How compensation teams keep sales pay benchmarks current during the year

Compensation teams keep sales pay benchmarks current during the year by pairing two inputs: a survey or official estimate that sets the level once a year, and a weekly reading of advertised pay that shows when the level moves. This post gives the official release calendar, a control chart with limits from the NIST/SEMATECH e-Handbook of Statistical Methods, a rule for a real move, and a worked quarter of readings.

How compensation teams keep sales pay benchmarks current between surveys

Take the level once a year from a survey or official occupational estimates. Between refreshes, chart the weekly median of advertised pay for each sales role and country against limits set on a baseline, treat a reading as a move only when it breaks a stated rule, and shift the benchmark by the size of the move.

The two inputs answer different questions. A survey or official estimate gives depth: a full distribution of pay for one reference date. A weekly median of postings gives timing: what employers offer now. The first places the benchmark, and the second says whether it still holds.

Chart each role and country separately, and base pay apart from on-target earnings, because a pooled median moves whenever its mix of roles changes. The entry on on-target earnings explains the split, and benchmarking sales compensation from advertised pay places a plan against the quartiles. This post keeps that placement current.

What official wage statistics publish, and how often

Of the official series below, those that give pay levels by occupation appear once a year or less often, months after the date they describe. The more frequent series are broader: a quarterly index of change by occupational group in the United States, and monthly average earnings by industry and sector in Great Britain. Every date was read on the publishers' own pages on 7 October 2026.

Three properties decide how to use them.

• Wages include commission paid. The Bureau of Labor Statistics counts commissions, incentive pay and production bonuses in occupational wages, and leaves out overtime, nonproduction bonuses and the employer cost of benefits (OEWS questions and answers). The survey behind the Employment Cost Index also counts commissions (BLS Handbook of Methods). An official sales figure reflects variable pay earned; on-target earnings state it at full target.
• The annual estimate is pooled. Each OEWS estimate combines six semiannual panels collected over three years. The Bureau does not encourage using OEWS data for time-series analysis, and notes that a sudden change reaches the estimates only gradually.
• Annual results are revised. The Office for National Statistics publishes each survey year as provisional, then revised; its occupation tables list both editions for 2024.

Annual or quarterly refresh of a sales compensation benchmark survey

Refresh the level annually, when a new survey edition or official estimate arrives, and review it each quarter against a faster reading. The quarterly review asks one question per role and country: has the weekly median broken a control rule since the last refresh? If not, age the level forward with an official index; if so, move it.

Ageing carries a level from the survey's reference date to the date you use it. In the United States, apply the change in the Employment Cost Index for wages and salaries of sales and related occupations between those two dates. In Great Britain, the monthly average weekly earnings series gives a rate by industry and sector.

The weekly chart decides between ageing and moving. A broad occupational index averages many roles, so a move confined to one role in one country can pass almost unseen in it. When the chart is quiet, age with the index; when a rule fires and its checks hold, move the level by the readings since the shift.

Refresh cadence for sales compensation benchmarking market data

Set a cadence for each input rather than one for the benchmark. Read advertised pay weekly, review the charts monthly, age the survey level quarterly with an official index of change, and reset levels annually when the survey and official occupational estimates are published. Between those dates, act only when a control rule fires and its checks hold.

The weekly reading in Fokals data

Sales Pay Benchmarks, in the hiring dataset, give what employers advertise for sales roles across companies, per closed week, role and country, with all countries together as a group of their own. Each group carries the lower quartile, median and upper quartile of advertised base pay and of on-target earnings on one yearly US-dollar scale, with the number of open sales postings behind it that state a base or on-target figure. A group is published once at least five postings stand behind it.

The figures are employers' own statements, read from the posting's own text and kept as written and as yearly US-dollar midpoints at a reference exchange rate. The role, from SDR or BDR to sales leadership and partnerships, means the same thing every week. Each Monday-to-Sunday UTC week is written once after it closes and never revised, so every reading is point-in-time and a past signal can be audited. The data dictionary defines every field and every role.

Putting the weekly median on a control chart

A weekly median is one number a week, the case the handbook's individuals control chart is built for. Section 6.3.2.2 of the NIST/SEMATECH e-Handbook measures variation by the moving range of two successive observations, and sets the limits from the mean of the readings, x̄, and the mean of the moving ranges, MR̄:

MR(i)        = |x(i) - x(i-1)|
Center line  = x̄
UCL          = x̄ + 3 (MR̄ / 1.128)
LCL          = x̄ - 3 (MR̄ / 1.128)

The handbook gives 1.128 as the value of d2 for n = 2. Section 6.3.1 explains the factor of three: for normal data, a point falls beyond three-sigma limits on either side with probability 0.0027, and a point outside the limits warrants a search for its cause.

What follows is this post's own algebra and choices, not the handbook's.

• Sigma. Call MR̄ / 1.128 the chart's sigma, so the limits sit three sigma from the centre line. Add lines at one and two sigma on each side for the rules below.
• Baseline. Take 8 to 13 weekly readings in which no rule fires and each clears your postings floor. Hold the limits fixed, and recompute only after a confirmed move.
• Floor. A group is published from five postings. Set a higher floor for a chart, and leave weeks below it off.
• Overlap. A group counts the postings open in its week, so a long-open posting sits in several readings. Successive medians share part of their sample, which narrows the moving ranges and the limits: expect more signals than the handbook's rates below.
• Exchange rate. Figures sit on a US-dollar scale at a reference rate, so a chart for a country paid in another currency moves with the rate too. Chart the rate beside it.

The query scores each week of one role and country against a baseline you choose, and marks the four rules below. Values after a colon are parameters: role, country, postings floor, and the baseline's first and last week.

with readings as (
  select week_start, postings, ote_median_usd as x
  from sales_pay_benchmarks
  where role = :role
    and country = :country
    and postings >= :postings_floor
),
baseline as (
  select x, abs(x - lag(x) over (order by week_start)) as mr
  from readings
  where week_start between :baseline_from and :baseline_to
),
limits as (
  select avg(x) as centre, avg(mr) / 1.128 as sigma
  from baseline
),
scored as (
  select r.week_start, r.postings, r.x, (r.x - l.centre) / l.sigma as z
  from readings r
  cross join limits l
)
select
  week_start,
  postings,
  x                                               as median_ote_usd,
  round(z::numeric, 2)                            as z,
  abs(z) > 3                                      as beyond_three,
  count(*) filter (where z > 2) over w3 >= 2
    or count(*) filter (where z < -2) over w3 >= 2 as two_of_three,
  count(*) filter (where z > 1) over w5 >= 4
    or count(*) filter (where z < -1) over w5 >= 4 as four_of_five,
  count(*) filter (where z > 0) over w8 = 8
    or count(*) filter (where z < 0) over w8 = 8   as eight_one_side
from scored
window w3 as (order by week_start rows 2 preceding),
       w5 as (order by week_start rows 4 preceding),
       w8 as (order by week_start rows 7 preceding)
order by week_start;

A rule for when a move is real

A point beyond the three-sigma limits is the basic signal. Section 6.3.2 of the handbook adds the Western Electric (WECO) rules, which catch smaller moves by looking at runs. Restated for either side of the centre line, a move is signalled by:

1. one point beyond three sigma;
2. two of the last three points beyond two sigma on the same side;
3. four of the last five points beyond one sigma on the same side;
4. eight points in a row on the same side of the centre line.

Sensitivity has a price, and the handbook states it: with the three-sigma rule alone, a false alarm comes about every 371 points on average, and with the WECO rules added about every 91.75 points. On a weekly chart that is, by this post's arithmetic, one false alarm in a little over seven years against one in about 21 months. Across charts they add up: base pay and on-target earnings for four roles in three countries make 24 charts, and if they were independent the full rule set would give 24 / 91.75 = 0.26 false alarms a week, about one every four weeks, before overlap adds more. A rule therefore opens a review, and the review decides.

For a slow drift, section 6.3.2.4 gives the EWMA chart, EWMA(t) = λ Y(t) + (1 - λ) EWMA(t-1), with λ usually set between 0.2 and 0.3. The handbook notes that it can be made sensitive to a small or gradual drift, where the Shewhart chart reacts only once the last point leaves its limits. This post's suggestion: the individuals chart for jumps, and an EWMA chart beside it for drift.

When to act on a signal

A fired rule starts four checks, and the checks decide.

When the checks hold, the two charts name the kind of move.

• Base and on-target earnings move together. The level of pay has moved. Update the reference for offers to new hires.
• On-target earnings move and base holds. Variable pay at target has changed. Review the variable part and the split first.
• Base moves and on-target earnings hold. The split between fixed and variable pay has shifted.

A confirmed move updates the reference for new-hire offers at once. Ranges and plan design wait for the annual reset, unless the move takes your plan past a tolerance set in advance, which opens an off-cycle review. Keep the chart, the rule and the checks with each decision.

Worked example: a quarter of weekly readings at Acme Robotics

This example is illustrative. Acme Robotics is an invented company, and every reading is invented to show the arithmetic; none is a Fokals figure.

Assumptions:

• Acme Robotics pays account executives in one country on-target earnings of 150,000 US dollars.
• The thirteen readings are weekly medians of advertised on-target earnings for that role and country, each above Acme's postings floor.
• Weeks 1 to 8 are the baseline. Weeks 9 to 13 are judged against limits held fixed from it.
• In the quarter the base pay chart fired no rule, the postings count stayed near its baseline level and the exchange rate was steady.
• Acme's policy: a move is confirmed when a run rule fires and the checks hold; a confirmed move that leaves the plan more than 1% below the median opens an off-cycle review of new-hire offers.

1. Centre line: the eight readings sum to 1,197,000, and 1,197,000 / 8 = 149,625.
2. Mean moving range: the seven ranges sum to 10,500, and 10,500 / 7 = 1,500.
3. Sigma: 1,500 / 1.128 = 1,329.79.
4. Limits: 149,625 + 3 × 1,329.79 = 153,614 and 149,625 minus 3 × 1,329.79 = 145,636. The one-sigma lines sit at 148,295 and 150,955, the two-sigma lines at 146,965 and 152,285.
5. No baseline reading lies beyond two sigma and none of the four rules fires, so the baseline stands.

No reading reached the upper limit of 153,614, so the three-sigma rule alone stayed silent all quarter. The run rules fired in week 12, the fourth week of the move, and week 13 confirmed it. The checks hold under the assumptions.

The provisional new level is the mean of weeks 9 to 13: 761,500 / 5 = 152,300, a move of 2,675, or 1.8%, from the centre line. Acme's plan sat 375 above the old centre and now sits 2,300 below the new level, which is 1.5% and past the 1% tolerance, so the off-cycle review opens. With base pay steady, the move is in variable pay at target, and the review starts there. The new centre line stays provisional until eight readings after the shift allow the limits to be computed again.

Frequently asked questions

The queries and code on this page are examples to adapt. Test them in your own environment before you rely on them.

What this page says about the products it names was checked against their public documentation on 7 October 2026. Product and company names are trademarks of their owners. Fokals is not affiliated with them or endorsed by them.