--- name: comp-band-read source: https://app.decimal.ai/s/comp-band-read@1/SKILL.md source_sha256: e4f928674d5f --- # Read a pay survey into a recommendation A market survey hands you a *distribution*, not a price. Each role/level row gives three columns — the 25th, 50th, and 75th percentile of what the market pays. The reflexive move is to grab the middle number and call it "the market rate." That throws away the two decisions the survey exists to inform: **where in the band to aim**, and **what the number is made of**. A recommendation names both. Four steps turn the row into an answer: read the percentiles as positions, adjust the band for the actual level and location, split the target into base plus variable pay, and state the read in one line. ## 1. The percentiles are positions, not a range to average Each column is a deliberate place to stand, tied to how hard you intend to compete for this hire: - **50th percentile — pay at market.** The default target for a solid role you can fill on a normal timeline. Half the market pays less, half more; you are neither buying nor discounting. - **75th percentile — pay to lead.** Aim here for a scarce or hot skill, a competitive counter, a backfill you must close fast, or a candidate you cannot afford to lose. You are paying a premium to win. - **25th percentile — pay below market.** Only defensible when something non-cash carries the offer — real equity upside, a magnet brand, a mission pull — or a hard budget ceiling forces it. Whenever you land here, call out the raised risk of declines and early attrition; do not present it as neutral. Do not average the three columns, and do not treat the median as automatically correct. Pick the column the hiring situation calls for, and say why. A target between two columns is fine — interpolate linearly (a 60th-percentile aim sits about 40% of the way from the median toward the upper column). ## 2. Adjust the band for level, then location — multiplicatively, in that order A survey row is cut for one specific level and one reference geography (often a national figure or a named metro). Using it for a different level or place means adjusting the whole band, not quoting it raw. - **Level first.** Match the target level to the surveyed level. If the role sits between two surveyed levels, interpolate the band between them — do not borrow an adjacent level's numbers unchanged, and do not stretch one level's band to cover a different scope of work. - **Location second.** Apply a locality factor to scale the level-matched band to the target market: a high-cost metro lands above the reference (a factor over 1), a lower-cost market below it (under 1). Scale all three percentile points by the same factor so the band keeps its shape. Apply the two in order — establish which band, then scale that band — and state both adjustments in the output. A single adjusted number with the reasoning hidden cannot be checked or defended. The locality factors themselves come from a geo-differential source you supply; treat any factor shown here as illustrative, not authoritative. ## 3. Split the number into base and total — and never confuse the two Before you quote anything, read the survey's column definition: a percentile figure is **either** a base-salary number **or** a total-cash / total-compensation number, and the two are not interchangeable. Getting this wrong is the most common way a survey read goes silently off. Total compensation is made of parts: - **Base salary** — the anchor. Offers are leveled and negotiated on base; raises are a percentage of base. This is the figure the candidate compares first. - **Target bonus** — expressed as a percentage of base, paid on performance. - **Equity** — an annualized grant value (total grant divided over its vesting period). - **Total = base + target bonus + equity.** So: - If the survey column is **total comp**, back the base *out* of it — subtract the expected bonus and annualized equity — rather than quoting the total figure as base salary. - If the column is **base only**, add bonus and equity *on top* to state total, and never compare that base-only figure against a candidate's current *total* package. Recommend **both** a base number and a total number, with the components listed, so the reader sees what they are actually offering. ## 4. State the read in one line Close with a single sentence that exposes every decision: the target percentile and why, the level-and-location adjustment applied, and the base/total split. For example: *"Targeting the 75th percentile because the skill is scarce and we're countering a competing offer; took the L4 band, scaled it +15% for the metro, which puts base at ~$185K and total near $250K with a 15% target bonus and equity."* One line like that turns a silent guess into a decision someone can audit. ## Worked example (illustrative numbers — run your own) Survey row, national reference, **total-cash** column: L4 Software Engineer — p25 $170K, p50 $195K, p75 $225K. Target: an L4 backfill in a high-cost metro, scarce skill, need to close fast. 1. **Position:** scarce + fast close → aim p75 = $225K total. 2. **Level:** target is a clean L4 match, no interpolation. 3. **Location:** metro factor 1.15 → $225K × 1.15 = ~$259K total. 4. **Split:** column is total cash. With a 15% target bonus and ~$40K/yr annualized equity, back out base: base × 1.15 (bonus) + $40K = $259K → base ≈ $190K, target bonus ≈ $29K, equity $40K. 5. **Read:** "p75 for a scarce fast-close L4; L4 band scaled +15% for the metro; base ~$190K, total ~$259K including a 15% bonus and $40K annual equity."