Robin: one publisher, two spellings, two ad populations

A proper name seems the most stable of data. This is what makes him a traitor. Unlike a number, which we distrust because we know that it can start again, or a state, which we know is subjective, the name of a publisher passes for a fact. We copy it without thinking about it, we filter on it, we regroup on it.

But an accent is a character like any other. Two strings that differ by an accent are two different strings, and nothing in a sort, filter, or search will spontaneously bring them together. What follows measures the extent of the phenomenon on a specific title on the French market, and draws the practical consequences.

Forty-five against twenty-seven

I surveyed the French market on August 30, 2026 on three queries around the character, then duplicated the results. Among them, 78 ads concern the same old title bringing together three characters, including Robin. The amounts quoted aresums claimedthat day: no sales are observed, and a reading taken on a single date does not show any development.

Of these 78 ads, the publisher's name appears 45 times without an accent and 27 times with it. The two spellings designate the same house, on the same title, often for similar numbers. Neither is marginal: the less frequent still represents more than a third of the mentions. So there is no “correct” form to fall back on, nor of a form rare enough to be neglected.

An announcement of the same title also names an entirely different publisher, for a series presented as complete from number 1 to number 16. I refrain from drawing a conclusion about who published what: a sales notice is not a publisher's document. What this isolated case shows is enough: the publisher field of an ad is no more reliable than the others.

This last point deserves attention, because it settles a question. If the bindings were numbered in the order of the issues they contain, two consecutive blocks would have consecutive numbers. This is not the case here: seven, then eighteen. The binding number is therefore not deduced from the content, and the content is not deduced from the binding number. The two pieces of information are independent and must be noted separately.

What an accent does to a search

You have to be precise about the mechanism, because intuition is misleading. Many search tools apply a normalization that ignores accents: typing one form then brings back both. Many others don't. And above all, a personal inventory almost never does this: a text field compares strings, and two strings that differ by one character are different.

The problem is therefore not so much research on a marketplace — where behavior depends on the tool — as what happensAfter, in your own base. Twenty copies entered by copying the advertisements will produce two publisher entries instead of one, distributed randomly according to the wordings encountered. Grouping by publisher will display two lines. One count will give two numbers. None will be false in the strict sense: they will count exactly what they have been given.

The same remark applies to the other variations that we encounter on this title: the mention of a second series, present in eleven advertisements, a “new series” out of two others, a “first series” on one. These formulations probably designate overlapping divisions, but they are written differently, and it is the writing which governs the arrangement.

Finally, a word on how this defect occurs, because it explains why it goes unnoticed for so long. No one enters twenty cards at once by comparing the labels: we register one copy today, another in three weeks, a third the following spring. Each time, we copy the announcement we have in front of us, and each time the gesture seems right. It is only when looking at the entire column, a long time later, that the two spellings appear side by side.

This is the signature of the most costly data defects: they do not arise from an error, but from the accumulation of correct actions taken one by one. No entry is at fault; it is their juxtaposition that is. And since nothing ever failed, no error messages were ever displayed, there is no record of when the divergence started.

Six consequences in an inventory

What this statement does not establish

It doesn't say which spelling is correct. Both circulate, one more than the other, and an announcement is not a source on the official spelling of a company name. The question, moreover, has no practical importance: what matters is to know that two forms exist, not to crown one of them.

Nor does he say how many bindings were published, or according to what plan. Only four advertisements mention a collection or an album; this is enough to establish that two designation systems coexist, far too little to reconstruct an overall logic. I therefore stop at the observation, and I am careful not to extrapolate a binding plan from two issues.

Finally, the price difference between the two lots – €30.00 and €24.99 for the same four numbers – cannot be interpreted. Condition, presentation, seller, age of the ad: too many factors come into play for a difference of five euros to mean anything. I cite it because it helps identify the two announcements, not because it teaches anything.

Translate this into a file

The rule:standardize the proper nouns at the time of entry, once and for all, and note the spelling encountered separately.Choose a publisher form and stick with it, regardless of what the ad is. It doesn't matter which one, as long as it is unique in your database: it is the uniqueness that makes the groupings usable, not the typographical accuracy.

This standardization is the exception to a principle that I have often defended elsewhere — that of recopying what is printed without retouching it. The exception is justified because a publisher's name does not identify a copy: it classifies it. Classification data must be homogeneous to be useful, whereas identification data must be faithful. Keep the spelling of the ad in your notes if you want to, but don't let it rule the arrangement.

For bindings, the instructions differ: they require two fields, not one. The binding number on one side, the list of issues contained on the other. Without the second, you will never be able to know if a booklet offered individually is already at home in bound form - and this is exactly the kind of duplicate that you discover when you reopen the boxes, years later.

When the ad does not provide this list, ask for it before purchasing, or explicitly note that it is unknown. A binding whose contents are unknown is not an inventory line: it is an object waiting to be described. Noting it as such costs a few words and avoids passing off ignorance as data.