Methodology

The rules in one paragraph

Every salary band on B2BMarketingSalaries.com is built from public data, and every row says which source it came from. We do not run our own survey, we do not collect salaries from readers, and we do not have access to any private compensation dataset. What we do is read the public sources that already exist, weight them by how good they are for a given question, and turn them into bands that are labeled by level, geography and company size. The site is sponsored by Metadata.io, whose AI marketing salary guide is one of the sources we cite, and it is held to the same rules as every other source. This page explains the rules.

Sources and how they are weighted

No single public salary source is good at everything, so we use several and weight them by the question being asked. The table below lists each source we draw on, what it does well, and what it does badly. The weights are a judgment, not a formula; we state them so you can disagree with them.

In practice, most bands come from two or three of these sources agreeing within a reasonable margin. When the sources disagree by more than about 20% at the midpoint, we show the disagreement in the row rather than averaging it away, and we explain the likely reason (usually a sample skewed toward tech, or a title that means different things at different company sizes).

How salary bands are built

A band on this site is a low, a midpoint and a high, labeled with the role, level, geography and company size it describes. It is built in four steps, and each step is reversible from the row note.

1. Pick the anchor. For each role and level, one source is chosen as the anchor based on the weight table above. For an ABM manager at a Series B company, that is Pave. For a marketing operations manager at a 2,000-person non-tech company, that is Robert Half, cross-checked against BLS.
2. Set the midpoint. The midpoint is the anchor's median (or 50th percentile) for the role, level and national geography. If the anchor publishes only a range, the midpoint is the center of that range and the row is flagged "range-derived."
3. Set the low and high. The low is the 25th percentile and the high is the 75th percentile where the anchor publishes them. Where it does not, we use the lowest 25th and highest 75th from the two secondary sources, which widens the band deliberately. We prefer a wide honest band to a narrow misleading one.
4. Apply normalization. Geography and company-size multipliers, described below, are applied to produce the by-city and by-size rows. The national row is always shown so you can see the unadjusted figure.

We do not publish a single "average salary" for any role. Averages hide the spread, and the spread is the useful information when you are negotiating.

Geography normalization

City rows on the salaries by city page are produced by applying a geography multiplier to the national midpoint. The multiplier for each metro area is the ratio of that metro's BLS median for marketing managers to the national BLS median (1), rounded to two decimal places, and checked against Glassdoor's location-adjusted figures for the same title. Where the two disagree by more than five points, we print the BLS-derived figure and note the disagreement. Remote roles are listed under the company's pay-location policy where a source states it, and under "national" otherwise. We do not adjust for cost of living; a salary band is what employers pay, not what it buys.

Company-size normalization

Company size changes both the level of pay and its composition. The salaries by company size page uses four bands: under 50 employees, 50 to 250, 250 to 1,000, and over 1,000. Pave's stage-based data is mapped to headcount bands using Pave's own published stage-to-headcount ranges (2), and Robert Half and BLS supply the larger bands where Pave is thin. The mapping is approximate, and every by-size row says so. The general pattern the sources agree on is that base pay rises with size while equity falls as a share of total comp; the specific numbers are in the tables, with the source per row.

What "total comp" includes

The site reports base salary as the primary figure on every page. Where a "total compensation" figure appears, it is defined the same way everywhere:

When a source's "total pay" is a modeled estimate rather than a reported figure (Glassdoor's is), we do not use it in the total comp column. Total comp rows draw on Pave and Levels.fyi, which report components separately.

Sample-size caveats

Public salary data is thin for exactly the roles this site covers. "Demand generation manager" is not a BLS occupation; "ABM manager" barely exists as a title outside B2B software; "RevOps" is a few years old. So the sample behind any band may be small, and we say so in three ways.

• Each row carries a confidence flag: solid (three or more sources agree, anchor sample in the hundreds), fair (two sources, or one strong source with a stated sample), or thin (one source, small or unstated sample).
• Where the anchor publishes a sample count, it is printed in the row note. Where it does not, the note says "sample not stated."
• Senior titles (VP and CMO) are flagged thin by default. The VP marketing and CMO pages explain why: fewer people hold the roles, fewer report their pay, and the ones who do are not typical.

A thin band is still shown, because a labeled thin band is more useful than nothing. It is not a figure to anchor a negotiation on without checking the source yourself.

The sponsor's guide under these rules

Metadata.io's AI marketing salary guide (8) is a vendor publication, and we treat it as one. It appears on this site in one place, the AI premium page, where it is the only public source we have found addressing that specific question. It is labeled as the sponsor's, its stated methodology is summarized, and its figures are never used to set a base band for any role. Nothing on this site draws on Metadata's internal financial or operating metrics, and nothing on this site reports any individual Metadata employee's compensation.

Update cadence and corrections

Role pages are reviewed twice a year, in the month after BLS releases its annual occupational data and again six months later. Pave and Robert Half publish annually, and their pages are refreshed within 30 days of a new edition. The date at the top of each page is the date of the last review. Bands that have not been reviewed within twelve months are flagged in the row note as "prior edition." If you believe a band is wrong, send the page URL, the row and a link to the public source you think is correct through the contact details on the about page; we check the source, and if it supports the change we update the row and add a dated note. Corrections from any employer, recruiter or reader are handled the same way.