scripts/02_compute_stats.py: aggregate statistics builder.
• process_files/scripts/03_make_figs.py: chart generation script.

Reproduce

The parsed outputs are included, so core stats and figures can be regenerated from the public CSV/JSON files.

python3 -m pip install -r requirements.txt
cd process_files
python3 scripts/02_compute_stats.py
python3 scripts/03_make_figs.py
python3 scripts/04_build_report_bilingual.py
python3 scripts/05_module_i_check.py

01_detect_widgets.py documents the original static detection flow. Raw HTML home-page caches are not included in this repository; the published source package focuses on structured outputs, scripts, figures, and report materials.

Method Notes

This report uses public home-page HTML signals from a DTC-heavy brand pool. The detector is intentionally transparent and rule-based.

Important caveats:

• Static HTML parsing can miss widgets injected only after JavaScript execution.
• Vendor detection should be read as a visible lower bound, not a full backend install-rate estimate.
• The DTC brand pool is Shopify-ecosystem-leaning, so vendor shares should be interpreted with sample-echo caveats.
• The report measures public storefront signals, not the full support backend.

Citation

If referencing this work, cite the published article and this repository:

• Article: https://thunderbit.com/blog/dtc-support-run-commerce
• Repository: https://github.com/thunderbit-operations/dtc-support-stack-benchmark-2026