Email software pricing, 2026: three findings from vendors' own pages

What are the three findings?

Each finding below gives its exact sample, the plans and pages it comes from, the formula with a worked example, what is left out, its limits and what it does not show. The numbers are calculated from the same records as the Value Index when this page is built, so a corrected record corrects this page.

Who publishes this, and what do they sell?

• The commercial relationship. Donezo Inbox sells an email client for Gmail and Google Workspace, and that client is one of the products in the Value Index. Donezo Inbox also publishes the Value Index and this page. Read the findings with that in mind.
• Its own plans are excluded. Three Donezo Inbox plans are in the Index. They appear in the downloadable data, marked as the publisher's, and are in none of the three findings.
• The vendors were not consulted. None was told about these findings, none reviewed a figure, and none has paid for or influenced a row. The publisher is not asking for coverage, a review, a rating or a link.
• Nothing was tested. The data records what vendors' pages state. It does not measure how well any product works, including the publisher's.
• The sample is hand-picked. It is the set of products the publisher chose to compare, not a random sample and not the whole market. A different set could give different numbers.
• It is a data note, not a study. It has not been peer reviewed. The method and the data are on this page so that you can check them.

What is in the sample, and what is left out?

The Value Index holds 31 plans from 13 products. Take away Donezo Inbox's own three plans and the two plans whose price is not confirmed and 26 plans from 12 products remain. Each finding then applies its own filter to those 26; the plans it keeps are listed under it. Five of the 26 (Mimestream, Front, Hiver Omni, Help Scout and Mailbird) are price-only rows with no second billing term or feature columns, so they fall outside all three filters.

Prices are list prices in US dollars, per person per month as the vendor's page prints them, with no taxes, discounts, promotions, regional prices or negotiated quotes. Euro prices are never converted. Billing terms are never mixed: a total uses the term the plan is listed on, and Finding 2 compares a plan's own annual and monthly prices.

Collection date. Each figure was copied by hand from the vendor's own pricing page, with billing toggles and plan tabs opened, on the date shown next to the plan. The plans used here were read between October 21, 2026 and October 31, 2026, across five products: Superhuman Mail, Fyxer, Missive, SaneBox and Inbox Zero. The Index withholds the totals of any row not re-read within 45 days.

Finding 1: What do ten people pay for a team plan with AI?

Result. For six plans from three vendors (Superhuman Mail, Fyxer and Missive), ten people cost $2,880 to $7,680 a year at the annual-billing list price. The dearest is 2.67 times the cheapest. The median is $4,140.

Sample. A plan is in if it is not the publisher's, its price is confirmed, it is paid, it is priced in US dollars and listed on annual billing, its page lists shared-inbox or team features, its page says AI is part of the plan (the plan's AI level is not "none" or "not stated"), and any seat cap on the plan is at least ten.

Method. Ten people a year = the monthly price on annual billing × 12 × 10. Example: Missive Productive is $24, so $24 × 12 × 10 = $2,880. The ratio is the highest total divided by the lowest: $7,680 ÷ $2,880 = 2.67. The six totals in order are $2,880, $3,000, $3,960, $4,320, $4,440, $7,680. The middle two of the six are $3,960 and $4,320, so the median is ($3,960 + $4,320) ÷ 2 = $4,140.

If billed monthly instead. The same six plans, at the monthly price their pages also print, cost $3,600 to $11,880 a year for ten people (median $5,100).

Exclusions. Donezo Inbox's own plans. Spark Plus and Spark Pro (price not confirmed). Plans listed on monthly or two-year billing, because their price is not an annual-billing price (their monthly-only case is below). Plans whose page lists no AI or whose AI is "not stated", and plans whose page lists no team features.

Sensitivity. Adding the four Shortwave plans that are billed monthly only and otherwise fit (Pro, Business, Premier and Max), at their monthly price, gives ten plans from four vendors at $2,160 to $14,400, median $4,140. That mixes billing terms, so it is not the headline.

• Limits. Three vendors and a hand-picked set of plans. A list price times ten is not a quote: the Fyxer page asks for a volume quote at ten or more members, and a negotiated deal may differ. The AI allowance differs from plan to plan (credits on Fyxer, no quota printed on Superhuman, credits or your own key on Missive), so a price is not a measure of value.
• What this does not show. What any team actually pays, that the cheaper plans do the same job as the dearer ones, or anything about plans not in the sample.

Finding 2: How much cheaper is annual billing than monthly?

Result. For thirteen paid plans from five vendors (Superhuman Mail, Fyxer, Missive, SaneBox and Inbox Zero) whose pages print both prices, annual billing is 10.0% to 43.2% cheaper than monthly billing, median 20.0%. Said the other way round, monthly billing costs 11.1% to 76.1% more (median 25.0%).

Sample. A plan is in if it is not the publisher's, its price is confirmed, it is paid, it is priced in US dollars, and its vendor's page prints both an annual-billing price and a monthly-billing price for that plan.

Method. Annual is cheaper by = (monthly − annual) ÷ monthly. Monthly costs more by = (monthly − annual) ÷ annual. Extra a year = (monthly − annual) × 12. Example: SaneBox Lunch is $15.99 billed monthly and $9.08 billed annually, so annual billing is ($15.99 − $9.08) ÷ $15.99 = 43.2% cheaper, monthly billing is ($15.99 − $9.08) ÷ $9.08 = 76.1% dearer, and monthly billing adds $82.92 a person a year. The median is the middle value of the thirteen plans' percentages. The two directions are not interchangeable: 10% to 43.2% cheaper is not the same as 10% to 43.2% more.

Exclusions. Donezo Inbox's own plans. Spark Plus and Spark Pro (price not confirmed). Free plans. Plans whose page prints only one billing term, which includes Shortwave: its page shows a monthly price only. SaneBox Snack, SaneBox Lunch and SaneBox Dinner print three terms (every two years, annually, monthly) and are compared on annual against monthly only.

Why SaneBox is in. SaneBox prints three terms for each of three plans, so its annual and monthly prices can be compared like the rest. Its gaps are the widest in the sample, so the top of the range is its.

• Limits. Five vendors and a hand-picked set of plans. Annual billing means paying for a year at once; the Index does not record refund or cancellation terms. The biggest gap in dollars is Fyxer Team at $420 a person a year, the smallest Inbox Zero Starter at $24.
• What this does not show. How many customers pay monthly or annually, whether a discount is available on negotiation, or anything about vendors that print one billing term only.

Finding 3: What does an agent connector (MCP) add to the price?

Result. Of six vendors in the sample, three (Superhuman Mail, Missive and Fyxer) state a connector for outside AI agents (MCP) on a plan. Each lists it only on a plan above its cheapest paid plan, at $96 to $324 more per person a year (32.0% to 73.0% more). n is small: three vendors, seven plans. Read it as "three vendors that state one", not as a market rate.

Sample. A vendor is in the sample if it has at least one plan that is not the publisher's, has a confirmed price, is paid, is priced in US dollars and had its features read: Superhuman Mail, Fyxer, Shortwave, Missive, SaneBox and Inbox Zero. Of those, a vendor is a finding if its pages state an MCP connector on at least one plan. Shortwave, SaneBox and Inbox Zero state none. "Not stated" is a statement about a page, not about a product.

Method. For each vendor, the yearly price of a plan = its monthly price × 12. Extra a person a year = the yearly price of the first plan with the connector − the yearly price of the cheapest paid plan. Step up = that extra ÷ the cheapest paid plan's yearly price. Example: Fyxer Pro is $444 a year and Team is $768, so the extra is $768 − $444 = $324, a step up of $324 ÷ $444 = 73.0%. The calculation stops with an error, rather than print a figure, if a vendor's plans without the connector are not all cheaper than its plans with it, or sit on different billing terms.

Exclusions. Donezo Inbox's own plans (its pricing page says its connector is on every plan; that is the publisher's claim and is not counted here). Spark (price not confirmed). Free plans, and plans whose features were not read. A mention of MCP that only describes a product's own AI connecting out to other apps is not counted as a connector for outside agents.

• Limits. Three vendors. The higher plan adds other features as well as the connector, so the step is the price of moving up a plan, not a price for the connector alone. A vendor with no connector listed may still have one.
• What this does not show. That the connector is why a plan costs more, that vendors with none listed lack one, or how good any connector is.

How can you check the numbers?

• findings-data.csv: one row for every plan in the Index, with the vendor page, the date read, the fields used, a flag for each finding it is part of and the figures calculated for it, including why a plan was left out.
• findings-data.json: the same rows, the written definition of each sample and the results.
• calculate.mjs: a script with no dependencies that reads findings-data.json and works out all three findings again from the prices. Run node calculate.mjs findings-data.json in a folder holding both files (Node 18 or later). It prints the same numbers as this page and stops with an error if they differ.
• The full list of plans, with each plan's features, is in the Email Software Value Index and its own downloads (data.csv, data.json).

A separate and older piece of work, a working paper with a different sample and method, is on its own page: AI features and pricing in email software. It is a preprint that has not been peer reviewed.

How should you cite it, and where do you report a mistake?

Donezo Inbox Research (2026). Email software pricing, 2026: three findings from vendors' own pages. https://donezoinbox.com/research/email-software-pricing-2026/ (vendor pages read October 21, 2026 to October 31, 2026). Please quote a figure together with its sample, for example "across thirteen plans from five vendors in a hand-picked set of products".

Licence. The text, tables and charts on this page, the calculation script and the way the data file is arranged and coded belong to Donezo Inbox Research and may be used under Creative Commons Attribution 4.0 (CC BY 4.0): copy, share and adapt them, including commercially, if you credit "Donezo Inbox Research" and link to this page. The licence covers only what Donezo Inbox has the right to license. It does not cover vendors' names, logos, trademarks, page text or screenshots, which stay theirs, and it does not cover the Value Index or Donezo Inbox's own name and logo.

Vendors, editors and readers are welcome to send a correction to ankur@kmfv.cc, ideally with the vendor page address and the figure you see. The row is read again and updated with its new date, and the page follows.