Process Length as a Metric: Why Seven Steps Work and Twenty Don't
The client-engagement process guideline at our agency had eight steps: request, scoping, spec, effort estimate, approval, delivery, handover, three days for acceptance. Everyone knew it by heart, including new hires in their second week.
That wasn't good writing. It was length. A process guideline nobody can reproduce from memory doesn't get opened during work — people follow their memory of it instead, and memories differ from person to person. Document length is therefore not a stylistic question. It's a measurable predictor of whether the guideline gets followed.
A process guideline gets followed to the extent it can be recalled in full at the moment of work. Experimental psychology gives a benchmark: the number of elements reliably held in working memory is closer to four than to the popular figure of seven. The practical takeaway isn't "write shorter" — it's "group steps into blocks that match the decisions being made," and move rare cases out of the core document into a separate reference guide.
What happens when a process guideline gets too long
A long document doesn't get ignored outright — that would be too visible. It degrades in stages, and each stage looks perfectly normal on its own.
First, people stop opening it. Not out of laziness: opening a twenty-page document to check one step costs more than asking a colleague. Once a colleague becomes the primary source, the document stops shaping how work actually gets done.
Next, the versions in people's heads diverge. Everyone remembers their own set of steps — usually the ones they've applied most often. The gap surfaces not right away, but at the first unusual case, when it turns out two people understood the process differently.
Then local exceptions appear. "We do it differently for this client" — and that's not a violation, it's a reasonable adaptation to a general document that doesn't cover a specific case. A year later, there are more exceptions than rules.
Eventually, the document gets rewritten. The new version is longer than the last one, because it absorbs every exception that accumulated. The cycle repeats, and each round makes the document less workable.
| Stage | What it looks like | What actually happened |
|---|---|---|
| Stopped being opened | "Everyone already knows the routine" | A colleague became the source, not the document |
| Versions diverged | Dispute over an edge case | There's no longer a single process |
| Exceptions appeared | "It's different for this client" | The document doesn't cover reality |
| Rewritten, longer | New version runs 30 pages | Compliance dropped another notch |
What matters is that no one makes a mistake at any stage. Every action is reasonable in its own context, which is exactly why the cycle repeats in organizations of every size.
How many steps a person actually retains
The number "seven" has become management folklore, and it's worth going back to the primary source — it's stated far more carefully there than in the retellings.
The span of immediate memory for one-dimensional stimuli and recall lists sits consistently close to seven elements — "chunks." The author himself calls the recurrence of this number across different tasks an "eerie Pythagorean coincidence," not a law.
I'll name the limitation myself: this is a 1956 review, not a single-sample primary experiment, and the author explicitly warns that "seven" is a rhetorical device. Using it to justify "a process guideline should have seven steps" is incorrect. What's valuable is a different concept from this paper — the chunk: the unit of memory is a meaningful block, not an individual item, and grouping increases how much you can hold.
Once the effects of chunking and rehearsal are stripped out, the true limit of the central capacity of short-term memory is around four elements, not seven.
The full text is paywalled, and I worked from the abstract and method description; the paper was published in a journal with open peer commentary, where several colleagues disagree with this estimate. This isn't consensus — it's a strong position in an ongoing debate. But the practical conclusion survives the debate: what people reliably retain is lower than commonly assumed, and grouping into blocks beats trimming the list.
What happens to long mandatory checklists
There's a recent measurement of how long mandatory checklists actually get completed — at scale, with a transparent method.
Across a set of 73,922 responses to a mandatory checklist, 44.9% of the justifications for negative answers were rated as weak or perfunctory — brief or blank; 53% of authors dismissed the risks of their own work with no substantive analysis.
This is an unreviewed preprint, and the subject is academic publishing, not a company's operating procedure — applying it elsewhere calls for caution, and I'm flagging that caution here. But the mechanism is recognizable: a long mandatory checklist gets completed pro forma, and the share of pro forma answers runs into the tens of percent. Mandating something doesn't create checklist compliance. It creates checkmarks.
How to cut length without losing substance
The practical method isn't crossing out items — it's splitting the document into three parts with three different fates.
The third part is the most underrated. A requirement built into a form as a mandatory field, or as a forcing function that blocks moving to the next step, doesn't need to be remembered at all. Every requirement moved from the document into the system shrinks the core sequence without losing any content — the same principle also governs the order in which you roll out systems.
Splitting the document into a core sequence and a reference guide produces a second, less obvious effect. You can test the core sequence for compliance: ask two employees to list the process steps independently. Wherever their answers diverge is exactly the part that isn't actually being followed.
Four steps to a process guideline people actually follow
Ask two people to list the steps from memory
The gap between their answers marks the edge of compliance more precisely than any document review.
Move rare cases into a reference guide
If a situation comes up less than once a month, you don't need to remember it — you need to be able to look it up.
Turn prohibitions into system-level forcing functions
A mandatory field, a block on advancing to the next step, an automated check. This removes items from the document entirely.
Group what's left into blocks
Not "cut it down to seven" — combine steps into meaningful blocks: intake, delivery, handover. A block is remembered as a single element.
If two employees can't independently list the same process steps, you don't have a process guideline — you have a document about one.
Frequently asked questions
How many steps should a process guideline have?
As many as an employee can recite from memory in their second week on the job — and that number gets tested, not assigned. The popular "seven plus or minus two" doesn't apply correctly to process guidelines: the author of that phrase called the number a rhetorical device himself, and later estimates of stable working-memory capacity put it closer to four elements. What works isn't trimming the list — it's grouping steps into meaningful blocks.
What do I do with a long process guideline that already exists?
Split it into three parts instead of cutting it down. The first is the core sequence — the steps you need to remember. The second is a reference guide for rare situations, consulted on demand. The third is requirements built into the system as mandatory fields and forcing functions, which you don't need to remember at all. After splitting, the core sequence usually ends up two to three times shorter than the original document.
How do I check whether a process guideline is actually being followed?
Ask two employees to list the process steps independently of each other and compare the answers. It takes ten minutes and yields more information than a document audit: the discrepancies show you exactly which parts of the process exist only on paper. The second method is to check the documented order against actual system records from the past month.
Does making something mandatory help people follow it?
It helps produce checkmarks, not checklist compliance. A large-scale measurement of mandatory checklists shows the share of pro forma and blank justifications reaching nearly half. Making something mandatory works when it's built into the system as a forcing function and physically prevents doing otherwise; where it relies on discipline alone, document length determines the outcome more than any requirement does.
The eight-step process guideline (request, scoping, spec, effort estimate, approval, delivery, handover, three days for acceptance) comes from Alego.Digital's client-engagement process guideline. It's the author's own company's internal document; it hasn't been verified by an independent party and is presented here as an illustration of the mechanism. The claim that new hires knew the process guideline by heart in their second week is the author's observation, not a measured figure. Splitting a document into a core sequence, a reference guide, and system-embedded requirements is the author's own practice, not formalized in the company's documents.
- Miller G. A. The Magical Number Seven, Plus or Minus Two. Psychological Review, 63(2), 1956. musanim.com
- Cowan N. The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 2001. cambridge.org
- Jinnath N., Zhao W. An analysis of compliance with a mandatory checklist across 73,922 responses. arXiv:2608.09280, August 2026. arxiv.org