Two forwarders quote the same container on the same lane and return numbers that cannot be compared. Not because one is lying — because the quotation is an all-in figure and the invoice is a stack of six separately billed components, three of which no forwarder prints anywhere.
I spent the last weeks building a dataset that tries to keep those components honest: 100 corridors, 150 companies, every number carrying a basis and an error band. This post is about what that structure buys you, using Rhenus Logistics and Scan Global Logistics as the worked example — and, more usefully, where the whole approach runs out of road.
The two companies, in the fields the dataset carries
Rhenus Logistics: 3PL and forwarder, Rethmann Group, headquarters in Holzwickede, Germany. Scan Global Logistics: 3PL and forwarder, CVC Capital Partners as majority owner, headquarters in Copenhagen, Denmark.
Both records, as published, carry the same two empty fields:
schedule_reliability_vessel_basis_sep2026: null # 5 of 150 companies publish this
free_time_tariff_published: false # 4 of 150 companies publish this
ownership_status: "checked_against_public_sources" # Rhenus: plausibility check only
That null is the interesting part. A dataset that quietly omits it invites a ranking that the underlying facts do not support. In September 2026, five ocean carriers published vessel-level schedule reliability at all. A forwarder-level punctuality comparison is therefore not "unavailable for now" but structurally unbuildable, and the honest representation is a null with a reason attached, not a blank.
Every figure needs a basis and a band
The corridor records look roughly like this:
| field | example (Busan → Los Angeles) |
|---|---|
| mode / spec | ocean, 40ft FEU |
| benchmark | $8,247 |
| range | $7,422 – $9,072 |
| transit | 12–15 days |
| basis | KCCI published route, ±10% |
| as_of | 2026-11-06 |
Six columns, and the last two do most of the work. Without basis, "$3,615 Shanghai → Rotterdam" is an unfalsifiable claim; with it, a reader can go to the underlying index and disagree with me on the record. Without as_of, a quote from three weeks ago looks like a contradiction when it is just a different week.
The band is not decoration either. Shanghai → Rotterdam on 2026-11-06: $3,615, range $3,254–$3,977. That is a 22% spread around the point value. Anyone comparing two forwarders to the dollar is comparing noise; anyone comparing to the band is asking the right question — is the quote inside the published range for this lane this week, and if not, which line item explains it.
Six numbers between an index and an invoice
The pipeline above is the mental model I ended up with. An index gives you one component. A quotation compresses the first two into a single number. Then three more bills arrive: surcharges at both ends, free time, and terminal storage. Each is billed by a different party, and only the first is public for most companies.
Concretely, on Korea import the numbers are these:
- THC, Busan, 40ft: ₩240,000 (Hapag-Lloyd Asia THC, August 2026 table).
- THC, Shanghai: CNY 672 (20') / 890 (40'), effective 2024-01-15.
- THC, Rotterdam: EUR 285 any size, 2026-10-01. Hamburg: EUR 300.
- US side: Los Angeles / Long Beach $500 / $650, New York $600 / $750, from 2026-10-01, plus PierPass TMF at $40.63 per TEU.
- Carbon: Maersk EMS/ESS $170 per 40ft from 2026-10-01 on Far East → North Europe; the Q1 2026 EU ETS surcharge ran $84/65 per TEU (Maersk), $82/87 (Hapag-Lloyd), $84/67 (CMA CGM), $93/42 (ONE) on North Europe and the Mediterranean.
None of that is in the first number. All of it is in the invoice.
The rate table, normalised
This is the one table where normalisation actually changed the answer. The published Korea import ladders for four carriers look similar in prose and diverge the moment you compute a cumulative total:
- Maersk and Hapag-Lloyd: 10 days free, then ₩30,000/day (days 11–15), ₩50,000 (16–20), ₩90,000 from day 21. Cumulative: ₩120,000 by day 14, ₩490,000 by day 21.
- OOCL: 12 days free, ₩30,000 for the first seven chargeable days, ₩50,000 for the next seven, ₩85,000 after. Cumulative: ₩60,000 by day 14, ₩310,000 by day 21.
- COSCO: 12 days free, ₩30,000, then ₩50,000, then ₩60,000. Same cumulative as OOCL at day 14 and day 21.
Two days of free time plus a tier boundary produces a ₩180,000 gap by day 21 on a single 40ft container. That is the kind of number that never appears in a quotation and always appears in a dispute.
And then there is the clock nobody models:
- Busan (PNC, BPT, BNCT, PNIT, BCT): 4 free storage days.
- Incheon (ICT, SNCT): 7. Gwangyang (GWCT): 14.
Terminal storage is invoiced by the terminal operator, separately from carrier demurrage, and starts on the same discharge day. On a 10-day Maersk free time at Busan, days 5 to 10 are storage-only. If your data model treats "free time" as one field, it will be wrong for every container discharged at Busan.
How the records are assembled
Four kinds of source, and the type is stored next to the value because they age differently.
Indices (Drewry WCI for ocean, TAC for air) give a benchmark and a direction of travel, dated to the week. They are the fastest to move and the easiest to misread: a WCI point value is a composite of spot assessments, not a rate anyone can book.
Route publications (KCCI on the Korea corridors, with a stated ±10%) give a lane-level anchor with a national basis, which is what makes the Busan figures defensible rather than modelled.
Carrier tariff tables (THC, free time, demurrage ladders) are primary documents with an effective date, and they are the only source where "not published" is itself a finding. Each one is transcribed with the date printed on it, because carriers revise them mid-year and a ladder without its effective date is a rumour.
Terminal publications are the least standardised of the four. Busan's five terminals have four different effective dates for the same 4-day free storage, and Gwangyang differs from both Busan and Incheon. It is the most boring source to collect and the one that most often explains a surprise invoice.
Where a figure is missing, the record says missing and names the party that would have to publish it. That is a design decision, not an oversight.
Where this approach runs out
Three places, and they are worth stating before anyone builds on it.
Reliability cannot be aggregated from public data. Five carriers publish a vessel-basis figure. Everyone else's null is a null. No amount of modelling turns it into a number.
Lane coverage is not price. The dataset says what a corridor generally costs. It does not say which forwarder actually moves volume on it, at what contract level, with what equipment availability. Rhenus and Scan Global both look identical in the fields above, and in practice they are not identical companies.
No price ranking. The comparison page for Forto and Scan Global Logistics makes the same refusal explicit: published figures side by side, and a written note wherever nothing is published. A ranking would be more convenient and less true.
Checking your own quote in ten minutes
If you have a quotation in hand, the useful sequence is:
- Put the quote in the dataset's unit (40ft FEU per container) and compare it to the band, not the point.
- Ask for THC at both ends as separate lines, with the tariff's effective date.
- Ask for free time in writing, plus the day the clock starts.
- Ask whether terminal storage is included or passed through.
- Ask who issues the B/L — that is the contracting party.
- Ask for the same validity date on both quotes.
- Ask for carbon as its own line.
Seven questions, one email, both forwarders. It does not make the freight cheaper. It makes the comparison real, which is the only thing a dataset can honestly promise.
Links
- Rhenus Logistics profile · Scan Global Logistics profile
- Forto vs Scan Global Logistics comparison
- Busan → Los Angeles lane · 100 corridors · 150 companies
Figures checked against published carrier tariff tables and public indices up to 15 November 2026 (tariffs, terminals) and 6 November 2026 (benchmarks), compiled by MMS Consulting Group for BtoB-GoGo. Sources for any specific line on request: saasradar.sponsor@gmail.com.


Top comments (0)