Splitting a song three ways: what to do with the rounding

26 · 2026-08-24 · 13 min read · Español

Write 33.34, 33.33 and 33.33, and put a name next to the 33.34 so everyone knows whose hundredth it is. One hundred does not divide by three and it never has, in any business, and the music industry solved it the way every industry solves it — with a tolerance and no announcement. The rounding is not what costs you money. Three people each solving the same remainder privately, in three different portals, is.

• An even three-way writer split is 33.34% / 33.33% / 33.33% on the convention where all writer shares total 100%.
• Shares do not have to total exactly 100%. ASCAP and BMI reconcile a work at 100% or within plus or minus 0.06%, and The MLC allows the same tolerance on a CWR registration.
• Three shares of 33.33% total 99.99%, which is inside that tolerance in either direction.
• What stops money is conflict, not rounding: when co-writers file different numbers, the societies hold a work they cannot reconcile.
• Claiming shares totalling more than 100% at The MLC is an overclaim, and resolving it is the claimants’ own responsibility.

What are the numbers for an even three-way split?

33.34%, 33.33% and 33.33%, on the convention where all the writer shares on a song total 100%.

You have probably seen those three numbers before. Our template prints them, along with the 100% and 200% conventions and every other field a sheet needs, and it does not explain them because a form is not the place to argue. This is the place.

One thing has to be added to those three numbers before they are useful, and it is not a number. The sheet has to say which writer takes the 33.34. Three people are going to type this song into three different systems, possibly months apart, and the one fact that cannot be reconstructed from memory is whose hundredth it was.

Everything below assumes the three of you already agree on an even split. If you do not — if the argument is whether the person who suggested the title is a writer at all — that is a different question and a harder one, and no arithmetic settles it. And if nobody writes anything down, the law divides the song for you in equal undivided shares, which is a sentence with no decimal places in it and no help at all when you reach a registration form.

Do the shares have to add up to exactly 100%?

No. ASCAP and BMI reconcile a work at 100% or within plus or minus 0.06%, and The MLC allows the same tolerance on a CWR registration.

This is the part nobody tells you, and it is published in both places. Songview is the database ASCAP and BMI built together so that one work has one set of ownership data across both:

A work will have the Songview checkmark only if all of the share percentages for the work total 100%, or are +/- .06%, which is the industry standard tolerance.

The MLC says the same thing in its own instructions, in the document that governs bulk registrations. From The MLC’s CWR User Guide:

The total ownership shares for all writers and publishers for each right must total either 100% plus or minus 0.06% or 0%.

Now do the arithmetic that has been bothering you. Three shares of 33.33% total 99.99%. That is short by one hundredth of one percent, and one hundredth is inside 0.06 by a distance you could park a bus in.

The whole industry’s answer to a number that will not divide is a shrug, formalised.

Two things neither source says, stated plainly and once. Neither publishes why the tolerance is 0.06 rather than some other number. And neither says what an individual society’s own web form does at the moment you type 33.33 into it three times — the tolerance is documented for reconciliation and for file-based registration, not as a promise about every text box on the internet. So write numbers that total 100.00 anyway. Not because 99.99 breaks anything, but because a total that is already correct never has to be defended by a person who has read this page.

What happens if the three of you file different numbers?

The societies end up holding conflicting information about one work, and the money waits while the people claiming it agree.

This is the failure. Not the hundredth — the three of you, each perfectly reasonable, each rounding at your own kitchen table. One writes 33.33 across the board. One writes 33.34 for everybody. One decides to be generous and takes 33 flat. Nobody has done anything wrong and the song now exists in three versions.

Songview states what it takes for a work to reconcile at all: the checkmark means “both ASCAP and BMI have the same information regarding the songwriters, composers and publishers on a work, as well as the shares of the work belonging to each.” And one of the two named reasons a work does not reconcile is that the information provided “is in conflict.”

BMI, in its Royalty Policy Manual, describes what a second, disagreeing registration triggers:

If a later registration is received for the same work which conflicts with the earlier registration, we will notify the party submitting the later registration and request documentation or written confirmation from all affected participants in conformity with BMI’s conflicting registration rules before changing our records.

Read that again with three co-writers in mind. All affected participants. Somebody now has to collect written confirmations from two people who have not thought about this song since the mix.

At The MLC the same disagreement has a name when it runs the other way. An overclaim “occurs when multiple claimants are claiming shares totaling over 100% on a musical work.” Every claiming member gets an automated email, and The MLC is direct about whose problem it is: “it is the responsibility of those claiming shares on the work to resolve the overclaim.”

Nobody calls you. Nothing is flagged in red on a dashboard you check. The work simply sits in whatever state the last submission left it in, and a collection period can get pissed away proving that three people who agree with each other agree.

If the numbers on file are already wrong, changing them is a procedure with an order to it and it is more work than getting them right was. And if the mismatch turns out not to be clerical — if one of the three actually thinks the split is different — you are no longer reading the right page. That is a disagreement, and it is treated as one.

Does it matter which way you round?

Rounding a hundredth up or down keeps the total inside the industry’s published tolerance of plus or minus 0.06%.

Check it both ways. Three shares of 33.33% is 99.99%, one hundredth under. Three shares of 33.34% is 100.02%, two hundredths over. Both are inside the tolerance, in a document that says plus or minus and means it.

What is not inside it is three writers each rounding to a whole number and each taking 34%, which claims 102% of a song. That is not a rounding error. That is an overclaim, with a name, an automated email and a resolution duty attached to it.

So the rule is not round down and it is not round up. It is: pick three numbers that total 100.00, agree them once, and stop relying on a tolerance to catch a thing you could simply not do. The tolerance exists for the industry’s convenience, not as a budget for you to spend.

Do you type the same numbers into every system?

Yes — the agreed percentages never change, but BMI, The MLC and SOCAN each ask for them in a different shape.

This is where a reader who has done everything right starts to doubt themselves, because the third form does not look like the first two. It is the same agreement every time. The boxes are different.

BMI counts the song as 200%. The manual: “BMI considers payments to songwriters or composers and to publishers as a single unit equal to 200%.” The writers share one 100% and the publishers share the other, so your 33.34 is 33.34 of the writers’ half and not of the whole. The template explains both conventions, and the reason it explains them is that “50%” means two different things depending on who is asking.

The MLC asks a publisher for a collection share. Its instructions say the figure “must be between 0.01% and 100%,” per right, per territory. Two things follow. That is a collection number rather than an authorship number — what somebody is entitled to collect, not who wrote what. And the floor is one hundredth, which is the real reason 33.333 is not an answer to anything: there is no box on earth that will take it.

SOCAN takes the division you give it. From its frequently asked questions: “You can register the song, indicating the division of ownership shares, and we’ll distribute your portion to you.” Your portion. Not the song’s — yours, which is a useful reminder that a society pays what it was told about you and takes no view on what it was told about anybody else.

Where each of these registers sits and what it actually pays out is its own map. What matters here is narrower: one agreement, three forms, and the same three numbers surviving all of them.

What do you actually write on the split sheet?

The three exact percentages, beside the three names, with the writer who takes the extra hundredth named as that writer.

Not “equally.” Not “in thirds.” Not “split three ways.” Every one of those is a word that has to be turned into a number later, by three people, separately — which is the mechanism that produces three registrations that do not match. A word that needs arithmetic done to it is not a decision. It is a decision postponed and then made three times.

Two more habits, and they cost nothing:

• The same writer keeps the extra hundredth everywhere. It does not rotate between registrations to be fair. It is worth one hundredth of one percent, and moving it around is how a matching set of filings becomes three that disagree.
• Write it while the song is still fun. The template exists to be filled in at the session, not recovered from a group chat in eighteen months.

What makes any of it stick is a separate question with a separate answer, and it is worth knowing before you assume a signature does more or less than it does.

What if it is six ways, or seven, or a third of a half?

Same rule at any size: exact percentages that total 100.00, with the leftover hundredths given to one named writer.

Every split that does not divide evenly has the same shape and the same fix. Here are the ones that turn up, with numbers you can copy. Every row totals exactly 100.00 and every row is arithmetic you can check yourself in about four seconds, which is the only kind of number on this page.

The last row is the one that catches people, because the remainder appears inside a half rather than inside the whole song. It behaves identically. Split 50 into three, one of the three takes the extra hundredth, and the four numbers total 100.00 the same way the three did.

Go and open the last song you registered with somebody else. Add up the writer shares as they are actually filed, at every society you filed them with, and see whether the three numbers are the same three numbers in all of them. If you would rather do the whole catalogue at once, there is a checklist for that — and if it turns out you have been carrying a mismatch for two years, the good news is the fix is a form and not a lawyer.

FAQ

Is an even three-way songwriter split 33.33% each, or 33.34/33.33/33.33?

Three shares of 33.33% total 99.99%, so one writer takes 33.34% and the three numbers reach 100.00%. Which writer takes the extra hundredth is a decision, and the split sheet should name them.

Do songwriter splits have to add up to exactly 100%?

Not to the hundredth. ASCAP and BMI reconcile a work whose shares total 100% or fall within plus or minus 0.06%, and The MLC allows the same tolerance on a CWR registration. Writing numbers that total 100.00 is still the right habit, because no society publishes what its own entry form accepts.

What happens if co-writers register different splits for the same song?

The societies end up holding conflicting information about one work, and the money waits. BMI requires documentation or written confirmation from all affected participants before it changes its records, and a total claimed above 100% at The MLC is an overclaim that the claimants themselves have to resolve.

Who should get the extra hundredth of a percent?

No society publishes a rule, so the writers choose. Pick one at the session, write the three exact percentages on the split sheet, and keep the same writer on the extra hundredth in every system you file it into.

Sources

• ASCAP and BMI, Songview — that a work carries the Songview checkmark only where the share percentages total 100% or fall within plus or minus .06%, described there as the industry standard tolerance; that the checkmark means both societies hold the same writer, publisher and share information; and that one of the two common reasons a work fails to reconcile is information in conflict. ASCAP’s own help pages refuse non-browser requests, so the ASCAP side of this piece is carried by the database the two societies publish jointly.
• The MLC, CWR User Guide (May 2024) — that total ownership shares for all writers and publishers for each right must total either 100% plus or minus 0.06%, or 0%, and that the same tolerance applies to collection shares. The PDF opens and reads normally; its text would not extract on the machine this piece was written on, so it was read by decoding the document’s own character maps, and the same tolerance is carried independently by the source above.
• The MLC, how to register works in the MLC Portal — that a publisher’s collection share must be between 0.01% and 100%. The floor is why a share cannot be stated in thousandths.
• The MLC, overclaims and the Overclaims Tool — that an overclaim occurs when claimants claim shares totalling over 100% on a musical work, that every claiming member is emailed automatically, and that resolving it is the responsibility of those claiming.
• BMI, Royalty Policy Manual — that BMI treats payments to writers and to publishers as a single unit equal to 200%, and what a conflicting later registration triggers.
• SOCAN, frequently asked questions — that you register the song indicating the division of ownership shares, and SOCAN distributes your portion to you.

This page is about arithmetic and registration, not about who owns anything. What a share should be, who qualifies as a writer, and what a signature does are separate questions with their own pages here. One limit worth stating rather than leaving you to find: the plus or minus 0.06% tolerance is published by ASCAP and BMI for reconciliation and by The MLC for CWR registrations. No equivalent figure is published by SOCAN, CMRRA or Re:Sound that this piece could open, so nothing above claims one exists in Canada. Every page cited here was read on the date at the top of this piece.

Keeping the register

The reason this is hard is that the numbers live in four places and only one of them is a piece of paper you control. CatalogTracker keeps master splits and publishing splits per track, as percentages summing to 100, enforced on the server when a split is shared for confirmation — which is stricter than the tolerance this piece just spent a section on, and deliberately so: a document generator that lets you save 99.99 is a document generator that hands you an argument later. Its confirmation is a cryptographic attestation from a device, not an electronic signature and not proof of identity, and the app says so itself rather than letting you assume. In development for iPhone.