DuskByte
Real Estate & PropTech · LegalTechFull case study

FreeTaxProtest

Property tax protest infrastructure for 90,000+ homeowners

Problem

Chandler Crouch, a licensed Texas property-tax consultant, wanted to offer free tax protests to homeowners over-assessed by county appraisal districts, removing the friction of a process built on a fixed annual deadline, valuation evidence most owners can't assemble themselves, and government portals never built for the public. The constraint was scale: doing by software what a consultant normally does by hand for a few dozen clients, for tens of thousands.

Solution

We built the system that runs the entire protest pipeline: an authorization lifecycle that tracks agent sign-off through county matching and routes exceptions (bad signatures, ownership mismatches, county-side errors) to the right resolution instead of a generic rejection; a regression-based valuation engine over MLS/IDX/RESO and county data that estimates fair market value against the assessment; an ML risk-scoring model that prioritizes which protests have the most to gain; and Puppeteer-based browser automation that files and tracks protests directly against county appraisal-district portals with no public API. Ticketing, approval workflow, and triggered homeowner communications keep tens of thousands of cases moving without manual tracking, on Dockerized services running on Kubernetes built to absorb millions of requests a month during protest season and sit quiet the rest of the year.

FreeTaxProtest feature map: user registration and property management, digital consent signing, document upload, automated data analysis, comparative market analysis, property value evaluation, protest hearing, and automated package generation, with homeowners saving 20-30% on assessments.
The protest pipeline end to end: registration, consent, valuation evidence, and automated filing packages.
The defensible core here is orchestration at scale and an auditable valuation engine: deterministic systems and trained models doing work an LLM shouldn't. Machine learning drives the valuation and prioritization; Puppeteer handles the portal automation. We name this plainly because in this domain a confident wrong answer costs a homeowner real money, and credibility comes from putting each tool where it actually belongs.

Outcome

The platform represents 90,000+ homeowners through a single annual filing cycle, cut manual appraisal research time by 75%, and drove a 150% increase in qualified leads after launch, with zero unplanned downtime through peak season. It's also the anchor of an ongoing relationship: what started as one engagement grew into the CRM and appraisal-automation infrastructure covered in the ChandlerCrouch and StartDeck case studies.

From the client

They delivered a usable product that performed almost flawlessly in every component. The site has been tested with high volume of user interaction. Users have rated our website with top level ratings. I have never missed a deadline with this company.
Chandler Crouch, Founder & CEO, Chandler Crouch Realtors. Clutch-verified, 5.0/5.0

Results for FreeTaxProtest

  • 90,000+ homeowners represented per cycle
  • 75% less manual valuation research
  • +150% qualified leads after launch
  • 100% uptime at peak-season scale

Tech & integrations

Puppeteerscikit-learnpandasIDX/MLS/RESOLaravelPythonVue.jsTypeScriptMySQLAWS (RDS, S3, EC2)DockerKubernetesSendGrid

Beyond this client

The transferable pattern here goes beyond real estate: browser automation against a portal with no API, ML scoring on domain data, triggered multi-channel communication, and a deadline-bound state machine that never drops a case. The same architecture applies to insurance verification, legal filing, compliance workflows, and any process that runs through a locked government or vendor portal at volume. If the system has no API, a hard deadline, and thousands of cases, that's the work we do.

Have something like this to build?

Book a call and tell us where you're stuck. We'll tell you honestly what it takes.