Win the tenders that really matter.
We model the scoring maths so your team knows which criteria separate the winner, what a point is worth under the price-quality split, and where each draft stands against the rubric.
Stream 1 shapes the bid analysis in stream 2, which shapes the AI evaluation criteria in stream 3, which determines what data you need from streams 4 and 5. Pick a stream to see what it produces.
What did the winner actually score?
We acquire actual evaluation scores from comparable past tenders. Real criteria breakdowns, real winning margins, real price-quality trade-offs. Not projections.
Where should you spend your next hour of writing time?
Some criteria separate the winner from the field. Others cluster, with every serious bidder scoring roughly the same. We classify each criterion and calculate the marginal value of one extra point under the tender's price-quality split. Your team gets a priority table they can work from.
What would the evaluator score this draft?
We score every section of your draft against the tender's sub-criteria and tell you where marks are being lost. When the panel asks a question your draft doesn't answer well enough, you find out now, not after submission.
Are we submitting evidence or assumptions?
Evaluators know the difference between a claim and evidence. We produce site-level analysis, coverage models, and demand forecasts you can cite directly. Data the panel can verify.
Is our bid feasible?
You've proposed a fleet size, a route structure, a service frequency. We build the models that prove those numbers work, with KPIs your team can defend under questioning.
Decision quality
depends on
instrument quality.
Bidding for a scored tender is a measurement problem. Every weakness in your evidence becomes a point gap. We treat the bid like an instrument: calibrate against real scoring data, stress-test where the panel is likely to deduct, and present numbers the evaluator can verify rather than prose they have to trust.
How we engage, what we produce, and what we explicitly don't.
Timesheets reward effort. Fixed budgets let us try multiple approaches and redirect effort to whatever produces the most value for the submission.
We produce maps, scorecards, models, and gap assessments that you can cite directly in the submission. We don't write your bid.
We acquire actual evaluation scores from past tenders through Freedom of Information requests, giving you real benchmarks rather than guesswork.
Three model families rather than three prompts on the same model. When the panel disagrees on a section, that's where your bid is genuinely weak.
Decision intelligence, geospatial analysis, and applied AI. The methodology is the same regardless of domain: turn data into scoring evidence.
We've supported over 21 tenders across waste management, infrastructure, logistics, and utilities, with the largest single tender in excess of $100M. The work combines decision intelligence, geospatial analysis, and applied AI; the methodology is the same regardless of domain.
Before Second Order AI, the team operated Waste Labs (2020 – 2024), building optimisation products and AI tools for waste and recycling operators. The analytical heritage goes further back, to simulation modelling for WorleyParsons and OXFAM, including port terminal analysis, construction traffic impact studies, and fleet optimisation for municipalities.
The core methodology is grounded in peer-reviewed research carried out during prior affiliations with the Singapore University of Technology and Design and the University of Geneva, covering optimisation, simulation, and applied AI for logistics and resource planning.
We'll tell you which streams matter for your tender and which don't. If we can't move the score, we'll say so before you commit.
info@second-order.ai →