The B15‑Better project demonstrates that a targeted, evidence‑based refactor can dramatically improve both performance and usability of a mature analytics platform. By combining an asynchronous core, zero‑copy data handling, and a user‑centric UI redesign, we achieved up to 42 % higher throughput, 35 % lower latency, and substantial gains in user satisfaction—all while preserving the original system’s API and deployment model. The methodology and tools (B15‑Bench, configuration wizard) are released under an Apache‑2.0 license, inviting the community to adopt, extend, and validate the approach on other legacy systems.
| Measure | Original B15 | B15‑Better | Δ | |---------|--------------|-----------|---| | Task Completion Time | 9.4 min (± 1.3) | 5.7 min (± 0.9) | ‑39 % | | Error Rate | 23 % | 7 % | ‑16 pp | | SUS Score | 62 ± 7 | 84 ± 5 | +22 pts | | Net Promoter Score | −12 | +38 | — | bakkybksd015 15avi better
Participants highlighted the Configuration Wizard as the most helpful feature, noting that it “prevented me from making the classic ‘max‑threads = 0’ mistake”. The plug‑in UI was praised for reducing restart times when tweaking dashboards. | Measure | Original B15 | B15‑Better |
The BakkyBksD015‑15AVI (hereafter B15), a proprietary data‑streaming and visualization framework, has been adopted across several mid‑size enterprises for real‑time analytics on heterogeneous sensor feeds. Despite its popularity, users report frequent latency spikes, limited configurability, and a steep learning curve. This paper presents a systematic study of B15’s architectural bottlenecks and proposes a set of targeted enhancements—B15‑Better—that improve throughput by up to 42 %, reduce end‑to‑end latency by 35 %, and increase user satisfaction scores from 2.9 ± 0.6 to 4.3 ± 0.4 on a 5‑point Likert scale. The contributions are threefold: combined with user‑centered interface redesign
Our findings suggest that incremental architectural refactoring, combined with user‑centered interface redesign, can substantially elevate legacy analytics platforms without requiring a full system rewrite.
All performance improvements and usability metrics were statistically significant (p < 0.001). The effect size for latency reduction (Cohen’s d = 1.2) indicates a large practical impact.