Construction tech · Full Stack / SaaS · Case study

Nova Insight: construction schedule-intelligence SaaS, with 5 formats normalised into one schema.

Planners get a baseline-vs-update comparison in minutes instead of days, and every number shows where it came from. The same pipeline now onboards new companies as SaaS tenants.

  • ClientNova Insight, Denmark
  • IndustryConstruction tech
  • Timeline10 weeks
  • Type of buildMulti-tenant SaaS
Project

What the project was about.

The challenge

Planners compared baseline and updated schedules by hand across Excel, PDF and MS Project exports. Every file format broke the tool, and a single comparison took days.

What we built

We built a schedule-intelligence SaaS for Nova Insight, with a universal schedule format pipeline: Detector → Extractor → Recognizer (heuristics + Claude fallback) → Normalizer → Validator.

How it works

1. Upload (portal): the schedule file. 2. Detect (Python): format and structure. 3. Extract (Azure Document Intelligence): tasks and dates. 4. Validate (rules): no silent corruption. 5. Analyse (Nova engine): delay and variance.

What was delivered

Universal schedule format pipeline: Detector → Extractor → Recognizer (heuristics + Claude fallback) → Normalizer → Validator
Parses Excel, CSV, MS Project .mpp and scanned PDFs via Azure Document Intelligence, with provenance and a confidence score per value
Golden-file test suite and a zero-silent-corruption rule: unclear values are flagged, never guessed
Turned the app into multi-tenant SaaS: per-company subdomain, isolated data, admin onboarding; 3 tenants live
Tools

What it was built with.

Each tool has one clear job in the system.

PythonDetects formats and runs the pipeline
FastAPIThe API behind the portal
PostgreSQLSchedules, tenants and results
RedisJob queue and caching
Azure OpenAIAI analysis of schedule data
Azure Document IntelligenceReads scanned PDFs: tasks and dates
ClaudeFallback recognizer for unclear layouts
Timeline

How the 10 weeks went.

  1. Start

    Scope

    We mapped the schedule formats planners use and the baseline-vs-update comparison they needed.

    Agreed scope
  2. Build

    Build

    The format pipeline, the portal and the Nova analysis engine were built and shown in regular demos.

    Working pipeline
  3. Before launch

    Test

    A golden-file test suite checked every format, so unclear values are flagged and never guessed.

    Tested parsing
  4. Week 10

    Launch

    The app went live as multi-tenant SaaS with a subdomain and isolated data per company.

    3 tenants live
Outcome

What changed.

Measured results from the project.

99.4%Excel parse accuracy
96.1%PDF parse accuracy
10 minBaseline vs update comparison
10 wksFrom kickoff to live

Excel parse accuracy

85%
99.4%
BeforeAfter

PDF parse accuracy

70%
96.1%
BeforeAfter

Baseline vs update comparison

2 days
10 min
BeforeAfter
Book a call

Want a result like this?

Tell us which case is closest to what you need and you get a fixed-scope quote within 24 hours.

What you walk away with:

  • A mapped funnelWhere your leads arrive and where they are lost.
  • A recommended buildMatched to your channels and tools, like the case above.
  • A fixed-scope quote within 24 hoursYours to keep, whether we work together or not.
  1. 1Book a free call

    Pick a time that suits you. The call takes 30 minutes.

  2. 2We map it and quote

    We map your funnel and send a fixed-scope quote within 24 hours.

  3. 3We build and launch

    Once you approve, we build it, test it and launch it.

Pick a time that suits you

Choose a day and time for your free 30-minute call. It is booked straight into our calendar.

Your details are used only to book this call. No spam, ever.

Client Reviews

5.0 average rating on Upwork and Fiverr