Home/Blog/Conversion Funnels/How to Run a Funnel Improvement Loop Eve…
Conversion Funnels

How to Run a Funnel Improvement Loop Every Month

How to run a funnel improvement loop: change one thing, be honest about whether you have the volume to read it, recheck later, then move to the next constraint.

Cover of the FIKR Space guide: How to Run a Funnel Improvement Loop Every Month

To run a funnel improvement loop, work through a monthly cycle: prioritise one loss, diagnose its cause, change one thing, read the result honestly, and then look for where the constraint moved next. It is the loop that turns a dashboard into money.

A funnel is not a project with an end date. Funnels behave like a chain: fix the worst link and the constraint moves somewhere else. That means the work is a habit and not a task, and the habit is what this post describes. The parts of the loop that need a deeper treatment have their own posts, linked where they come up.

How to run a funnel improvement loop: the four stages

The guide's third checklist is called the improvement loop, and it is meant to be used monthly. It has three groups of boxes, which can be read as four stages in order.

StageThe questionWhat you leave with
PrioritiseWhich loss is the largest and cheap enough to address?One step chosen, with segments checked
DiagnoseIs it friction, doubt or mismatch, and what is the evidence?A written hypothesis
ActWhat single thing will I change?One change, made
EvaluateDid it work, can I tell, and where did the constraint go?A judgement, and the next target

The first two stages are covered at length elsewhere. Ranking by absolute loss is in how to find where a funnel loses people, and the three causes are in how to fix funnel drop off. This post concentrates on what happens after you have chosen a step and decided what you think is wrong with it.

Write the hypothesis down first

Before you change anything, write down what you believe is happening and what you expect your change to do. The checklist includes this line: I have written the hypothesis down before making the change.

It looks like paperwork. It does a specific job. A hypothesis written beforehand is something your result can contradict. Without it, whatever happens can be explained afterwards as a success, and you lose the ability to learn from being wrong. A single sentence is enough: which cause you believe it is, what evidence you have, and what you will change.

Free guide, 30 pages: The Funnel Playbook. Send me the guide.

Change one thing

Change one thing at the step, not several. Multiple simultaneous changes to the same step produce an outcome you cannot interpret, and you will not know which change to keep. Worse, you may keep the wrong one and build on it.

This can feel slow, especially when you can think of four improvements at once. But a loop that teaches you something each time is faster over a quarter than one that produces a muddle. If you have several ideas, queue them. The step will still be there next month, and by then you will know what the first change did.

ApproachWhat you can say afterwards
One change at the stepThis change moved the step, or did not
Several changes at onceSomething moved it, and we cannot say what to keep

Be honest about the volume

The next box in the checklist is the one people skip: I know whether I have the volume to read the result, and I have said so honestly.

Most early funnels do not have enough traffic for a small improvement to be told apart from ordinary noise within a useful time frame. That is not a failure. It is a fact about your stage. It leaves you two legitimate choices.

  • Make a change large enough to produce an obvious effect. A big difference can be seen even in a small sample.
  • Accept that you are making a reasoned judgement and not running an experiment. Say so plainly.

Both are fine. The one thing that is not fine is pretending a small sample proved something. That is how teams end up confidently maintaining changes that did nothing. A useful phrase to adopt in your own notes is this is a reasoned judgement, which keeps you honest about the strength of what you know.

Knowing your ordinary variation helps here. The month of baseline data you took before changing anything tells you how much the numbers wobble from week to week, so you can recognise a change that is bigger than a normal wobble. That baseline is built as described in how to track funnel events.

Recheck the step a few weeks later

Fixing a step is not the end. Re-examine it after a few weeks. There are two reasons.

  1. Some gains decay. A change that looked good at first may fade, and you want to know whether it held.
  2. The loss may have moved, not vanished. Occasionally a fix at one step simply moves the loss to the next one, because the people you saved were never going to convert anyway.

That second outcome is worth knowing about. It usually indicates a mismatch problem that the fix did not address. If you see it, stop working at the step and look upstream, in the message or the targeting.

Then look for the next constraint

When a fix works, the constraint moves. This is success and not a setback. The step that was your largest loss is no longer the largest, and a different step now sits at the top of the ranking.

So the loop starts again. Rebuild the ranking from fresh numbers, check the segments in case the new problem is one audience, and pick the largest cheap loss. Over time, each pass improves one more link in the chain. Because every improvement multiplies against all future acquisition, the effect accumulates in a way that spending on traffic does not.

FIKR Content, multi-channel publishing: reads, listens, opens, clicks in one dashboard. See FIKR Content.

Know when the funnel is fine

The loop has an exit, and a good loop knows when to use it. Sometimes the funnel is not the problem.

What you seeWhat it points to
Every step converts reasonably, totals still too smallA volume problem, not a conversion problem
Every step mediocre, none catastrophicThe offer, rather than the sequence
One step clearly worse than the restA specific cause at that step: back into the loop

If every step converts reasonably and the totals are still small, further optimisation is misdirected effort. If every step is mediocre, suspect the offer. The funnel is a good enough instrument to tell you when it is not the right instrument, and knowing that boundary saves you from the most common failure of measurement work, which is polishing the thing you can see because you can see it.

The first post in this series, what is a conversion funnel, sets out the same honest limit from the start.

Ways the loop quietly breaks

Most failures of the loop are failures of discipline and not of tooling. Three are worth watching for.

  • Reaching for traffic. When a month passes without a visible win, the pull is to buy more visitors. That pours volume into a process that is still losing most of what enters it, at a cost that only rises.
  • Declaring victory too early. A small sample can make almost any change look good. Say it is a reasoned judgement and keep the recheck in the diary.
  • Polishing the wrong funnel. If the offer is wrong, percentage gains on a number that cannot work are motion, not progress.

The common thread is that each one feels productive while it happens. The loop is built so that a written hypothesis, a single change and a later recheck make these easier to catch.

Why a loop beats a project

A project has an end, and a funnel does not. The constraint moves each time you fix a link, so there is always a next one, and the gains from the last pass carry forward into the next. Waiting for a full sales cycle to learn whether a change worked is slow. Watching whether it moved the step it was aimed at can be done within days, which is why teams that watch the steps can learn once a week while teams that watch only the final outcome learn a handful of times a quarter.

That faster feedback is one of the few real advantages a small company has, and the loop is how you use it.

Free guide, 30 pages: The Funnel Playbook. Send me the guide.

A monthly agenda

The loop needs a regular moment. A short, fixed agenda is enough.

  1. Rebuild the ranking of losses from the latest cohort counts.
  2. Check the segments, in case the problem is one audience.
  3. Review the change from last time: did it hold, and where did the constraint go?
  4. Pick the largest cheap loss and name the cause you believe it is.
  5. Write the hypothesis, then make one change.

Keep a short written record of each pass. It gives you a history of what you tried and what you learned, and it stops the same idea being tried twice without anybody noticing.

FIKR Content, multi-channel publishing: reads, listens, opens, clicks in one dashboard. See FIKR Content.

Frequently asked questions

How often should I run the loop?

Monthly, as the guide's checklist suggests. Rechecking an improved step a few weeks after the change fits into the same rhythm.

What if I do not have enough traffic to read a small change?

Make larger changes whose effect would be obvious, or accept that you are making a reasoned judgement and say so. Do not claim an experiment proved something it could not.

Why only one change at a time?

Several changes at the same step give an uninterpretable result, and you will not know which to keep.

What if the fix works but the total does not improve?

The loss may have moved to the next step because the people you saved were never going to convert. That usually points to a mismatch to fix upstream.

The full method, with the checklists, is in The Funnel Playbook, a free 30-page playbook.

LG

Luis Goncalves

// Founder & CEO at FIKR Space

Three-time founder. Built and exited Evolution4All before this. Now building FIKR Space — the operating infrastructure underneath every innovation ecosystem (startups, accelerators, governments, investors). Lisbon-based, works global.

Claim your founding member spot.

Every founding member is hand-reviewed. Tell us who you are and how you want to co-create — we get back inside 48 hours.

Role
Job Title
Country
Cookies

We use cookies.

Essential ones keep the site working. The rest only run if you say so. No dark patterns, no pre-ticked boxes.Privacy policy

Co-create with us. It's about building it together.Co-Create With Us