In The B2B Numbers Census, a website-content audit of 614 B2B companies observed in September 2026 and published in October 2026 by ClaimTheStory Research, 7 in 10 companies used someone else’s numbers, and 1 in 3 never published a number from their own data.
To build a benchmark report from your own records that outside readers can trust and check, you need clear rules before you count anything: a precise metric definition, full source and group counts, a fixed inclusion threshold, fair comparisons within matching groups, plain limits, a separate publishable table, and a complete methods note. If you’re still deciding whether a research project is worth the time and cost, start with original research for content marketing. Otherwise, this guide takes you through building a published benchmark in eight sections, in order:
- What a published benchmark report is, and how it differs from an internal one
- Define exactly what your benchmark figure counts
- Describe the data behind the benchmark: source, period, population and counts
- Set inclusion rules and a minimum group size before you compare
- Compare groups fairly
- Say what your data represents, and what it does not
- Decide what you can publish and what stays private
- Make the report checkable and easy to reference
What a published benchmark report is, and how it differs from an internal one
A benchmark report compares a measured figure against a reference point across groups or over time. This guide covers the kind a company builds from its own records and publishes so customers, peers and journalists can use the figures as a reference.
Many benchmarking guides teach an internal report: measuring your own team’s performance against competitors, running a gap analysis, and assigning action plans with owners and deadlines. That internal process is outside this page’s scope. An internal benchmark still needs clean definitions, inclusion rules and limits. A published report has to show all of those to readers outside your company, along with a methods note, a citation line and a clear data-access policy.
Define exactly what your benchmark figure counts
A benchmark is comparable only when every group is measured by the same rule, written down before anyone computes a number. Every metric definition needs four parts:
- The denominator: the exact unit being counted.
- The numerator: the event that counts toward the rate or share.
- The period rule: which date places an item inside the reporting window.
- Exclusions and missing items: how canceled, unfinished and incomplete records are handled.
In the HVAC dispatch example, the team writes its definition before pulling totals: the on-time arrival rate equals jobs where the contractor’s technician checked in within the arrival window the homeowner booked, divided by completed jobs with a recorded check-in. Jobs enter the period by their scheduled date (1 January to 30 June 2026). Canceled jobs are excluded, and completed jobs without a recorded check-in are left out of the rate and reported separately as missing data.
- Denominator Completed jobs with a recorded check-in.
- Numerator Jobs where the technician checked in within the arrival window the homeowner booked.
- Period rule A job enters the period by its scheduled date, 1 January to 30 June 2026.
- Excluded Canceled jobs.
- Missing Completed jobs without a check-in: left out of the rate, reported separately.
Keep the direction of the action clear in every sentence: the homeowner booked the arrival window, and the contractor’s technician arrived at the homeowner’s home. Keep the exact base wording too. Calling the denominator “completed jobs with a recorded check-in” rather than “all jobs” tells readers that canceled and unlogged visits are not in the rate. And an on-time share only measures how often technicians arrived within the booked window. It says nothing about how late the other arrivals were.
Describe the data behind the benchmark: source, period, population and counts
Readers and reporters need to see where your numbers came from before they trust a single figure. State five facts about your dataset upfront:
- Source: which system and log produced the records.
- Period: the exact start and end dates.
- Population: who is covered and how they entered the dataset.
- Total records: the full count before and after exclusions.
- Group counts: both the number of companies and the number of records in every comparison group.
If you’re still taking stock of what systems your company holds, see proprietary data for content. If you’re building a benchmark from a questionnaire rather than operational logs, the same five items apply, and survey research for content covers how to design and run the survey. If your own records are too thin for a benchmark, see how to get data for content.
- Source The software’s job log, extracted with the data owner’s approval.
- Period Jobs scheduled 1 January to 30 June 2026.
- Population Residential HVAC contractors using the software, in three regions.
- Total records 56,200 jobs scheduled; 50,000 completed jobs with a recorded check-in after removing 3,800 canceled and 2,400 without a check-in.
- Group counts Six region-and-size groups, 417 contractors, each with its contractor count and its base.
Set inclusion rules and a minimum group size before you compare
Inclusion rules decide which companies and records enter the comparison, and you must set them before looking at the results. There is no universal minimum sample size for a benchmark report. How many records or companies you need depends on what you are measuring and what your report says about it.
Set one minimum group-size rule for your dataset, label it as your own choice, and apply it the same way to every table and every total. If you suppress a small group in one table but include its numbers in an overall total, a reader can subtract the visible rows from the total and recover the hidden group.
Compare groups fairly
A fair comparison uses the same definition, the same calendar window, and comparable groups on both sides. Never compare your internal figures against outside industry numbers whose definitions and collection methods you cannot verify.
Check whether suppressing a group tilts your top-line totals, and test whether missing records could skew the rates. Keep your wording strictly descriptive: a benchmark shows how often an event happened, not what caused it.
Say what your data represents, and what it does not
A company’s own records describe its customers, not the entire industry. Never present customer logs as an industry-wide standard. Name who is covered in your title and headline, using phrasing such as “among contractors using [Product]” rather than “US HVAC contractors”.
List the questions your columns cannot answer, and the extra field each one would need. In the HVAC dispatch example, the benchmark covers only residential HVAC contractors using this software in three regions between 1 January and 30 June 2026, and only completed jobs with a recorded check-in. It cannot show how many minutes late the late arrivals were (that needs arrival timestamps past the window), why technicians missed a window (that needs delay reason codes), or whether homeowners were satisfied (that needs post-visit ratings).
Can show
How often technicians arrived within the booked window, among contractors using the software, on completed jobs with a check-in, January to June 2026.
Can’t show
- How many minutes late the late arrivals were: needs arrival timestamps.
- Why a window was missed: needs delay reason codes.
- Whether homeowners were satisfied: needs post-visit ratings.
Decide what you can publish and what stays private
Sorting records into raw rows, publishable aggregates and private fields is covered in proprietary data for content and turning data into content. For a benchmark report, keep your internal working table and your publishable table as separate documents so suppressed groups never slip into published totals.
Published
- Rates and base counts for the five qualifying groups.
- Totals by contractor size (4+: South and Midwest only).
Not published
- Contractor names and individual contractor rates.
- West, 4+ technicians, and any West regional total.
Before publishing, assess disclosure risk with the data owner and against your customer agreements. A minimum group-size rule alone does not make data anonymous: if one large contractor accounts for most of the jobs in a region-and-size group, readers may still recognize it.
Make the report checkable and easy to reference
An inspectable benchmark lets a journalist, analyst or buyer check your source, definitions, dates, group counts and limits without having to email your team first. Publish a complete methods note with the report, covering:
- Source system and data owner
- Exact observation dates and period rule
- Population covered and exclusions
- Metric definition, numerator and denominator
- Inclusion rules and minimum group-size choice
- Group counts for both companies and records
- Known limits and missing-data rates
- How to cite the report, and where to download the publishable table
Our guides to turning data into content and writing up survey and research findings show how to structure a methods note in detail.
A report that’s easy to check and cite saves readers work when they want to reference your numbers. It doesn’t guarantee coverage, backlinks or citations. If you plan to pitch the findings to reporters, see data-led PR for how to shape a press angle without overstating what the records show. To see how recurring benchmarks fit into a broader content plan, visit our hub on data-driven content marketing.
If you have data and don’t know what to do with it, we find a question it can answer that your customers would care about. Book a demo