How to read a funnel drop-off before you fix anything
A big fall between two pages is a question, not an answer. Here is the order to work through it, so the next marketing your team does is aimed at the real cause.
On this page
The short version
The short version- Rank the falls by how many people you lose, not by the biggest percentage. A dramatic rate on a quiet page is rarely the best use of a week
- Falls should get smaller as people get closer to enquiring. One that gets bigger further down is the first place to look
- If one marketing source falls much harder than the rest, the page is probably fine and the promise made upstream is wrong
- Check what you changed before you look for a cause, then watch a handful of real visits
- Under about a hundred people at a step, the honest answer is "not enough yet". Widen the date range and wait
A funnel drop-off is the share of people who leave between one step and the next, such as between your prices page and your enquiry form. To read one you need three things the chart will not give you on its own: how many people it is, how keen they were when they arrived, and who they were.
The example here is a wedding venue. Couples arrive on the venue's gallery, look at the packages and prices page, open the enquiry form and send it. Their marketing is a listing on a wedding directory, Instagram, a stand at two wedding fairs a year and an advert in a regional bridal magazine. The owner sees that most visitors leave on the prices page and wants to know whether to redesign it. All the figures below are illustrative.
Here is the order to work through before anyone touches the page.
Find the fall, then count the people
Finding the fall takes seconds: it is the step where the bar shrinks most. Deciding whether it deserves the team's week takes longer, because a percentage and a headcount answer different questions.
Say 3,000 couples reached the gallery in a season and 900 went on to the prices page: a 70% fall, losing 2,100 people. Then 900 reached the prices page and 180 opened the enquiry form: an 80% fall, losing 720. Of those 180, 120 sent it: a 33% fall, losing 60.
Percentages tell you how well a step works. Counts tell you what fixing it is worth.
The prices page has the worst percentage, but the gallery loses three times as many people. Shortlist the two biggest losses by headcount, then ask why each one leaks.
Is the fall big for where it sits?
There is no universal "bad" drop-off. A fall only means something next to the step before it, and next to how keen people were when they reached it.
Intent climbs as people move down.
A couple browsing the gallery is dreaming. A couple on the enquiry form has a date in mind. So the falls should get smaller as you go down, and the first place to look is any step where the fall gets bigger instead.
| Where the fall is | A large fall here usually means | When to worry |
|---|---|---|
| Gallery to prices | Plenty of people browsing who are not planning yet | Only if it changed, or one source is much worse |
| Prices to enquiry form | The price or the package did not match what they expected | Often worth a look. This is the venue's real question |
| Enquiry form to sent | Something about the form put them off | When it is bigger than the fall above it |
When the bottom leaks more than the top.
A fall that grows near the end usually means people were surprised after they had already decided. For an enquiry form that is usually one of a few things: a required field nobody expected (a budget, a guest count, a phone number), a form that breaks on phones, or a date picker that only offers dates already taken. All are quick to check by filling the form in yourself, on a phone.
Check who started at the top
Not everyone enters a funnel at step one. Plenty of the venue's couples follow the directory's "see prices" link and land straight on the prices page, never seeing the gallery. If they are counted as if they walked the whole path, the prices page looks busier than the gallery could have made it, and its fall looks worse than it is.
Keep two questions apart:
- "How well does our path work?" Count only people who started at the first step
- "How many enquiries did we get?" Count everyone. An enquiry is an enquiry however the couple arrived
If your analytics can split the first step by entry page, do it. A prices page where half the visitors came straight from the directory is two audiences sharing one bar, and they may need different fixes.
Is it the page, or the people the marketing sent?
A drop-off is an average, and two groups behaving very differently can produce an ordinary looking number. Split the fall by where people came from, one source at a time. This is why UTM tags matter, and why even simple analytics tools read them: the small labels (such as utm_source=bridal-magazine) on the links you control are what let your analytics separate the magazine from Instagram.
If one source falls much harder than the rest, the page is probably fine.
Say couples from the directory leave the prices page at 60% and couples from Instagram leave at 95%. The prices page works for people who arrived expecting the venue's prices. Instagram is sending people who were not, perhaps because the posts show the venue at its most lavish. The fix is in the posts, not the page.
If every source falls about the same, it is the page.
Less comfortable, and more useful.
Sometimes nothing changed except the mix.
If Instagram's share of visitors doubles after a popular post, the overall drop-off rises without any page getting worse. The tell is that each source is flat while the total moves. Then the question is about the marketing, not the website.
Check what changed, then watch a few visits
A drop-off that started three weeks ago is a different problem from one that has always been there. Before looking for a cause, ask what changed: a new prices page, a new package, a form field added, a new campaign that started sending different people. A change on Tuesday and a fall starting Tuesday is not proof, but it is a very short list of suspects.
Then watch what a few real visitors did. Numbers are good at showing where people leave and poor at showing why. Many analytics tools show individual visitors' paths, and free recording tools such as Microsoft Clarity show clicks and scrolling on a page.
Five is usually enough.
Jakob Nielsen's case for testing with five users, from 2000, estimated that five people reveal about 85% of the usability problems in a design. The same holds here: people fail in patterns, and the sixth mostly repeats the third.
Decide what to try next
The point of reading a drop-off is choosing the next piece of work. Match what you found to the move:
- Fix the source
- One marketing source falls much harder than the rest
- Fix the page
- Every source falls by about the same amount
- Undo or rethink the change
- The fall started the week something on the page or the offer changed
- Wait and gather more
- Fewer than about a hundred people reached the step
When the honest answer is "not enough people yet".
If 12 couples reached the enquiry form and four sent it, the rate is 33%. One more and it would be 42%. Watching that daily means reacting to noise. Evan Miller's warning about checking test results too often makes the point with numbers: in his worst case, checking after every visitor pushed the false alarm rate from an assumed 5% to 26.1%. Widen the date range, and look monthly.
Write the choice down with a date to look again, so the answer survives a busy month. If you want the arithmetic behind each rate, how to calculate funnel conversion rate works through it.
Frequently asked questions
How big does a drop-off have to be before it is worth fixing?
There is no threshold that works for every business. The useful test is whether the fall is bigger than the one before it, because people should get keener as they move down. After that, rank the falls by how many people you lose rather than by percentage.
What is a normal drop-off rate for an enquiry form?
There is no reliable figure that travels between businesses, because forms, prices and visitors differ. Compare your form with itself last month, and compare your marketing sources with each other. If one source leaves the form far more often, that is the lead to follow.
Why do people leave my prices page?
Usually because the price or the package did not match what they expected when they arrived. Split the drop-off by where visitors came from. If one source leaves far more often, the expectation set by that marketing is the problem, not the page.
How many visitors do I need before a drop-off means anything?
Roughly a hundred people at the step before the rate is worth an opinion, and a few hundred before comparing sources. Below that, widen the date range and watch individual visits, which stay useful at any volume.
Should I redesign the page with the biggest drop-off?
Not before checking three things: whether one marketing source causes most of it, whether something changed when it started, and whether enough people reached it to trust the number. Often the fix is in the marketing that sent people, not the page.
Read it in five moves
Rank the falls by people lost. Read each one against the step before it. Take out the people who skipped the start. Split by source until it stops being an average. Then check what changed and watch a few real visits.
Most of the time that points at the marketing as often as the website, which is why knowing which marketing worked is how you decide what to try next. That is the loop Mole Marketer is built on, with analytics as one of its three inputs: every initiative gets its own UTM tags, the included Funnel Analytics shows what each one brought in, and what happened feeds the next week's initiative pitch. Your team still decides what to run.
Written by Oli
Founder of Mole Marketer. Writes about running marketing as a steady weekly process with a small team, and checks every third-party price against its own page before publishing.
