How fast can Orbid process a large tender?

Orbid completes semantic matching on a 150-row tender in approximately 46 seconds. The full end-to-end cycle — parsing, matching, compliance checking, and draft generation — typically finishes in under 60 seconds for tenders of this size. Processing time scales roughly linearly: a 300-row tender takes approximately 90 seconds, while smaller tenders under 50 rows often complete in under 15 seconds.

What accuracy does Orbid achieve on spec matching?

Orbid achieves 97% accuracy on semantic spec matching with a false-positive rate of just 0.3%. This means that out of 150 line items, you can expect roughly 4–5 items flagged for human review rather than incorrectly matched. The system is deliberately calibrated to err on the side of caution — it will flag an uncertain match rather than commit a false positive, because in medical device procurement a wrong match can mean disqualification.

What file formats does Orbid support for tender documents?

Orbid supports XLSX (including multi-tab workbooks), PDF, DOCX, and CSV tender documents. For XLSX files, Orbid intelligently parses multiple tabs — identifying which tabs contain requirement specifications, which contain scoring criteria, and which contain administrative instructions. PDF support includes both native text PDFs and scanned documents via OCR. Orbid also handles ZIP archives containing mixed document types, which is common in European public procurement portals.

← Journal

AIProcurement

How Orbid Processes a 150-Row Tender in Under 60 Seconds

17 June 2026

It is Friday morning. Your inbox contains a new hospital tender notification: 150 line items for diagnostic ultrasound systems and accessories, multi-tab XLSX workbook, two PDF attachments with clinical specifications, and a submission deadline in five business days. Under the old workflow, your bid team would spend three to four days just reading, cross-referencing, and filling in the response template. With Orbid, the mechanical work is done before the first coffee gets cold.

This walkthrough shows exactly what happens inside Orbid when it processes a real-world 150-row ultrasound tender — step by step, with the actual numbers.

Step 1: READ — Parsing the tender documents

The moment you upload the tender package, Orbid begins its READ phase. This is not simple text extraction — it is structural parsing that understands procurement document conventions.

For the XLSX workbook, Orbid identifies which tabs contain the actual requirement specifications (often labelled "Lot 1 Technical", "Specifications", or similar), which tabs contain evaluation criteria and scoring weights, and which tabs contain administrative or contractual instructions that are irrelevant to product matching. In this ultrasound tender, the workbook has six tabs: Orbid correctly identifies Tab 2 ("Technical Requirements") and Tab 4 ("Clinical Specifications") as the primary matching targets.

The two PDF attachments are parsed simultaneously. One contains detailed clinical workflow requirements for obstetric and cardiac imaging modes. The other contains installation and training requirements. Orbid extracts structured requirements from both, cross-referencing them against the XLSX rows to build a unified requirement matrix.

The output of the READ phase is a normalised requirements list: 150 line items, each with a requirement description, requirement type (mandatory vs. scored), any referenced standards or certifications, and the evaluation weight if specified. This phase completes in approximately 8 seconds.

Step 2: MATCH — Semantic matching against your product catalogue

The MATCH phase is where Orbid's AI capabilities are most visible. Each of the 150 requirements is semantically matched against your product catalogue — not by keyword lookup, but by understanding what the requirement actually asks for and whether your product genuinely meets it.

Consider a typical requirement: "Real-time tissue harmonic imaging with at least 3 selectable frequency ranges for abdominal applications." A keyword search might match any product mentioning "harmonic imaging." Orbid understands that this requirement has three distinct components — real-time capability, minimum three frequency ranges, and abdominal application suitability — and verifies all three against your product specifications.

Each match receives a confidence score. Scores above 95% are auto-matched. Scores between 75% and 95% are flagged for human review with an explanation of what triggered the uncertainty. Scores below 75% are marked as potential non-compliances, with a suggested alternative product if one exists in your catalogue.

For this 150-row tender, the MATCH phase completes in 46 seconds. Results: 138 items auto-matched at greater than 95% confidence, 9 items flagged for review, 3 items identified as non-compliant with suggested alternatives. That is 97% accuracy with a 0.3% false-positive rate.

Step 3: COMPLY — Regulatory and certification verification

Medical device tenders are not just about product specifications. They require proof of regulatory compliance — and a single expired certificate or missing clearance can disqualify an entire submission.

Orbid's COMPLY phase cross-checks every matched product against its regulatory status database:

CE marking under EU MDR 2017/745: Verifies that the CE certificate is valid, that the Notified Body designation is active, and that the certificate scope covers the specific device models referenced in the tender response.
FDA 510(k) or PMA status: For tenders requiring US market clearance, verifies clearance status, predicate device chain, and any pending supplemental submissions.
MHRA UKCA status: For UK market tenders, checks UKCA marking or MHRA registered MDD-era CE certificates under transitional provisions.
Certificate expiry monitoring: Flags any certificates expiring within the tender contract period, so your team can address renewal timelines in the submission narrative.
Evidence chain assembly: For each compliance claim, Orbid assembles the documentary evidence — certificate PDFs, declaration of conformity references, EUDAMED registration links — ready for inclusion in the submission package.

The COMPLY phase runs in parallel with the tail end of MATCH, adding approximately 6 seconds to the total processing time.

Step 4: DRAFT — Generating the submission-ready response

The final phase is DRAFT. Orbid generates the tender response in the buyer's own template format — populating the original XLSX workbook or generating a structured document that mirrors the buyer's required layout.

For each line item, the draft includes: the matched product with model number and catalogue reference, a compliance statement mapped to the requirement language, the confidence level, and references to supporting evidence documents. Scored requirements include suggested response narratives optimised for the evaluation criteria.

The DRAFT phase completes in approximately 4 seconds. Total elapsed time from upload to draft: under 60 seconds.

The numbers

Here is the performance summary for this 150-row ultrasound tender:

READ phase: 8 seconds — 3 documents parsed, 6 XLSX tabs analysed, 150 requirements normalised
MATCH phase: 46 seconds — 150 items matched, 97% auto-match rate, 0.3% false-positive rate
COMPLY phase: 6 seconds (parallel) — CE/FDA/MHRA status verified, 2 certificate expiry warnings raised
DRAFT phase: 4 seconds — submission-ready response generated in buyer's template format
Total: 58 seconds end-to-end

Compare this to the typical manual process: 3–4 days of analyst time for a tender of this complexity, with significantly higher error rates on specification matching and compliance verification.

What your team does after Orbid finishes

Orbid does not replace your bid team — it removes the mechanical burden so they can focus on what actually wins tenders.

After Orbid delivers the draft, your team reviews the 9 flagged items (roughly 15–20 minutes of expert review), addresses the 3 non-compliant items with strategic decisions about alternative products or exception narratives, adds the strategic commentary and competitive positioning that differentiates your submission, and reviews the pricing strategy against the tender's evaluation weighting.

The total human time drops from 3–4 days to approximately 2–3 hours of high-value strategic work. Your team spends their time on judgement calls and competitive narrative — not on cross-referencing spreadsheet rows against product catalogues.