Research › Methodology

How Shipherd research is produced

Shipherd ResearchLast reviewed 25 August 2026Reviewed quarterly

What informs our figures

Shipherd operates inside its customers' systems, so the figures we publish come from operational records rather than surveys or estimates. The categories are:

• Order records across all sales channels, including marketplaces
• Marketplace transaction and settlement records
• Carrier scan events
• Carrier invoices and contracted rate cards
• Supplier invoices, purchase order history and supplier contracts
• Warehouse management system processing events
• Helpdesk tickets and their resolution history
• Media submitted as evidence in damage claims

We do not use survey data, panel data or third-party estimates in any figure we describe as measured.

Scope

Figures published as measured results are drawn from connected accounts under commercial agreement with Shipherd, covering over £40M of annual merchant revenue and the shipment, ticket and invoice records associated with it, over a rolling twelve-month period.

All accounts are multi-channel retailers selling across two or more sales channels. Figures should not be read as applying to single-channel operations or to businesses outside the £2M–£100M range. Marketplace and carrier rules differ by region, so a figure derived from one market should not be assumed to hold in another.

The denominator matters

A recovery rate is meaningless without knowing what it is a percentage of. Where we publish one, we state the base explicitly.

• Marketplace reimbursement rates are expressed against seller-fulfilled marketplace revenue, not total revenue and not total marketplace revenue. Seller-fulfilled and platform-fulfilled orders have different reimbursement mechanisms and are not comparable.
• Carrier claim rates are expressed against shipment count, and their monetary value against annual carrier spend.
• Invoice discrepancy rates are expressed against carrier spend.
• Support automation rates are expressed against total inbound ticket volume, not against tickets of a selected type.

How we label a number

Four categories. Every figure we publish carries one, and we do not mix them.

An illustrative figure is never presented as a measured one. Where a page carries a worked example with numbers in it, those numbers are labelled as illustrative even where they are realistic.

What we exclude

• Test orders
• Cancelled orders
• Internal transfers between locations
• Any metric that falls below the minimum sample for its type, set out below

Minimum sample

The threshold differs by metric, because base rates differ. Support tickets arrive in thousands a month; reimbursement claims arise on a small fraction of orders. A sample large enough to make a ticket automation rate meaningful would be an unreachable bar for a claim rate, and one number applied across both would either exclude valid findings or admit unreliable ones.

Each threshold counts records examined, not records that met the criteria. A rate of two per cent is derived from every record checked, not from the two per cent that qualified.

Where a metric cannot meet its threshold, we either do not publish it or we publish it as a single measured result with the account scope stated, rather than presenting it as a range. A range implies a distribution, and a distribution needs enough accounts to be meaningful.

What these figures do not tell you

Shipherd sees the post-purchase operation. Our figures say nothing about:

• Warehouse operational efficiency
• Pricing strategy or its effectiveness
• Marketing performance, acquisition cost or conversion
• Anything on the sell side of the business
• Product margin or category profitability

A recovery rate is a measure of leakage in fulfilment and support. It is not a measure of how well a business is run.

Limitations

These figures are observational. They describe what was found in live operations, not the result of a controlled study, and there is no comparison group of businesses running without recovery in place.

Recovery rates move with how long an account has been connected. The first period of analysis on any account includes everything still claimable in the lookback window, so an early figure reflects a backlog being cleared as well as an ongoing rate. Where that distinction matters to a published number, we state which one it is.

Category matters. A retailer shipping bulky, heavy goods has a materially different carrier cost profile from one shipping small items, and recovery rates move accordingly. Where we know a figure is category-sensitive, we say so.

Marketplace and carrier rules change. A figure that held under one policy may not hold after it is revised, which is why every published figure carries a review date.

Review

Published figures are reviewed quarterly. Where a figure changes materially, the page carrying it is updated and its last-reviewed date changes. We do not silently revise a published number.

Every proprietary figure published by Shipherd links to this page.