The One Step Most Applicants Stall On, and How to Find Yours

Most application funnels lose people at one step, not evenly across all of them. Here is a one-hour method for finding yours from data you already have, and what to write to the people stuck there.

Read

9 min read

Published On

Sep 25, 2026

Read: 9 minutes

Last updated: September 25, 2026

In this guide

One step usually explains most of the drop-off
A four-column count that finds it in about an hour
How to tell a hard step from a badly timed one
What to write to the people stuck on it

Why One Step Usually Explains Most of the Drop-Off

Open any application funnel and the shape is the same. Many people start, fewer reach the middle, fewer still submit. The loss is rarely even. One step, sometimes two, holds a large share of everyone who leaves.

That changes what you do. If people leave evenly, the form is simply long, and the only fix is a shorter form. If they leave at one step, you have one screen to fix or explain and one group of people to write to. A message aimed at that step beats a generic nudge sent to everyone.

Build the Four-Column Count

You need one row per applicant: the furthest step they reached, the date they started and whether they submitted. If your system tracks events instead of steps, take the last event per applicant and group events into stages you can name.

Export the last 90 days of applicants who started, including those who submitted.
Keep four columns: applicant id, start date, furthest step reached, submitted yes or no.
Number the steps in the order a person meets them.
Remove test records and anyone who started under three days ago. They have not had time to stall.

For each step, count how many people reached it and how many stopped there. Then divide the stoppers by the people who reached it. That share is the number to read, because an early step with the most people looks bad on raw counts and fine on shares.

Step	Reached it	Stopped here	Share who stopped
1. Account	1,000	60	6%
2. Business details	940	90	10%
3. Link a bank	850	170	20%
4. Upload ID	680	290	43%
5. Sign	390	40	10%

These figures are an example. In this made-up funnel the ID upload loses 43 percent of the people who reach it and the bank link loses 20 percent. Every other step is ordinary.

A Worked Example: Rosa's Bakery at the ID Step

Rosa's Bakery stops at ID upload (illustrative)

Rosa runs a single-location bakery with about $38,000 a month in card deposits. She starts a funding application on the shop's back-office computer and reaches the ID step. It asks for a photo of her licence. The photo is on her phone. She closes the tab.

The platform's ops lead splits the ID step by device. People who started on a phone finish at about the funnel's normal rate. People who started on a desktop mostly stop.

The cause is a handoff, not a hard step. The platform adds a way to continue on a phone, and the people who already stopped get one short email about that step only.

Hard Step or Badly Timed Step

This distinction saves the most effort. A hard step is one where people see what is asked and decide it is too much, such as three years of tax returns. A badly timed step is easy but arrives when the person cannot do it, such as a bank login at ten at night.

Marketplaces see this constantly. A seller applying for capital is usually between orders, so the application is the first thing dropped when a new order lands. Those sellers are willing and busy, which makes timing matter more than copy.

You can tell the two apart by how people come back. Add one column, the date of each stopper's next event if there is one, and count returns inside seven days. If a good share return unprompted, the step is badly timed. If almost none do, it is hard. A hard step needs simplifying: fewer fields, clearer wording, a better example. A badly timed step needs a better moment, or a message that arrives at one.

Three Cuts That Explain the Number

Cut	What to look at	What a gap usually means
Device	Phone against desktop at the stall step	A handoff or permission problem, not a hard step
Time of day	When people start against when they stop	They start in a quiet moment and get pulled away
Source	Where they came from, such as in-app or email	Different expectations about how much work this is

What to Write to the People Stuck There

Write one message per stall step, not one for the whole funnel. It names that step, says how long it takes and promises nothing. The full pattern, with five rewrites, is in The Case Against the Reminder Email.

Hi Rosa. Your ID is the last thing left on your application. You can finish it from your phone in about two minutes. Here is the link.

Example message for applicants who stopped at ID upload

There is no amount, no mention of approval and no deadline. A follow-up that includes them can be read as an offer. Send it a couple of days after the person stops, not the same hour. Check the rules that apply to your emails first. Follow-Up Email Compliance covers consent and content.

How to Know Whether It Worked

Hold back a random group that gets nothing. Then compare the share who submit in each group over several weeks. How big the groups must be and how long to wait decide whether you can trust the result, and the holdout guide covers both. If your groups are small, say so and keep going. Do not announce a lift from forty people.

If your applicants live in a spreadsheet, a Google Sheet with a holdout column is enough to run the whole loop.

Mistakes That Waste the Hour

Reading the raw count. The step with the most stoppers is often just an early step with the most people. Divide by who reached it.
Fixing the wrong step. If people stop at ID because the bank link before it wore them out, shortening ID changes little. Look at the steps on either side.
Treating one funnel as all funnels. A marketplace seller and a contractor may stall in different places. Run your two largest groups separately.

A One-Hour Version

Export 90 days of applicants with start date and furthest step.
Make the four-column count and find the step with the highest share who stopped.
Split that step by device and by start time.
Decide: hard step or badly timed step.
Draft one message for that step in three short sentences.

If you do only two of these, do the first two. Knowing the step is most of the answer. It matters most when the funnel is a funding application, because every stalled application is revenue your platform never sees. How Embedded Capital Works for Platforms explains the program behind it.

Slate Rally writes follow-ups like these for platforms. See how Rally works.

Common Questions

How many applicants do I need for this to mean anything?

A few hundred is enough to see a large gap between steps. For small differences you need more. If a step looks bad on thirty people, treat it as a lead to check, not a conclusion.

What if my funnel has no clear steps?

Use events. Take the last thing each applicant did and group those into stages you can name, such as started, linked a bank, uploaded documents and signed.

Is it fine to message everyone who stopped?

Only people who started an application with you, and only within the rules that already apply to your emails. Opt-out and sender identification are your responsibility.

Should the message come from a person?

It can come from your brand with a real reply address. A reply that reaches a person is worth more than any clever wording.