TARE

clock_comparison · evaluate · results_only

tare evaluate --holdout --no_method_disclosure

Request evaluation

Pilot — evaluation-as-a-service for timing comparison

Proprietary transformation and evaluation-as-a-service for timing comparison systems. Before/after statistics are the deliverable; the engagement is evaluation-only—not a live product or performance guarantee. Patent pending (priority Feb 2026).

Program identity

CategoryEvaluation-as-a-service on timing comparison data
TransformationProprietary; applied off-line; method not disclosed
Domain todayLong-baseline GNSS clock comparison (IGS-class products, agreed baselines)
DeliverablePost-processed series + before/after statistics + written evaluation pack

Other timing domains may follow; this pilot envelope is GNSS clock-comparison only.

Evaluation frame (definitions only)

What is transformed: an agreed primary clock-comparison or residual series from your submission (definition fixed at onboarding or in the pilot agreement)—we return post-processed outputs and statistics; we do not disclose internal processing.

What you compare to: Before = same stream, same metric (typically RMS on that primary series), before our confidential post-process. After = identical metric on what we return. Same inputs, same metric definition—no moved goalposts.

What “% improvement” means: relative reduction in that agreed primary statistic from before to after on your panel; where we cite published holdout runs, no recalibration on the evaluation window (see technical note).

What Tare returns (form only, no mechanics): Evaluation outputs—post-processed time-aligned traces (series/residuals) and a written pack; summary statistics (principally RMS, plus any metrics named in onboarding); optional review-level indicators where specified—not formal confidence/uncertainty statements unless explicitly included in your agreement.

Published runs vs your pilot

Published runs are fixed, public IGS-class comparison panels—pinned windows, null tests, evidence room. Your pilot is the same evaluation protocol on your agreed inputs, private under agreement. Public statistics show what we observed on those panels; your pilot measures whether similar movement appears on yours. Not a separate product line: published runs are the open slice of the same service. Representative of how we evaluate and what we have seen in this domain—not a uniform gain on every dataset.

Published holdout results (orientation only)

On IGS-class GNSS clock-comparison panels we have published, held-out evaluation reported typical RMS reduction around ~45%, with strict-window examples up to ~81% and post-process residuals commonly in the 17–22 ns range. Those figures come from specific years, analysis centers, and windows—not a uniform product curve across every dataset.

Pilot terms (procurement)

Commercial form: 90-day invite-only pilot—not open API, not live feed, not method license. You send comparison data under agreement; we return post-processed outputs and before/after metrics. No formulas, parameters, or code are disclosed.

The pilot is invite only (by application). Pilot fee: $45,000 (academic) or $195,000 (commercial) for the 90-day period. Scope limits, data handling, and ownership: FAQ · Pilot terms.

← Home