The engineering behind TaxMind | UAE
About 136 function points per person per month, against an industry median of about 14 (ISBSG 2023). Here’s how the platform was built, and how we measured it.
← Back to the Nexdigm storyKickoff to production, June to September
Most of the calendar went on tax decisions and client testing, not coding. Nexdigm’s tax team was running full test rounds within six weeks of kickoff.
- 18 JunKickoffscope, tax rules and the 30 Sep deadline
- 5 JulFirst commitafter two weeks on the spec
- 16 JulFirst PR merged704 merged by mid-September
- Within 6 wksClient testing startsfull test rounds
- 14 SepIn production12.5 weeks after kickoff
Every change went through the same five steps
As the build went on, we mapped about 1,280 pages of tax-authority guidance and worked through 500+ requirement rows, then wrote the tests from that guidance.
- 01Ask before building
Anything unclear went back to the person who asked, in the same thread, with a suggested answer. CraftHQ never guesses a tax rule.
- 02Impact brief
Before building, we listed every screen, endpoint, shared rule and user role the change would touch, so side effects showed up before the build, not after.
- 03Risk tier
Each change was tagged UI-only, logic or core. Anything touching money or tax is always core.
- 04Build and test
A separate step wrote the tests from the requirement. One failing test on a money or tax path stops the work.
- 05Three checks, then a person merges
Two local checks and a SonarCloud scan run on every pull request. Each one carries its requirement, its tests and the results, and a person approves every merge.
What the numbers say
The platform is about 780 function points (IFPUG, ±15%). At the industry median, that’s roughly nine months for a six-person team. Two people did it in 12.5 weeks.
Weeks start Monday. Source: git history and GitHub.
Why so fast? People spend their time on decisions. CraftHQ’s agents build, test and check their own work, and only come back to a person when a decision is needed. Most of the thinking is done up front.
Fast, and built properly
After go-live, we measured the build from its own code history. Speed only matters if the work is right, so here’s what the checks found.
14,600+ automated tests
Running on every change; 47% of all shipped code is test code.
Fewer than 1 in 500 changes undone
0.19% of changes were reverted over a 30-day window.
SonarCloud A / A / A
Security, Reliability and Maintainability all rated A by an independent scanner.
Three checks on every pull request
Two local checks, then a SonarCloud scan, before a person merges.
Data protection & ISO/IEC 27001
A data-protection impact assessment (DPIA) was completed; data is kept in the UAE; BetaCraft is ISO/IEC 27001:2022 certified.
Owned by Nexdigm
Code, tests and documentation live as plain files in Nexdigm’s repository, with no dependency on CraftHQ.
An independent third-party penetration test before go-live found no high or critical issues, and every finding was closed.
What CraftHQ did
- ✓Mapped the tax authority’s guidance, about 1,280 pages, and worked through 500+ requirement rows.
- ✓Built the whole filing journey: onboarding, document requests, statement mapping, an 8-step return, tax calculated to the last fil, the advisory memo, two-level review and a 7-year archive.
- ✓Wrote the tests from the requirement in a separate step, and ran three checks on every pull request.
- ✓Kept a person on every merge. The engineer reviewed and approved each pull request.
- ✓Left everything with Nexdigm: code, tests, documentation and the project’s rule book, as plain files in their own repository, with no dependency on CraftHQ.
How we measured
Function points (IFPUG) counted from the project’s own code history, both people’s time and the requirements work included. Platform size is an estimate (±15%), so the six-person equivalent is 8–11 months. Industry median: ISBSG 2023, 7.9 hours per function point (about 14 per person-month at an assumed 110 productive hours a month). Our figure: ~780 FP ÷ (2 people × ~2.87 months) ≈ 136 per person-month. Figures describe this build, not a forecast. The ISO certificate and the security-test summary are available under NDA.
Could your team work like this?
Start with a free baseline report on one repository (we only need read access). Or scope a 3-month pilot at early-adopter pricing for teams of up to 20, with the output measured against what your team’s work costs today.