Miller's law on a landing page: the real point isn't "7 items", it's chunking

There's a rule that has circulated in design circles for decades: "no more than seven items in a menu, seven plus or minus two, that's Miller's law." The rule is reasonable in its practical consequences, but its foundation is wrong: the 1956 paper isn't about elements visible on a screen, it's about how much information a brain can hold at once for a few seconds. And its most useful conclusion for a landing page isn't the number seven — it's the finding that grouping information into meaningful units, chunking, lets far more fit into a very small mental space. That nuance shifts the question entirely: instead of counting the items on a page, you ask what the page forces the visitor to memorize.

What the 1956 paper actually says

The founding paper is "The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information", published by George A. Miller in Psychological Review (1956, vol. 63, pp. 81-97). Miller pulls together work on two distinct capacities: absolute judgment (how many levels of a stimulus — the pitch of a tone, the intensity of a taste — you can tell apart without error) and immediate memory span (how many items you can repeat back after a single hearing). He observes that both plateau around the same zone, somewhere near seven, and he openly jokes about it: he speaks of being "persecuted by an integer" rather than of a law of nature. The genuinely important passage comes next: Miller shows that the limit isn't on the amount of raw information but on the number of units, of chunks. A string of twenty-one binary digits can't be held as such; grouped in threes and recoded, it becomes seven units and becomes memorable again. In other words, capacity is fixed, but what you put inside each unit is not — and that's where design has room to work.

The "7 ± 2 menu items" myth

Two errors crept into the popular version of the law. The first is the number. Later research revised the estimate sharply downward: in "The magical number 4 in short-term memory: a reconsideration of mental storage capacity" (Behavioral and Brain Sciences, 2001, vol. 24, pp. 87-114), Nelson Cowan shows that once you neutralize mental rehearsal and grouping strategies, the real capacity of working memory is closer to three to five units. Miller's seven was a high estimate, obtained in conditions where participants could already chunk. The second error is more serious: the limit is about what has to be remembered, not what is displayed. A comparison table with twelve rows visible at once saturates no working memory, since the visitor can re-read; but asking someone to compare a feature in row 2 with a price displayed three screens higher saturates it immediately. That's the distinction drawn by cognitive load theory, formalized by John Sweller in "Cognitive Load during Problem Solving: Effects on Learning" (Cognitive Science, 1988): what costs isn't the amount of information present, it's the mental work required to relate the pieces to each other. The article on cognitive load on a landing page covers this point, and the one on the navigation menu shows why the real reason to cut the menu has nothing to do with a magic number.

Consequence #1: ask for no memory effort at all

If working memory holds three to five units for a few seconds, the practical conclusion is the reverse of the myth. It isn't about trimming what's displayed to fit under a quota, it's about making sure the visitor never needs to remember anything in order to decide. Every time a decision depends on information read higher up and now off-screen, the page is betting on a memory that won't support it — and the visitor, rather than scrolling back, gives up or hesitates. The fixes are mechanical and cost almost nothing.

• The price stays visible at the moment of commitment — repeat the price, the duration, and what's included right next to the button, instead of relying on the pricing table two sections above, and do it at every repeated button on the page.
• The form restates the offer — a bare signup form with no reminder of what you're signing up for forces the visitor to hold the offer in memory while typing an email address. A one-line summary above the fields is enough to offload that.
• No comparison at a distance — if two pieces of information have to be related (a price and what it includes, a feature and its benefit), they must be physically adjacent, not separated by a screen of scrolling.
• Labels make sense on their own — a "Continue" button requires remembering what you're continuing; "See all 10 templates" makes sense with no prior context.
• Nothing to copy over — a reference number, a promo code, or an email to carry from one screen to the next is the most expensive case: it has to be shown at the moment it's needed, or pre-filled.

Consequence #2: chunking, the actual technique

The other side of Miller's paper is constructive: since the limit is on the number of units and not on the amount of information, you can push through far more content by grouping. Twelve product benefits lined up as a twelve-bullet list read as twelve things to process; the same twelve benefits split into three titled blocks of four — "Save time", "Reassure your prospects", "Stay flexible" — read as three things, each one developed. The content is identical, the perceived effort is divided. The same mechanism explains why a phone number formatted 06 12 34 56 78 is easy to hold and re-read without error, while 0612345678 forces you to count digits: visual grouping creates the units on the reader's behalf. On a landing page, this applies to form fields (phone, card number, tax ID), to section headings that act as labels for a bundle of paragraphs, and to text formatting — a ten-line wall of text is one unreadable unit, the same text in three short paragraphs becomes three. The textbook case is the form: twelve fields on a single screen present twelve units, whereas the same form split into three thematic steps of four fields presents only four at a time, each carrying a label that gives it meaning — that's the reasoning in the article on the multi-step form, with one important caveat: splitting only helps if the groups are meaningful, since three steps of four randomly mixed fields lighten nothing and merely add clicks. And chunking doesn't remove the prior question, covered in the article on how many form fields to use: the best grouping is still deleting the field nobody needs.

Miller's law or Hick's law? Two laws often confused

Both push you to simplify, but they don't describe the same phenomenon and they aren't fixed the same way. Hick's law is about decision time: the more options there are, the longer the choice takes, and that slowdown happens even if every option is visible and there's nothing to memorize. Miller's law is about retention capacity: it describes what stays in mind for a few seconds, regardless of how many choices have to be made. A page can fail on one without failing on the other: three very similar offers saturate the decision (Hick) while demanding nothing of memory; a single offer whose price sits a thousand pixels above the button demands no choice but saturates memory (Miller). The remedies differ: reduce or rank the options in one case, bring the information closer together and group it in the other. The question of a landing page with several offers clearly belongs to the first category, while the three-column pricing table belongs to both — three columns limit decision time and form three memorable units.

Test it empirically rather than counting items

None of these principles can be verified by counting elements in a mockup: the only useful check is measuring what a visitor actually retains. The 5-second test does exactly that: show the page for five seconds, hide it, then ask what the company offers, to whom, and what to do next. What the person reports back is, near enough, the contents of their working memory — three or four units at best. If the main promise isn't among them, the problem isn't the amount of information displayed but its grouping and hierarchy, particularly in the upper zone described in the anatomy of a landing page that converts. The corollary is that you have to decide in advance which three units you want to survive — because beyond that, nothing will.

LanderKit templates (€89 each, €229 for the pack of 10) are built on this principle: benefits already grouped into titled blocks, the price restated next to every button, forms split by theme — and no navigation menu to scatter attention. It's especially visible on SaaS Waitlist, where the offer fits into three units before the form, and on Training Course, where the curriculum is chunked into modules rather than lined up as an endless list.