Every piece of information on a screen asks something of the people looking at it. The more you ask, the slower and less accurately they think.
Working memory is the mental workspace where we process information in the moment — the place where decisions get made, instructions get followed, and forms get filled in. The problem is that working memory is small: most researchers put its capacity at roughly four chunks of information simultaneously.
Cognitive load is the measure of how much of that workspace an interface consumes. When load is low, users move confidently and make good decisions. When load approaches the ceiling, they slow down, make errors, or abandon the task entirely. The ceiling doesn't change — it's a hardware constraint of human cognition. But how much of it a design consumes is entirely within a designer's control.
The inherent complexity of the task itself. Filing taxes, learning a language, and configuring a server are genuinely complex. No amount of good design makes them trivial — some of this load is irreducible.
Mental effort added by poor interface design. Confusing navigation, cluttered forms, irrelevant information, and unnecessary decisions all generate load that has nothing to do with the actual task. This is entirely avoidable.
Productive cognitive effort that builds understanding, skill, or memory. Learning how a product works, building a mental model, developing mastery — this is load worth spending. Good design creates space for it by eliminating the other two.
The practical implication is simple: every time you reduce extraneous load — by removing an unnecessary field, simplifying a navigation, grouping related information — you free up mental budget that users can spend on the task itself. The task doesn't get easier. But the experience of doing it does.
“Your job is not to show everything you know. It's to show everything your user needs.”
Registration forms are where cognitive load violations are most consistent. The team wants to collect information. The result is a single screen asking for email, password, name, phone number, date of birth, referral source, and marketing preferences — all before the user has experienced a single feature of the product.
Every field on a form is an independent cognitive task: read the label, recall or decide the information, validate the format, and enter it correctly. Eight fields on one screen is eight sequential cognitive tasks performed under the anxiety of uncertainty — the user doesn't know how long the form is, whether their password will be accepted, or what happens next. Progressive disclosure eliminates most of this by showing one task at a time. Try filling in both forms.
8 fields + 2 checkboxes visible at once. Most people abandon here.
One question per screen, 4 steps total. Tap Continue through all steps.
The total information collected by both forms can be identical. The difference is entirely in when and how it arrives. Spreading questions across screens doesn't make the form longer in real time — each step takes seconds. But it eliminates the anxiety of an unknown endpoint and reduces each moment to a single, low-stakes task.
Analytics dashboards are the most common site of extraneous cognitive load in B2B products. Sixteen metrics with equal visual weight and no hierarchy aren't presenting data — they're presenting a puzzle. Users who open a dashboard to answer one question (“is our conversion rate improving?”) are forced to scan a grid of equal-sized numbers just to locate it.
Visual hierarchy is cognitive load management. A dashboard that surfaces three primary metrics at large scale, with secondary metrics below, does the prioritisation work for the user. The same data is present. The brain processes far less to extract the answer.
16 metrics at equal size. Users must scan the whole grid to locate the one they need.
Same sixteen metrics. Three have been promoted to large scale so users see what matters first.
The second dashboard contains the same sixteen metrics. Not one number is missing. The difference is that three have been promoted to a size that makes them immediately readable — users know which numbers matter before they've consciously looked. The secondary metrics are still present for anyone who needs them, at a scale that signals their supporting role.
Project management tools all solve the same core task: let someone record a piece of work. The cognitive load required to do that varies dramatically depending on how the form is designed.
Some project management tools show every possible field on a single form. Before a developer can describe what they're actually building, they must navigate: project selector, issue type (bug, story, epic, task, sub-task…), summary, description, assignee, reporter, priority, label, sprint, story points, epic link, fix version, component, linked issues, and any number of custom fields added by the organisation. Most of these fields will be left blank. All of them consume working memory.
A more focused approach shows only what's necessary to capture the work — title, description, assignee, priority. Everything else is one click away if needed. The same issue gets created. The cognitive cost is a fraction. Try both forms below.
15+ fields before you’ve described the work. Most will be left blank.
4 visible fields. Everything else is a click away via “+ More”.
The focused form has an explicit design philosophy: remove everything from the issue creation flow that can be filled in later, deferred by default, or skipped entirely. The result is a form that takes under ten seconds to complete. The overloaded form can be completed just as fast — but only by people experienced enough to know which fields they can safely ignore. The focused form makes that judgment for everyone, by default.
The practical question cognitive load asks of every design is: whose job is this? Every decision currently required of the user can potentially be made by the system, deferred until later, or eliminated entirely. When you place a decision on a screen — a field in a form, a parameter in a settings panel, a metric in a dashboard — you are choosing to spend your user's cognitive budget. That budget is not infinite and it's not recoverable once spent. Spend it only on things that genuinely require the user's judgment.
AI has made this question more answerable than ever. Configuration decisions that previously required user expertise can now be handled by a model. Visual hierarchy decisions that previously required editorial judgment can now be generated from data. The trend in product design is not toward more powerful interfaces — it's toward interfaces that absorb more of the cognitive work themselves so users don't have to.
Sweller, J. (1988). Cognitive load during problem solving. Cognitive Science, 12, 257–285. · Miller, G. A. (1956). The magical number seven, plus or minus two. Psychological Review, 63(2), 81–97.