How we research and compute our figures

Home › About › Methodology

Last updated: 31 October 2026

In short: competitor prices come from each firm's own pages, dated and linked, and a script checks every week that they are still there. Our data studies use our own campaign exports, say which dataset each number comes from, and leave out any group too small to read.

This page explains how the research on roimaxi.com is done. Who writes it, and how we handle conflicts of interest, is on the editorial policy. Every change we have made to a published figure is on the corrections log.

How we research agencies and their prices

ROI Maximizer is an agency, and a lot of this site compares us with other agencies. So the rules for those pages are strict.

The firm's own page first. A price, a term or a guarantee is quoted from the company's own website wherever it publishes one, with a link to that page.
Third-party figures are labelled. Where a firm publishes nothing, we may cite a review site, a directory or another agency's article, and the sentence says whose figure it is. We never present it as the firm's own price.
Dated. Prices are written "as of" the month we read them, and comparison pages say when their sources were read.
Nothing published, nothing invented. If a firm does not publish a price, the page says so. We do not estimate one and call it theirs.
Our own arithmetic is shown as ours. When a page works out a cost per meeting from a firm's published fee, the inputs are on the page and the sentence says it is our calculation.

The claims ledger and the price watch

Two scripts keep those figures honest after a page goes live.

The claims ledger (scripts/claims-ledger.mjs) reads every blog post, finds each outside link that sits in the same sentence or table row as a dollar, pound or euro figure, fetches the source once and records what it found:

ok: the cited figure is on the source page, word for word. That pair goes into the price watch.
manual: the site blocks automated fetches, so Moe checks it in a browser and dates the check.
missing: the page loaded but the figure is not in it. A person reads the source and fixes the post or the citation.
gone: the page no longer exists.

On 31 October 2026 the ledger covered 93 posts: 320 post-and-source pairs on 63 domains, 274 confirmed word for word, 21 to check by hand and 23 marked missing for a person to read. It runs again after any post adds or changes a cited price, and at least once a quarter.

The price watch (seo/price-watch.json) is the list of exact strings each source must still show, and for some, strings it must not show again. Every Monday a scheduled check fetches each page. If a firm's page has changed, the post that cites it is re-checked and corrected, and the change goes in the corrections log.

How our data studies are computed

Our studies use exports from our own sending platforms, not surveys or estimates.

One dataset per number. Each table names its source: per-step analytics (sends and human replies), the export of people who replied positively, or the full-text reply classes. We never add two datasets together.
n is always stated. Every table and every chart caption gives the count it rests on.
Rates need a denominator. A rate appears only where we have emails sent or contacts. The positive-reply export has no send counts, so it gives shares, never rates.
Labels mean one thing. "People who replied positively" counts each person once. "Positive-reply events" counts replies, with one client's campaigns excluded. "Human replies" means any reply a person wrote, yes or no, with auto-replies and out-of-office messages removed. The three are never mixed in one sentence.
Exclusions are written down. If a client, a campaign or a period is left out, the method section says which and why.
Small groups are suppressed. In the reply-rate studies, any group under 30 is left out, and a rate built on fewer than 30 replies is shown but marked. In the fintech email security benchmark, results appear only for groups of 20 domains or more.
Aggregates only. The script that computes a study reads the source files and writes only totals. No name, email address, company or reply text leaves it, and no per-company result is published.
Clients are named only with permission. Five clients may be named: PR Noir, P1 Communications, Blinkoo, StartMeUp.ai and Deus Tech 2.0. Any other client is described by its market.

Each study's own method section gives the dates, the sample and its limits. Two examples: cold email reply rates by industry and the fintech email security benchmark. Where the aggregate table is public, it is a CSV you can download and reuse with credit.

Keeping figures current

The data posts are refreshed on the first working day of each quarter, starting 4 January 2027. The figures are recounted under the same rules, at the same URLs. A figure that cannot be recounted the same way keeps its old value and its old date.

A post's "Updated" date moves only when a figure or a substantial passage changes. It never moves for a typo or a layout change. Its publish date never moves at all.

Before every build, a content check stops the site from shipping figures we have withdrawn, client names we may not use, and our own price written as a single number or in pounds. Our price is a range in US dollars.

Questions about a figure

Email moe@roimaxi.com with the page and the number. If you want the source data behind a study, ask: we share aggregates, never individual records.

ROI Maximizer is a trading name of M Ventures LTD, registered in England and Wales, company no. 15554493. Registered office: 4 Beau Street, Bath BA1 1QY, United Kingdom. moe@roimaxi.com