Methodology

Data is collected through free calculator tools, not surveys. Every input is a practice-level financial or operational aggregate drawn from practice management software, accounting records, or HR systems. No patient-level or patient-identifiable data is ever collected, transmitted, or stored.

Inputs are stored as raw values submitted by the practice owner or administrator. All ratios and derived metrics — such as overhead rate, hygiene capture rate, or days in AR — are computed server-side from those raw inputs. Practices never submit a ratio directly; they submit the underlying components, and the index derives the metric.

Every submission passes a five-step pipeline before it is eligible for inclusion in benchmark aggregates. First, each derived metric is checked against published valid range bounds — values outside plausible operating ranges are flagged. Second, cross-field logic checks verify internal consistency: for example, net collections cannot exceed gross production by more than 10%, and claimed hygiene hours cannot exceed total available chair hours.

Third, plausibility heuristics flag statistically unusual combinations — such as a collection rate above 100% paired with unusually low AR. Fourth, duplicate detection checks for submissions that are suspiciously similar to a prior submission from the same practice in the same period. Finally, a composite quality score from 0 to 100 is assigned. Submissions scoring below 60 are held in a review queue and excluded from aggregation until manually cleared.

Benchmarks are computed using PostgreSQL's PERCENTILE_CONT function, which performs linear interpolation between values to produce smooth percentile estimates at P10, P25, P50, P75, and P90. Before statistics are computed, values are winsorized at the 2nd and 98th percentile to reduce the influence of extreme outliers on the distribution.

Each practice contributes at most one value per metric per period. When a practice submits multiple times in the same quarter, the most recent validated submission is used. Aggregation runs nightly for live benchmark cells and is frozen at the end of each calendar quarter for publication in the Index.

No benchmark cell is published with fewer than 10 distinct practices from 5 or more distinct organizations. This threshold ensures that no individual practice's data can be reverse-engineered from a published percentile. If a user's exact segment falls below this threshold, the comparison group automatically falls back to a broader segment — for example, from a state-level peer group to a national one. The comparison group is always labeled on the scorecard so users know exactly what they are being compared against.

Practice identifiers and user identifiers are never included in aggregation queries. The aggregation engine operates on validated metric values only, with no join to practice names, locations, or user accounts.

At launch, benchmarks are seeded from DSI composite estimates drawn from published industry sources, DSI consulting experience, and publicly available dental industry reports. Seed data is clearly labeled with a "Seed" badge wherever it appears in the index and on scorecards. Seed benchmarks represent our best estimates of national medians for general practices with 4–6 dentists; they are not derived from live practice submissions.

As live submissions accumulate, individual benchmark cells flip automatically from seed to live once the n≥10 threshold is reached and the cell passes a consistency check against adjacent percentiles. The transition from seed to live is logged and does not retroactively change previously issued scorecards.

Live benchmark cells refresh nightly based on validated submissions received since the last aggregation run. These updates are reflected immediately in new scorecard calculations but do not change the quarterly Index snapshot.

Quarterly: at the close of each calendar quarter, a full aggregation freeze runs, producing an immutable snapshot of all benchmark values for that period. This snapshot is published in the DSI Benchmark Index report, accompanied by editorial commentary generated by the DSI team. Once frozen and published, quarterly aggregates are never modified — they serve as the permanent record for that period.