Measurement framework
Which import operations KPIs can you trust?
Trust an import KPI only when its numerator, denominator, time boundary, source, owner, completeness rule, and path back to shipment-level evidence are defined. Tyllus can report and export from recorded shipment, invoice, cost, supplier, lane, container, and status data; a polished chart cannot repair missing, late, or differently defined inputs.
By Ayhan Karaca, Co-Founder · Updated: August 27, 2026
A KPI is a measurement contract, not a colored tile
Before naming a target, define the population and decision. 'Document readiness' could mean all required records received, all received files reviewed, or all broker-required evidence accepted by a cutoff. Those produce different numbers. 'Average cost per shipment' is also incomplete until currency treatment, included cost types, shipment state, and period are fixed.
CBP's FY2024 scale—38.4 million entries, $3.37 trillion of imports, and more than $88 billion in duties, taxes, and fees—shows why definitions matter across trade data. It does not provide a performance target for an individual importer.
| KPI family | Required definition | Tyllus evidence path | Critical boundary |
|---|---|---|---|
| Shipment portfolio | Active, archived, completed, period and creation/status rule | Shipment counts and underlying records | State discipline changes the result |
| Document readiness | Required record family, accepted state and cutoff | Shipment documents and review state | Received is not automatically accepted |
| Invoice and cost | Currency, cost categories, inclusion and date basis | Invoice rows, cost lines and shipment drill-through | Do not sum unlike currencies |
| Supplier and lane | Identity normalization, lane definition and population | Supplier spend, shipment count and trade-lane records | Mix and volume can explain movement |
| Exception response | Trigger, assigned-at, closed-at, owner and excluded pauses | Alert or exception linked to operational record | Late external data can move the clock |
| Data completeness | Required fields and eligible population | Missing values by shipment and source | Publish coverage beside the outcome |
Use three layers: coverage, timing, and outcome
Coverage asks whether the evidence required for a decision exists. Timing asks how long a defined state transition took. Outcome asks what happened to cost, service, compliance review, or workload. Report them together: an apparent outcome improvement is weak if half the source dates or invoices are missing.
The World Bank's Container Port Performance Index compares vessel time in port and explicitly notes that external factors affect results and rankings do not assign blame. Apply the same discipline internally: a port benchmark is context, not root-cause proof for document readiness, cargo dwell, inland execution, or one shipment.
Read the KPI design steps
Name the decision. Define the eligible population. Specify numerator, denominator, time and currency rules. Identify source records and owners. Publish completeness and known exclusions. Drill into outliers, review the operational cause, and version any definition change.
Why external statistics should stay context—not your target
The FMC reports about $15.4 billion collected in detention and demurrage across nine reporting carriers from April 1, 2020 through March 31, 2025. For Q1 2025 it reports billings down 24%, collections down 19%, and waivers up 7% quarter over quarter. Those aggregates can motivate a disciplined container-charge evidence record; they cannot forecast your charges, prove avoidability, or set a savings target.
Use external data to frame a question, then answer it from your population. For example: what portion of charged containers has free-time evidence, availability and return events, invoice dates, dispute status, and final outcome linked to a shipment? The denominator and coverage rate are as important as the amount.
- Label authority, reporting population, period, and publication date beside every external benchmark.
- Keep descriptive movement separate from causation and forecast.
- Never convert an agency total into assumed avoidable spend or Tyllus ROI.
- Compare like currency, lane, supplier, shipment state, and time window.
- Investigate record-level outliers before setting policy from an average.
What reporting does Tyllus support?
From recorded workspace data, Tyllus supports shipment counts by state, invoice counts and rows, currency-separated cost and spend totals, freight and duty categories, average cost per shipment, supplier analysis, trade-lane analysis, container and entry counts, trends, alerts, and report exports. Filters and drill-through references help connect a summary to operational records.
The report reflects what the workspace contains. Missing documents, late partner responses, unrecorded costs, inconsistent supplier names, and stale shipment states can change the interpretation. Tyllus surfaces a basis for review; it does not make incomplete source data complete or establish the right business target for you.
Which dashboard review should happen every week?
Review definition changes, data coverage, largest outliers, new exceptions, unresolved owners, and reconciliation differences before discussing targets. Sample a few records from each headline KPI and prove that another authorized user can reproduce the value from its source evidence.
CBP says targeted ACE reports can support internal audits and identification of systemic errors. ACE is the government trade system; Tyllus does not replace it. The transferable principle is that reporting should expose recurring control failures and lead to the record where corrective work is owned.
Bring one disputed metric and its source records.
Tyllus can help map the definition, fields, owners, completeness rule, and shipment drill-through needed to turn a recurring reporting argument into a reviewable measurement contract.