How to Reduce Drop-Off in the Job Application Process: Measure It by Step First

Most teams cannot reduce application drop-off because they cannot say where it happens. They track two numbers, ad clicks and submitted applications, and everything between those two numbers is a single unexplained gap. The fix is not a shorter form. The fix is to instrument the apply flow as six discrete steps, fire one event at each step, and compare step to step conversion. Once you can see which step loses the most people in absolute terms, the shorter form either matters or it does not, and you stop guessing.

This guide sets out the six steps, the event to fire at each, the four places where your numbers will legitimately disagree with each other, and the arithmetic that tells you which step to fix first.

Why can't most teams say where candidates drop off?

Because the apply flow usually crosses three systems that each count something different. The ad platform counts clicks. Your analytics counts sessions and events on your careers site. Your ATS counts records created. Nobody owns the joins between them, so the only number everyone agrees on is the one at each end.

The market makes this expensive rather than merely untidy. UK vacancies fell by 6,000 to 707,000 in the three months to July 2026 while the unemployment rate sat at 4.9%, according to the ONS labour market bulletin. A looser market means more applicants arrive per opening, so the costly loss is no longer raw volume. It is the qualified candidate who starts your form, hits a step you cannot see, and leaves.

What are the six steps of a job application funnel?

Define the funnel once and use the same six steps for every campaign and every role.

1. Ad click. Recorded by the channel's own attribution system.

2. Apply page view. The first measurable event on your property.

3. Application start. First interaction with a form field.

4. Screening complete. The candidate has answered your knockout questions.

5. Submission. The candidate presses send.

6. ATS record created. The application exists in the system a recruiter actually opens.

Steps 1 and 6 are the ones most teams already have. Steps 2 through 5 are where the money leaks, and they are invisible unless you fire an event at each one.

Which events should you fire at each step?

Google Analytics 4 gives you two of these for free. Enhanced measurement collects form_start, which fires "the first time a user interacts with a form in a session", and form_submit, which fires "when the user submits a form". Those cover steps 3 and 5 with no code. Note the documented catch: "You can only use the parameters in your reports if you create custom dimensions for them", so the events arrive but the form name and destination stay invisible until you configure the dimensions.

Steps 4 and 6 need custom events. Fire one when the last knockout question is answered, and one when your ATS confirms the record. Then build the funnel in funnel exploration, which "lets you visualize the steps your users take to complete a task" and supports up to 10 steps. Use a closed funnel, where "users must enter the funnel in the first step", so you are measuring the candidates who actually arrived from the ad rather than everyone who wandered onto the page.

Where do the numbers legitimately disagree?

Four boundaries produce gaps that are measurement artifacts, not candidate behavior. Treating them as drop-off sends you optimizing something that is not broken.

Ad clicks will exceed apply page views. Google documents the reasons plainly in its clicks and sessions troubleshooting guide: two clicks within 30 minutes may register as one session, users "close their browser or navigate away before the Google Analytics tracking code on your landing page fully loads", invalid clicks are filtered from Google Ads but their sessions are not filtered from Analytics, and campaign parameters "like GCLID or UTM parameters" are lost when a landing page redirects. Expect a gap here and investigate its size, not its existence.

Cross-domain gaps are self-inflicted. If your ad lands on a careers site and the apply flow lives on your ATS provider's domain, GA4 without cross-domain measurement configured will count "two users and two sessions instead of one user and one session", per Google's cross-domain documentation. This one misconfiguration makes steps 3 through 5 unreadable, and it is the first thing to check before you believe any funnel report.

Lead form submissions exist before your ATS knows about them. Meta lead ads can be retrieved either by a real time leadgen webhook or by bulk read polling of the Graph API, as set out in the lead retrieval documentation. If you poll on a schedule, the lead sits on the channel for minutes or hours before it reaches a recruiter. Timestamp step 5 by the channel's creation time and step 6 by your ingest time, and the gap between them becomes a response time metric rather than a mystery.

Conversion windows shift yesterday's numbers. A Google Ads conversion window can be set "anywhere from 1 to 90 days" and defaults to 30 days for click-through conversions, per Google's conversion window documentation. A report you pull today therefore contains conversions from clicks up to a month old. Compare cohorts by click date, never by report date, or an improving campaign will look flat for a month.

What does the drop-off math actually look like?

Here is the arithmetic on an illustrative campaign. These are not benchmarks. They are the shape of the calculation, and your own first month of instrumentation replaces every number in it.

The instinct is to attack the 45%, because it is the lowest rate. That instinct is right here, but only by coincidence. The correct rule is to rank steps by absolute candidates lost, not by percentage. This funnel loses 4,700 people between page view and application start, 1,600 between start and screening complete, and 400 between screening and submission. The first number is worth three times the second.

Now price it. Lifting page view to start from 45% to 60% produces 5,160 starts instead of 3,900, which is 1,260 extra candidates on identical media spend. Doubling the ad budget instead produces 7,800 starts, costs twice as much, and leaves the 45% exactly where it was. The step fix is cheaper every time, and it compounds into every campaign you run afterward.

What should you fix first at the start step?

Two things move the page view to application start rate more than anything else.

Page speed, measured at the right percentile. The Core Web Vitals thresholds are a Largest Contentful Paint within 2.5 seconds, an Interaction to Next Paint of 200 milliseconds or less, and a Cumulative Layout Shift of 0.1 or less, all assessed "at the 75th percentile of page loads, segmented across mobile and desktop devices". That percentile matters for frontline hiring. Your slowest quarter of visitors are disproportionately on older phones and weaker connections, which is exactly the population applying for warehouse, logistics and hospitality roles.

Field count. Baymard Institute's checkout research found the average checkout carries 11.3 form fields while "most sites need only 8 form fields in total". A job application is a checkout with worse motivation, since the candidate pays in effort and receives nothing immediately. If your apply flow asks for more fields than an e-commerce purchase, the field count is the hypothesis to test first.

How should screening questions sit inside the funnel?

Screening is not overhead in the funnel. It is the step that decides whether a submission is worth anything, and it belongs before the ATS rather than after it.

Wonderkind's terms and conditions define a qualified applicant as a candidate who completes the application flow, passes all knock-out questions as configured by the customer, meets the job description criteria as configured by the customer, and is delivered into the customer's ATS. All four conditions have to hold, and disputes are raised within 10 business days of the dashboard report. That is a contract definition rather than a marketing one, which makes it a usable standard to hold any vendor to, including your own team.

The practical consequence for your funnel is that step 4 should reduce step 5. A screening step that never rejects anybody is a delay, not a filter. Wonderkind's product approach puts this before the ATS with "gamified, mobile first screening flows that replace traditional application forms", assessing availability, intent and role-specific criteria, and lead ads keep the whole interaction inside the feed with one-click lead forms.

Where does Wonderkind sit in this funnel?

Wonderkind prices against the same funnel this guide describes, which is why the billing unit matters more than the rate attached to it. Its performance budget runs on three units, and each one settles at a different step. Cost per click pays for people who open your job, and counts when a click is recorded by the channel's attribution system, which is step 1. Cost per lead pays for applicants who start, and counts when a lead form is submitted on the channel, a lead form version of step 5. Cost per qualified applicant, defined above, does not settle until the record reaches your ATS at step 6.

So the unit you buy is a statement about which step's losses you absorb. On cost per click, everything from apply page view onward is yours. On cost per qualified applicant, an abandoned screening is not billable. The caveat is that buying a later unit does not repair an earlier step you still own. Candidates who leave your careers page before the flow starts reappear as a higher price per qualified applicant, and you still need the step events to see where they went.

Frequently asked questions

What is a good application completion rate?
There is no credible universal benchmark, because completion depends on role, channel, country and how much screening sits inside the flow. Your own step to step baseline from the first 30 days of instrumentation is the only benchmark worth optimizing against.

Should a lead form submission count as an application?
Only if you define it that way consistently and hold your vendors to the same definition. A lead form submit and a completed screened application are different events with different values, and merging them is how a cost per application figure becomes meaningless.

Why do my ad platform and analytics numbers never match?
Because they measure different things and always will. Google lists at least eight causes, including invalid click filtering, redirect loss of tracking parameters, and tags that fail to load before the user leaves. Track the size of the gap over time and treat a sudden change as the alarm, not the gap itself.

How many form fields should a job application have?
Fewer than your checkout. Start from the minimum needed to contact the candidate and evaluate the knockout criteria, then add a field only when you can name the decision it changes.

The short version

Drop-off is not a single number and it cannot be fixed by advice. Instrument six steps, fire an event at each, separate the four measurement boundaries from real candidate behavior, and rank your fixes by candidates lost rather than by percentage. The team that can say "we lose 4,700 people between page view and application start" is a month ahead of the team that can only say applications are down.