SOW 2.1 to 2.3, 3.1.4, 4.2, 4.6 and 5.8 · methodology

The same benchmark, three ways, so the County chooses with the numbers in front of it

The Statement of Work asks for several compensation methods, including unadjusted salaries and regional or geographic adjustments, with the final method approved in writing before work relies on it. Below, one benchmark runs through three methods using published federal indexes, and the rest of the page sets out how matches, alignment and the data record work.

Method comparison on a sample benchmark (top of range)

Unadjusted: each county’s published maximum as is. Regional price adjustment: the maximum scaled by the ratio of San Diego’s Regional Price Parity to the comparator metro’s (BEA). Regional wage adjustment: scaled by the ratio of San Diego’s all-occupations median wage to the comparator metro’s (BLS OEWS). Each index, year and source is listed below the table. A demonstration of method; nothing here is a finding.

The County’s 2025–2028 labor agreements describe the market target as the median of comparable classes on a cost-of-labor-adjusted basis. A cost-of-living adjustment is the common alternative. The two can point in different directions for the same county, which is why we show both before the County chooses (SOW 2.3.1). Under County Charter section 501.2 the Board weighs prevailing pay without tying County pay to another agency by formula, so every method here is information for that judgment.

CountyMetro used for the indexPublished maximumPrice factorPrice-adjustedWage factorWage-adjusted
SOW 4.2, 4.6

How a match is made and recorded

  • Matches are made on duties, level and minimum qualifications in the published specification. Titles are a starting point only.
  • A match needs substantial overlap in duties and the same level of responsibility. Where the peer class is a level above or below, we record that and use it only if the County approves the rule.
  • Where no comparable class exists, we write down why, as SOW 4.6 requires, and the county drops out of that benchmark’s calculations.
  • A benchmark needs a minimum number of valid matches, set with the County at kickoff, before its median is reported.
  • Every decision carries the analyst’s name, date and reason, and a second analyst checks a sample and every outlier.
SOW 2.2, 4.2

Internal alignment of the classes not surveyed

  • Each non-benchmark class is tied to a surveyed anchor through its series (for example the I, II and Senior levels of one occupation), its supervisory relationship, or a comparable class in another unit.
  • The County’s current range relationships are the starting point: if a class sits a set percentage above its anchor today, that relationship is shown and either kept or flagged with a reason.
  • Recommendations are stated as a percent movement to the class’s range and listed with the anchor used, so a reviewer can trace each one.
  • Classes in different units that share an anchor are checked together, so represented and unrepresented classes with the same work stay consistent.
SOW 5.5 to 5.8

What the County receives and owns

  • Attachment 2 sheets for every benchmark, and the summary spreadsheet of all results
  • The match register: every county, every benchmark, matched class, level, reason, source link and effective date
  • Internal alignment table for every non-surveyed class
  • Technical report and final report, outline and draft first
  • Raw and processed data, the codebook and the decision log, in native Excel and Word formats
Board Policy A-140 · Draft Agreement 10.13

How AI is used, and who decides

  • AI drafts first-pass matches from class specifications and flags likely level differences. That is its whole role in the survey.
  • A credentialed analyst accepts, changes or rejects every draft. No match, alignment or recommendation is made by software alone.
  • The prompts and outputs behind each first-pass match are kept with the match record and delivered to the County on request.
  • We disclose the AI functions in use at kickoff and report any change to the COR before it is made.