The challenge
The client is a firm of project managers and chartered cost consultants. On projects, the client is responsible for coordinating the supply chain to prepare construction bills of quantities (BQs) estimates.
To verify and refine BQs, the client has compiled a reference catalogue of material and construction rates, extracted from historic, approved BQs. In a highly competitive business landscape the reference catalogue gives the client a critical advantage.
The client needed to continuously update the catalogue by mining BQs for benchmark rates. But as the client grew, so did the volume of BQs – and the team's capacity risked becoming a bottleneck. Meanwhile, customers were asking whether the client could help them extract similar value from their own historic data.
The client approached Hoppa to explore how AI could extend their benchmarking capabilities.
Hoppa’s solution
Hoppa designed a data mining workflow for the client, capable of automated rates extraction from Excel and PDF documents. The workflow was able to extract 200x more benchmark rates per day than a human operator.
Addressing the technical complexities of the use case required close attention to detail and collaboration between Hoppa and the client’s subject-matter-experts. This included:
Varied document structure
BQ document layout and breakdown varied considerably between projects, and even engineering disciplines. Estimates were often prepared by other suppliers, and the client had little control over how information was presented.
Hoppa’s workflow was designed to be able to handle either PDFs or Excel documents, ensuring the client could process the information as it was received from other suppliers. Hoppa developed specialist AI content extractors capable of handling non-standard input formats, combined with intermediate schema validation and automated retry mechanisms to ensure outputs conformed to the expected structure.
Long context windows
BQs can often run into the tens of thousands of rows, detailing rich supporting information alongside quantities estimates. When chunking BQs up for analysis there was a risk that quantity estimates would become separated from headings/sub-headings earlier in the document.
Hoppa developed a multi-step workflow to scan through the BQ and maintain a context log whilst isolating chunks containing estimates. This semantically dense context log allowed for workflow parallelisation of estimate schema normalisation steps – speeding up BQ analysis without impacting performance. This allowed Hoppa to tackle BQs up to 800 pages in length.
Technical jargon
BQs commonly contained technical and project-specific jargon that meant benchmark rates were unexplainable and of very limited value in the reference dataset.
Hoppa classified all rates to the Principles of Measurement International (POMI) system, an industry standard taxonomy. The client quantity surveyors could use this widely recognised taxonomy to search and find rates, irrespective of their source description. The Hoppa workflow also exported other contextual metadata such as references to construction specifications and drawings, and the type of asset the rate applied to so that downstream users could verify the suitability of the benchmark.
Customer outcomes
Through engaging Hoppa, the client opened up new strategic business opportunities:
- Slash database ingest times for near real-time rates insights
- Scale existing team without headcount constraints
- Free-up key personnel to re-focus on benchmark quality assurance
- Free-up key personnel to deliver new client-facing services for rates benchmarking
- Quickly establish rates benchmarks when expanding into new regions
Blending AI and industry expertise has ensured the client can realise near-term value return from AI investment while positioning itself as the project and cost management consultancy of the future.
As the automated rates benchmarking capability moves into scaled deployment, Hoppa and the client will continue to apply the methodology to other project / cost management and advisory use cases.
Feeling inspired?
See Hoppa in action and learn how it can make the difference to your workflows.
