Test Results YourValidator 1.0.2
How the download edition was tested and what came out of it: automated tests, counter-checks for every fixed defect, and an independent external review.
Download edition 1.0.2 · German and English · complete file and modular edition · as of November 2, 2026
The package includes a test suite that operates the tool in the browser the way a person would: entering values, creating channels, saving, loading, exporting. It runs in all four forms and compares the complete file with the modular edition. Buyers can run it themselves.
| Area | DE | EN |
|---|---|---|
| Calculation: budget, goal back-calculation, simulation | 22 | 22 |
| Halo, saturation, ramp-up curve, seasonality, plans, saving, PDF, performance | 112 | 112 |
| Campaigns, stress tests, texts | 34 | 38 |
| Benchmarks, CRM, custom channels, random calculations | 37 | 41 |
| End-to-end flow, uploads, header area | 103 | 104 |
| Network isolation, origin of values, numbers, currency | 33 | 33 |
| Use without a licence, basic functions, console | 16 | 16 |
| AI functions: analysis, interpretation, budget proposal, checking the responses | 66 | 66 |
| Reference values from the recalculation, one test per finding, field-by-field round trip | 118 | 117 |
| Security: injected code, encryption, network requests | 35 | 35 |
| Total per form | 576 | 584 |
| Form comparison complete file = modular edition (per run) | 2 | 2 |
| Per full run (both forms + form comparison: 2 × 576 + 2 and 2 × 584 + 2) | 1,154 | 1,170 |
A full run only counts as passed if every block matches the expected figures exactly, without a single test that only passes on a second attempt. Acceptance was done from the delivery files in an empty folder, following the instructions in the package exactly.
Firefox and WebKit
In addition, the full suite ran once in Firefox and WebKit. This revealed three findings that were only visible there (including a button that swallowed clicks in WebKit); they are fixed in 1.0.2.
| Browser | passed DE / EN | not green DE / EN | Classification |
|---|---|---|---|
| Firefox 150 | 1,140 / 1,156 | 2 / 2 | Test data generated in Firefox does not fit into storage; the tool correctly reports "too large". Not a defect in the tool. |
| WebKit 26.4 | 1,136 / 1,152 | 6 / 6 | Images differ by single pixels (±1) after redrawing; the tests demand byte identity. Not a defect in the tool. |
Per browser and language, 12 tests that are only possible in Chromium were skipped (PDF printing, memory and listener measurement). WebKit is the engine of Safari, but Safari itself was not tested.
A test is only worth as much as its ability to find a defect. That is why every test for a fixed finding was counter-checked:
- Against the old state. Every new test ran on the state before the fix and had to fail there – in all four forms. The only exceptions are tests that deliberately secure existing correct behaviour; they must pass on both states.
- With falsified expectations. In a copy, one expected number or message per test was deliberately set wrong. Every test falsified this way had to fail. This rules out a test that always passes regardless of the result.
- Against real defects. Where tests measure stability, such as memory and event listeners, an artificial leak was built in. The test had to find it.
Separately from the test suite, the tool was reviewed from the outside: in a separate environment, with its own tools and its own, independently programmed calculation model that takes no code from the tool.
Recalculation
Random, valid inputs entered into the tool, all result values read from the user interface and compared with the independent calculation model – simple scenarios and combined ones with saturation, ramp-up curve and halo.
Random test
Random sequences of actions with empty, negative, huge and decimal values. After every fifth step, fixed rules are checked, such as revenue = sum of the channels and chart = figure.
Field-by-field round trip
Random values in all 269 fields, switches and selects, then reload, backup into a fresh browser, saving and loading a plan, and “Delete saved data”.
Boundary values
Each of the 256 number fields with 14 values: empty, 0, negative, decimal, up to 1015 and 100 times the maximum.
Coverage
It was measured which of the 359 functions the checks pass through. The areas not reached until then were checked in a targeted way: AI responses, encryption, deletion, imports.
Security
Injected HTML and scripts via channel names, backups and CRM files; foreign and damaged files; operation behind a Content Security Policy; encrypted backups with a wrong and the right password.
Code provenance
Comparison with public open-source collections (only anonymous fingerprints transmitted) and a local licence scan. The libraries Chart.js and d3 compared with the official files by checksum.
Delivery
Checksums of the zips and of every single file; comparison of the delivery files with the reviewed state.
The delivered edition 1.0.2 was reviewed again from the outside in all four forms.
| Check | Result |
|---|---|
| Idle: changes to the button "Confirm AI inputs" in 2 s without input | ✓ 1.0.1: 156 per form · 1.0.2: 0 per form |
| Recalculation: 240 simple and 120 combined scenarios | ✓ 0 deviations |
| Random test: 4 × 400 user actions | ✓ no rule violated, no hang |
| Field by field: reload, backup, plan, delete in 4 forms | ✓ all 269 fields and results as expected |
| Stress test DE and EN: 3,000 inputs, 500 recalculations, 150 channels, extreme values | ✓ results identical, no NaN; 150 channels 0.12 s per recalculation |
| Boundary values: 256 number fields × 14 values | ✓ no NaN, no slow calculation |
| Custom channels: 42 twin scenarios | ✓ 42 / 42 exactly identical |
| Console errors / outbound network requests | ✓ 0 / 0 |
Before 1.0.2, the modular edition 1.0.1 of both languages was tested under extreme conditions. The calculation is unchanged in 1.0.2.
| Check | German | English |
|---|---|---|
| Input storm: 3,000 inputs without pause, including empty, negative and up to 1012 | ✓ result identical to a fresh page; memory 4.1 → 6.3 MB | ✓ result identical to a fresh page; memory 4.2 → 6.4 MB |
| 500 recalculations in a row | ✓ median 13 ms, max 42 ms; no memory growth | ✓ median 15 ms, max 37 ms; no memory growth |
| 150 custom channels, all active | ✓ recalculation 0.16 s; sensitivity analysis 15 s | ✓ recalculation 0.15 s; sensitivity analysis 13 s |
| 150 channels: backup and import into a fresh browser (427 KB) | ✓ all 150 channels, identical result; import 4.1 s | ✓ all 150 channels, identical result; import 4.1 s |
| Delete all 150 channels | ✓ result exactly as at the start again | ✓ result exactly as at the start again |
| All fields at once: 100× the maximum / −1 billion / empty | ✓ no NaN, calculation under 30 ms; empty = defaults | ✓ no NaN, calculation under 31 ms; empty = defaults |
| Recalculation: 150 simple and 75 combined scenarios | ✓ 0 deviations | ✓ 0 deviations |
| Random test: 4 × 500 user actions | ✓ no rule violated, no hang | ✓ no rule violated, no hang |
| Boundary values: 256 number fields × 14 values | ✓ no NaN, no slow calculation | ✓ no NaN, no slow calculation |
| Field by field: reload, backup, plan, delete (3 random states) | ✓ all 269 fields and results unchanged | ✓ all 269 fields and results unchanged |
| Custom channels: self-created channel against built-in channel with identical values, 70 random scenarios (all channel types, ramp-up curve, order value, saturation, campaign, halo) | ✓ 70 / 70 exactly identical** | ✓ 70 / 70 exactly identical** |
| Backup exchange between DE and EN | ✓ all numbers and fields identical* | ✓ all numbers and fields identical* |
| Console errors / outbound network requests | ✓ 0 / 0 | ✓ 0 / 0 |
* The customer label (e.g. “Mitglieder” / “Members”) and the currency pound sterling, which only the English edition offers, are set to the default of the other edition when switching language. The calculation does not change.
** Custom channels use the same formulas as the built-in ones. They start with neutral defaults; custom referral channels take the number of referrals directly instead of own customers × rate.
All findings from the manufacturer's testing and the external review were fixed, each with its own test and counter-check. From finding 15 on, none concerned the main calculation with valid inputs any more, only edge cases in inputs, limits, imports and saving.
| File | SHA-256 |
|---|---|
| YourValidator_tool_deutsch_1.0.2.zip | 2c92e7b0f7b96f8e082d4a0530b21c9b9e3bd394c271e4c50a3f17702eeee1ae |
| YourValidator_tool_english_1.0.2.zip | d3504e09319316681ef35071eb23c70501a7e76fb0c9331faae7668b20789e4b |
With the checksum, anyone can verify that they received exactly the reviewed edition.
- The full suite ran in Chromium (Chrome, Edge), and once also in Firefox and WebKit. Safari itself was not tested. Real PDF printing is only tested in Chromium.
- The AI functions were tested without real API keys, using prepared responses. The real responses of the AI providers are not part of the review.
- A review of this scope shows that no errors occur any more in the areas and cases checked. No review can prove that software is completely free of errors.
- The results apply to download edition 1.0.2 with the checksums listed above.
The independent external review was carried out by Claude, an AI model by Anthropic, on behalf of YourValidator.de. It does not replace an audit by an accredited body and is not an endorsement or confirmation by Anthropic.
