The annual discount you cannot check: when 'two months free' is provable and when it is not
Paying yearly is the most widely advertised discount in software, and whether it can be verified at all depends on one field: the yearly total. Thirty-one plans in our record quote a per-month annual rate and publish no annual total anywhere, which leaves the saving unprovable. This guide works through a vendor whose arithmetic closes exactly, a vendor whose billing toggle changes the label and not the figure, and a third whose identical-looking numbers mean something else entirely.
Almost every subscription page offers a discount for paying yearly, and almost none of them is presented in a form you can check. There are two numbers involved: a per-month rate that applies if you commit to a year, and the total that will actually be charged. Publishing the first without the second is common, and it makes the discount a claim rather than a figure. In our record, fifty-seven plans publish both, thirty-seven publish only the yearly total, and thirty-one publish an annual per-month rate with no yearly total anywhere on the page.
Migadu shows what the checkable version looks like. Its entry mailbox plan is $9 a month, or $90 for the year. The difference is $18, which is exactly two months of the monthly rate, and the same relationship holds on every rung: the yearly figure is precisely ten times the monthly one across the whole ladder. The vendor's 'pay ten months, get twelve' framing is not a slogan there, it is a description of two printed numbers, and anybody can confirm it in a few seconds without trusting us or them.
Tally advertises the same offer and it cannot be confirmed at all. Its billing switch offers a monthly state and a yearly state labelled two months off, and the per-month figure displayed is identical in both. The label promises a discount that the number beside it does not show. No annual total is printed anywhere on the page, so the actual yearly charge is not readable, and the saving cannot be tested against anything. It is the first decorative toggle we recorded, and it is the reason the capture rule is to walk every toggle and record every state rather than read whichever one the page happens to open on. A page opening in its discounted state and a page whose discount is only a label look identical to a scraper that visits once.
The reasonable inference — that the yearly state simply bills the same monthly figure twelve times, offer notwithstanding — is not one we publish, and the restraint is the point. The page does not print a yearly total, so we do not hold one, and inventing the obvious one would put our arithmetic in a column of vendors' prices. What is recorded is what was seen: two states, one figure, one label, and no total. A reader can draw their own conclusion from that, and it is a better position than being handed ours dressed as a reading.
Now the complication, because the same numeric signature has more than one cause. Internxt's three personal tiers also record an annual per-month rate identical to the monthly one, which looks like Tally's case and is nothing like it. There is no monthly plan on that page at all: the states are annual and lifetime, the page opens on lifetime so the first figures a visitor sees are one-time, and the annual plan is billed monthly. The advertised per-month figure is a first-month rate, and the page's own note under each plan says it renews at roughly five times that from month two. Equality between the monthly and annual fields there means the plan has one recurring rate, not that a discount is fictional. Two vendors, the same shape in the data, opposite meanings, and no way to tell them apart except by reading the note beside the number.
There is a third case, where the per-month figure exists but nobody at the vendor wrote it. Eight plans in our record carry a provenance string on the annual rate, recording that the vendor printed only a yearly total and the monthly equivalent was produced by dividing it by twelve. That string is kept in the record on purpose, so a figure that came out of our arithmetic is never mistaken for one that came off a page. It is a small piece of bookkeeping and it is the difference between a per-month rate the vendor is offering and a per-month rate we worked out for readability.
None of this argues against paying annually. Where it is real it is usually the best-value decision on the page, and Migadu's ladder is a fair example of a vendor making the case honestly by printing both figures and letting the arithmetic speak. The argument is about what to do when only one of the two numbers is published. In that situation the discount is not verifiable, and the correct response is neither to assume it is fake nor to take it on faith, but to note that the vendor has chosen not to publish the number that would settle it, and to price the decision on the figure they did print.
The yearly total is the field that makes a discount checkable
Fifty-seven plans in our record publish both a per-month annual rate and a yearly total. Thirty-one publish the rate alone, and for those the actual annual charge is not readable from the page. A saving quoted without the total it is a saving from is a claim rather than a figure.
Migadu's arithmetic closes exactly
$9 a month against $90 a year is a saving of $18, which is two months to the penny, and the yearly figure is ten times the monthly one on every rung of the ladder. The 'two months free' claim is confirmable by anyone in seconds.
A toggle can change the label without changing the number
Tally's billing switch offers a yearly state advertising two months off, and the per-month figure it displays is identical to the monthly state's. This is why every toggle gets walked and every state recorded: a page whose discount is only a label is indistinguishable from a genuine one if you look once.
Identical monthly and annual figures have more than one cause
Internxt records the same equality as Tally for an unrelated reason: it sells no monthly plan, its annual plan bills monthly, and its advertised figure is a first-month rate that renews around five times higher. The data looks the same; the meaning is opposite; only the note beside the number separates them.
A derived per-month rate is labelled as derived
Eight plans carry a provenance string recording that their annual per-month figure came from dividing a printed yearly total by twelve. Keeping that string means a number we computed for readability is never presented as a rate the vendor quoted.
An unpublished total is not filled in with the obvious answer
Where a yearly state shows no total, the likely charge is easy to guess and we do not record a guess. The record holds the states, the figures and the labels that were actually on the page, which lets a reader reach the obvious conclusion themselves without us having laundered it into a price.
Questions people actually search
How do I check an annual discount myself?
Click the toggle both ways and compare the figures rather than the labels, then look for a yearly total. If the total is printed, subtract it from twelve times the monthly rate and see whether the difference matches what the badge claims. On a ladder like Migadu's it will match exactly. If no total is printed, the check cannot be completed, and that is worth knowing before you commit to a year.
If the toggle shows the same price in both states, is the discount fake?
Not necessarily, and that is the awkward part. It may be a decorative toggle, or it may be a page that sells no monthly plan and bills its annual plan monthly, in which case one recurring rate is the honest presentation. We record both cases identically because the figures are identical, and the difference lives in the note printed beside them. Reading that note is the only reliable way to tell.
Why not just multiply the monthly figure by twelve when no total is published?
Because the result would be our arithmetic sitting in a column of the vendor's prices, and a reader has no way to tell the two apart once they are printed in the same typeface. Where we do compute a per-month figure from a printed yearly total, the record keeps a note saying so. The reverse direction is not done at all, because an annual term can carry conditions that twelve monthly payments do not.
Is paying annually usually worth it?
Where the discount is real it is often the best-value option on the page, and two months free is a common and genuine shape. The caution is narrower than a rule against annual billing: a discount you cannot verify sits alongside a commitment you cannot easily exit, so the case for taking it should rest on a figure the vendor published rather than on a badge.
Do you earn a commission from any vendor on this page?
No. We hold no affiliate relationship with Migadu, Tally or Internxt, and this page carries no affiliate links. Migadu appears as the example that makes its claim checkable, which is a favourable mention nobody paid for. Our method is on the method page.
Every price on this page resolves from a dated vendor record rather than being typed into the text; the records cited were read between 2026-08-27 and 2026-09-03. Counts of which plans publish a yearly total are taken from the record as it stood on 2026-09-03. Where a vendor printed no annual total, none is stated or inferred here. This page carries no affiliate links and we hold no commercial relationship with any vendor named on it. Related: annual versus monthly, how to read a price record and why a renewal is not the sticker price. The method →
Found a price that no longer matches the vendor’s page? Tell us — corrections are dated and stay on the page.