Conversion Rate Optimisation Across the Funnel
Compare two groups at every funnel step, put a value on each possible fix, and choose what to try first with simple scoring, using a worked phone versus desktop example.
In this lesson you will
- Compare two groups step by step
- Value a fix in extra orders
- Prioritise ideas with ICE
- Remove worry and effort at the moment of buying
About 25 minutes
Phones are most of the visits and few of the sales
Noor splits her shop’s funnel by device and notices something. Phones bring most of her visits, but desktop visitors place twice as many orders. She could blame the phone, but that is not an answer. Conversion rate optimisation, often shortened to CRO, means finding where people are lost and fixing the leak that matters most. This lesson shows how to compare two groups at every step, put a value on each fix, and choose what to try first.
Find the leaks
Lesson 3 introduced the funnel. Here we apply it to two groups, so the better group shows what is possible. The program below uses made-up numbers for five steps: visit, product page, basket, checkout and paid. It works out the share who move on at each step for each device. Then it asks, for each step, how many extra orders mobile would get if it matched desktop at that one step only, and what those orders are worth at an 18 pound average order.
# MADE-UP funnel by device: visits, product views, baskets, checkouts, paid
stages = ["visit", "product", "basket", "checkout", "paid"]
dev = {"mobile": [6000, 3600, 540, 240, 90], "desktop": [4000, 2800, 600, 360, 180]}
AOV = 18.0
rate = {d: [b / a for a, b in zip(v, v[1:])] for d, v in dev.items()}
for d, v in dev.items():
print(f"{d}: {v[-1]} orders from {v[0]} visits")
for i in range(4):
print(f"{stages[i]} to {stages[i+1]}: mobile {100 * rate['mobile'][i]:.1f}%, desktop {100 * rate['desktop'][i]:.1f}%")
# extra orders if mobile matched desktop at ONE stage
gain = {}
for i in range(4):
r = list(rate["mobile"]); r[i] = rate["desktop"][i]
n = dev["mobile"][0]
for x in r:
n *= x
gain[stages[i] + " to " + stages[i + 1]] = n - dev["mobile"][-1]
for k, v in gain.items():
print(f"fix {k}: +{v:.0f} orders = {v * AOV:.0f} pounds")
print("biggest prize:", max(gain, key=gain.get))
# MADE-UP ideas: (idea, impact, confidence, ease) each 1-10
ideas = {"shorter checkout form": (9, 7, 6), "show delivery cost early": (8, 8, 8), "new logo": (2, 3, 4), "guest checkout": (8, 7, 5)}
ice = {k: v[0] * v[1] * v[2] for k, v in ideas.items()}
for k in sorted(ice, key=ice.get, reverse=True):
print(f"{ice[k]}: {k}")
mobile: 90 orders from 6000 visits
desktop: 180 orders from 4000 visits
visit to product: mobile 60.0%, desktop 70.0%
product to basket: mobile 15.0%, desktop 21.4%
basket to checkout: mobile 44.4%, desktop 60.0%
checkout to paid: mobile 37.5%, desktop 50.0%
fix visit to product: +15 orders = 270 pounds
fix product to basket: +39 orders = 694 pounds
fix basket to checkout: +32 orders = 567 pounds
fix checkout to paid: +30 orders = 540 pounds
biggest prize: product to basket
512: show delivery cost early
378: shorter checkout form
280: guest checkout
24: new logo
Mobile makes 90 orders from 6000 visits, desktop 180 from 4000. Mobile is behind at every step, but not equally. At the product to basket step, mobile reaches 15.0 per cent against 21.4 on desktop, and fixing that alone would add about 39 orders, worth about 694 pounds. That is the biggest prize: product to basket. Fixing the last step, checkout to paid, where mobile has 37.5 per cent against 50.0, would add about 30 orders. Notice that the biggest percentage gap is not always the biggest prize. What counts is how many people reach that step and how much a gain is multiplied by what happens after it.
Choose what to try first
Ideas are easy to find and time is short. A simple scoring method, often called ICE, rates each idea from 1 to 10 for impact, confidence and ease, and multiplies them. The last lines of the output score four made-up ideas. Top of the list at 512 is ‘show delivery cost early’. Notice that the new logo scores very low, which matches a common experience: the changes that feel like marketing often matter less than the changes that remove worry or effort at the moment of buying. The scores are opinions, so write down why you gave each one, and revisit them when you learn something.
- Remove worry. Show delivery costs, returns and reviews early.
- Remove effort. Fewer fields, guest checkout, saved details.
- Remove doubt. Clear photos, honest descriptions, a real contact.
- Fix speed. A slow page loses people, especially on phones.
- Test on a phone. Try buying something yourself, on a slow connection.
- Split your funnel by device, source or new versus returning visitors.
- For each step, estimate the extra orders from matching the better group.
- List ideas, score them with ICE, and write the reasons.
- Test the top one with a fair A/B test (lesson 18), or at least compare before and after carefully.
- Record what you learned, win or lose.
Redesigning everything at once. If you change the whole site you will never know which change helped or hurt. Change one thing at a time, in the order of the likely prize.
The matching-
Practice for this lesson
Check what you learned
Five quick questions. Pick an answer and you will see straight away whether it is right and why.
What does CRO try to do?
- Increase visits only
- Lower prices
- Redesign the logo
- Find and fix where people are lost
It is about improving the share of visitors who complete a goal.
How does the program value a fix?
- By colour
- Extra orders if one step matched the better group
- By cost
- By likes
It estimates orders from matching desktop at one step.
Which step was the biggest prize?
- Visit to product
- Basket to checkout
- Product to basket
- Checkout to paid
It gave the most extra orders in the sample.
What does ICE stand for?
- Impact, confidence and ease
- Income, cost and effort
- Ideas, content and email
- Image, copy and edit
Each idea is rated on those three and multiplied.
Why avoid redesigning everything at once?
- It is illegal
- It costs nothing
- You will not know what helped or hurt
- Visitors prefer it
One change at a time shows the cause.
Summary
- Split the funnel by device or source to find leaks.
- The biggest gap is not always the biggest prize.
- Score ideas for impact, confidence and ease, and write reasons.
- Change one thing at a time and keep the checkout honest.
Questions
Is the extra-orders figure a promise?
No. It is an upper estimate that shows where the gap is largest. Real fixes seldom close the whole gap.
Are the funnel numbers real?
No. They are made-up sample numbers to show the method.
Where do ICE scores come from?
They are your honest opinions, ideally backed by data and customer comments. Write the reasons so you can revisit them.
