In the 1950s, changing a die on one of Toyota's large stamping presses, swapping the tooling to produce a different car part, could take the better part of a day. The machine sat idle the entire time. Shigeo Shingo, the engineer Toyota brought in to fix this, didn't start by asking workers to move faster. He started by categorising every single step in the changeover into one of two types: steps that could only happen while the machine was stopped, and steps that could happen while it kept running.
That distinction, internal setup versus external setup, is the entire insight behind SMED, Single-Minute Exchange of Die. Once Shingo separated the two, he moved as many internal steps to external as physically possible: pre-staging the next die nearby while the current one was still running, pre-heating tools, pre-assembling fixtures. Changeover times fell from hours to single-digit minutes, without a single piece of new equipment. The machine wasn't faster. The sequencing was.
What SMED actually changed
The genuinely useful part of SMED isn't the outcome, fast changeovers, it's the method that got there: a hard, honest audit of which steps truly require the machine to be stopped, and which only seemed that way because nobody had questioned the assumption. Most of what Toyota classified as "internal" before Shingo's audit turned out to be external once someone actually checked.
A B2B conversion form has exactly the same structure, and exactly the same unexamined assumption. Every field feels like it has to happen right there, at the moment the visitor is filling it in, because that's simply when forms have always been filled in. Most CRO work accepts that assumption and tries to make the internal process faster: shorter labels, fewer clicks, a cleaner layout. Shingo would ask a different question first.
SMED's insight was never "work faster while stopped." It was "question whether stopped was ever necessary in the first place." Most CRO effort skips straight to the first question without asking the second.
Why most CRO effort attacks the wrong half of the process
Field reduction is the most common CRO fix applied to lead forms, and it works, but it's optimising the internal process rather than questioning why so much sits inside it. A five-field form is faster to complete than a nine-field form. It still asks the visitor, at their single moment of highest intent and highest drop-off risk, to do work that in many cases the business could have done for them beforehand.
Applying external setup to a B2B lead form
The practical version of Shingo's audit is straightforward: for every field on the form, ask whether the answer already exists somewhere, or could exist, before the visitor arrives.
Company name, size, and industry can often be inferred from the visitor's IP address or email domain the moment they land on the page, before they've typed a single character. Pre-filling or entirely removing these fields moves genuinely internal work to external.
If a visitor arrived through a LinkedIn campaign targeted at a specific job function or industry, that targeting data can be passed through the URL and used to pre-populate or skip the equivalent form field, rather than asking a question the ad platform already answered upstream.
A visitor who's filled in a form on a previous session shouldn't have to re-supply information already on file. Recognising them and pre-filling known fields is external setup applied to repeat traffic specifically.
What has to stay internal, and how to make it faster
SMED never eliminated internal setup entirely, some steps genuinely can't happen while the machine is running. Its second lesson, after moving what can be moved, is to streamline whatever has to remain, rather than treating it as untouchable simply because it's internal.
- Audit every field with the internal/external question before touching layout: Layout and copy improvements matter, but they're the equivalent of working faster while stopped. Ask the sequencing question first, it usually finds more slack.
- Treat consent and identity verification as the irreducible core: These need active input at the moment of submission. Make them as fast and clear as possible, but don't try to automate them away, that's a compliance risk, not an optimisation.
- Pre-fill rather than pre-fetch where data quality is uncertain: Reverse-enrichment isn't always perfectly accurate. Show it as an editable, pre-filled field rather than silently skipping the question, so a wrong inference doesn't quietly submit bad data.
- Measure external-setup coverage as its own metric: Track what proportion of a typical form submission was pre-filled versus manually entered. Rising coverage over time is a genuine efficiency gain, distinct from and additive to conventional field-count reduction.
- SMED's insight was sequencing, not speed, separating what genuinely requires a stopped process from what only seemed to
- Most CRO effort on lead forms optimises the internal process rather than questioning how much of it needs to be internal at all
- Firmographic enrichment, ad platform targeting data, and returning-visitor recognition can convert genuinely internal fields to external
- What has to stay internal, chiefly consent and identity verification, should be streamlined rather than treated as fixed
Our conversion research work audits every field on a form against what could realistically be known before the click, so field reduction targets genuine friction rather than guessing at it.
Explore Conversion Research →