Two experiments at Rocket Mortgage moved billions in loan volume by changing the order of things. A deep dive on friction, cognitive load, and experiment design.
The most expensive step in most funnels is a form. It sits at the end of everything you paid for: the ad click, the landing page, the pitch. Then it quietly turns away a third of the people who were ready to say yes.
Forms fail silently. Nobody complains about a form they abandoned. The money just never shows up.
This is the deep dive behind my Rocket Mortgage case study: what the two experiments actually tested, why order matters as much as it does, and how to run the same play on your own funnel. It maps to lens three of the 30-day growth audit.
The setup: a form worth billions
Rocket Mortgage is one of the largest mortgage lenders in the United States. Its application starts with an online questionnaire that asks about your home, your finances, and your goals, then routes you toward a loan.
At that traffic volume, the questionnaire is one of the most valuable pages in the company. A single percentage point of completion is a very large number of loans.
Which is exactly why it made a perfect testing ground. Small changes, measured properly, at a scale where the answer is unambiguous.
Experiment one: easy questions first
The hypothesis was simple. People were quitting early because the hard questions came too soon. Move the low-effort questions to the front, let people build momentum, and more of them will finish.
The variant changed no copy, no design, no fields. Only the order.
Completion rose 4%. More interesting: the close rate, the share of applicants who went on to become funded loans, rose 12%. Measured out, the reorder was worth $1.5 billion in incremental loan volume, on the same traffic and the same ad spend.
An obviously simple change, a massive result. That combination is what a well-run experimentation program is for.
Experiment two: the menu, and the flashy variant that lost
The second test targeted a menu where existing homeowners chose between loan options, including cash-out refinancing. The reorder variant simply resequenced the options so the most relevant choices for that audience came first.
It ran against a second variant that added a Trending badge to steer attention, the kind of clever flourish that demos well in a design review.
The plain reorder won. Completion rose 5% and close rate rose 18.8%, worth $1.1 billion in loan volume. The badge variant lost to it.
That result generalizes. Structural changes that reduce the work a visitor does tend to beat decorative changes that try to redirect their attention.
Why order matters this much
Cognitive load is the mental effort a task demands. Every question in a form has a price: remembering a number, finding a document, deciding whether you trust the site with your income.
Visitors run a quiet calculation the whole way through: is the effort ahead worth the progress so far? When an expensive question arrives before any progress exists, the math says quit.
Reordering changes the math without changing the questions. Cheap answers early create progress, and people protect progress they have already made. By the time the expensive questions arrive, quitting means losing something.
Experiment design notes
The mechanics matter as much as the idea. Three notes from these tests worth stealing:
- One variable per test. Both winning variants changed order and nothing else. When a variant changes five things and wins, you have learned almost nothing you can reuse.
- Measure past the form. Completion alone can lie. A change can push more unqualified people through and improve nothing downstream. Close rate is where these experiments proved their value.
- Let the boring variant compete. The reorder beat the badge. If only the clever variant had been built, the team would have shipped the weaker idea and celebrated.
Reorder your own form
You can run a first pass on this in an afternoon:
- List every field in your highest-traffic form, in current order.
- Price each field from the visitor's side: effort to answer, discomfort to share. A name is cheap. Income and phone number are expensive.
- Find where people quit. Step-level analytics or session recordings will show the exact question where drop-off spikes.
- Move the expensive fields as late as the flow allows, and open with two or three questions anyone can answer without thinking.
- Test the reorder against your current form and judge it on a downstream metric, whatever your version of close rate is.
If your form opens by asking for a phone number, an email, or an income figure, you are almost certainly paying for traffic that quits at question one.
Want help finding what your funnel is leaving on the table? Let's talk, or email me at karran@karrangupta.com.