Methodology
We track 18 South African couriers across price, trust, and market signals. This page documents every source, refresh cadence, sample size, and limitation behind our published data.
Direct API integrations where available (live quoting from courier partners we integrate with) and structured rate-card extraction for couriers without public APIs. All published prices are VAT-inclusive at 15% South African VAT.
Pulled live from the HelloPeter consumer API. We use HelloPeter's own weighted trust index (1–5 scale) which accounts for review recency, response rate, and rating distribution. We do not modify or re-weight these scores. We also report the raw review count and industry rank (HelloPeter's position vs all SA businesses in that category) for transparency.
Google News RSS query per courier, 7-day rolling window. Mention count = unique news stories indexed by Google News mentioning that courier name in the past 7 days. We do not classify sentiment — this is a volume signal only, useful for outreach timing and competitive narrative tracking.
Stats SA bulletin endpoints for CPI (transport, freight), retail trade, and e-commerce quarterly data. These provide sector-wide signals affecting parcel volume that contextualise individual courier performance.
Google Analytics 4 for publisher-side traffic, Microsoft Clarity for heatmaps + session replay (POPIA-safe with input masking on all PII fields), PostHog for product event analytics. Aggregate metrics only — never republished at user level.
Every row we publish carries a data_source_tier label and a numerical confidence value. The four tiers below are locked — additive-only. The legacy free-text source column is retained on every row for backward compatibility. New integrations should prefer the tier vocabulary.
| Tier | Confidence | Meaning |
|---|---|---|
| Live-observed observed_live | 0.95 | Confirmed via a live courier API call at capture time. |
| Rate card rate_card_published | 0.85 | Sourced from the courier's own published rate card (public PDF or web page). |
| Model estimate estimated_model | 0.55 | DeliverAI's model-based estimate where no live API and no published rate card is available. |
| Seed / backfill seed | 0.40 | Historical backfill / seed data used to bootstrap the dataset before daily live-capture began. Retained for continuity. |
Methodology contract. Tier names and confidence values are locked. We do not rename or remove tiers — new tiers can only be added additively, and would trigger a fresh sprint with downstream consumers updated in lockstep.
Live-observed rows also carry an observed_at ISO timestamp recording the moment a courier API confirmed the price. Rows in other tiers leave this field null.
Backward compatibility. The historical source column has not been renamed, dropped, or overwritten. Consumers still reading the legacy free-text values continue to work. All 23,992 pre-sprint rows were backfilled into the new tier vocabulary; zero rows carry a null tier or confidence.
Machine consumers. The tier vocabulary is emitted verbatim under the provenance block of /api/data/dataset.json and /api/data/sa-courier-pricing-2026.json. The CSV export at /api/data/dataset.csv carries the tier and confidence as columns with a preamble comment.
Canonical URL for this vocabulary: https://deliverai.co.za/methodology
Where there is uncertainty we say so explicitly. Sample sizes, refresh dates, and source endpoints are disclosed on every dataset.
We do not modify or re-weight third-party trust scores. HelloPeter's number is HelloPeter's number; we publish it unchanged. We never average across sources without disclosing the weighting.
All published datasets are released under CC-BY 4.0. Methodology is public. Refresh cadence is documented. The pipeline that builds /insights/weekly is the same pipeline we sell to Pro Data subscribers — only the cadence and the depth of historical access differ.
Courier ranking in our quote engine is by price + transit time + customer trust. Rankings cannot be bought. Partner badges and co-marketing arrangements are disclosed separately.
No PII is stored in our analytics layer. Booking data is encrypted at rest. Customer phone numbers and addresses are never exposed to third parties beyond the chosen courier for fulfilment.
All datasets and reports are released under Creative Commons Attribution 4.0 (CC-BY 4.0). Free to share, adapt, and republish with attribution.
Suggested citation: DeliverAI, South African Courier Intelligence — [Month Year], https://deliverai.co.za/insights/weekly, accessed [YYYY-MM-DD].
Raw dataset: /api/data/sa-courier-pricing-2026.json
Contact for corrections: hello@deliverai.co.za