TapCub is live — analytics, insights and live chat, free on one platform
Growth & Marketing

CRO tips: eight changes, each with the metric that proves it

Eight CRO tips in metric-and-check form, the sample size you need to see a result, and how to rank tips by expected conversions.

Sophia Bennett Published Nov 21, 2026 10 min read Growth & Marketing
Sample data: with a 2.0% baseline, 5% significance and 80% power, a +10% lift needs about 80,700 visits per group (8.1 months at 20,000 visits), +25% about 13,800 (1.4 months) and +50% about 3,800.Sample data

Good CRO tips name the metric that should move and how to check it. Show costs early (check the cost-step rate), cut form fields, offer guest checkout, speed up pages and fix errors. At a 2.0% rate, a 25% lift needs about 13,800 visits per group to detect and a 10% lift about 80,700, so small sites should make bigger changes.

What you will learn

  • Every tip needs a metric you can read before and after the change
  • At a 2.0% rate, detecting +25% needs about 13,800 visits per group; +10% needs about 80,700
  • Rank tips by share of conversions affected × expected lift, not by how easy they sound

A CRO tip is only worth following if you can verify it

Most CRO tip lists give you the idea and stop. A usable tip names the metric that should move, where to read it, and how much change counts as proof. Without that, you cannot tell whether the change worked, and a conversion rate that goes up for a week after a redesign is as likely to be noise or seasonality as a result. This article gives eight tips in that form, plus the arithmetic for how much traffic you need before you can see a result. The wider process is in the conversion rate optimization guide, and running experiments is in CRO testing.

Eight tips, each with its metric and how to check it

#TipMetric to verifyCheck it by
1Show total cost (shipping, tax, fees) before checkoutStep rate at the step where costs first appearCompare that step's rate for 4 weeks before and after
2Cut form fields to what you needForm completion rate; time to completeCount started vs submitted forms; look at the last field people touch
3Offer guest checkout or sign-up laterStep rate at the contact or account stepClicks on sign-in and forgot-password before vs after
4Make pages load fasterLargest contentful paint at the 75th percentile; step ratesLCP at 2.5 s or less, then step rate by page speed band
5Match landing page to the ad or linkSignup or purchase rate by source and campaignSplit by campaign: does the page repeat what the ad promised?
6One primary call to action per pageClick share of the primary buttonPrimary clicks ÷ all clicks on the page
7Answer objections where people hesitateExit rate and chat starts on the pricing or checkout pageQuestions asked in chat on that page
8Fix errors on forms and paymentError rate on conversion pagesErrors ÷ pageviews on those pages

Read down the last column: every check is a number you can pull before the change and after it. If a tip has no such number for your site, either start tracking it or skip the tip.

TipMetric to verify
Show total cost earlyStep rate where costs appear
Cut form fieldsForm completion rate
Guest checkoutContact-step rate
Faster pagesLCP at p75; step rates
Match page to adRate by campaign
One primary call to actionPrimary click share
Answer objectionsExit rate and chat starts
Fix errorsError rate on conversion pages

Each metric can be read before and after the change.

What the published numbers say, and what they do not

Two sourced figures are worth knowing, and neither is a promise for your site.

  • Baymard Institute lists reasons shoppers gave for abandoning at checkout (excluding "just browsing"): extra costs too high 40%, site wanted an account 18%, checkout too long or complicated 17%, website errors or crashes 17%. Its checkout review found an average of 23.48 form elements shown by default in US checkouts, against an ideal of 12–14. These are the evidence behind tips 1, 2, 3 and 8.
  • A study commissioned by Google and run with Deloitte across 37 brand sites and over 30 million sessions found that a 0.1-second improvement in mobile site speed was associated with 8.4% more conversions for retail sites. That is the evidence behind tip 4, and it is for retail, measured at one point in time.

We do not list "typical uplift ranges" per tip. Published uplifts come from specific sites with specific problems, and the biggest ones come from fixing the worst problem; the same change on a site without that problem does nothing. Your own baseline is the only range that counts.

How much change counts as proof: the sample-size arithmetic

Sample data: a site gets 20,000 visits a month and 400 conversions, a 2.0% rate. To tell a real change from chance with the usual settings (5% significance, 80% power), the sample needed per group is roughly:

n = 7.85 × [p₁(1 − p₁) + p₂(1 − p₂)] ÷ (p₁ − p₂)²

Lift you want to detectNew rateVisits needed per groupMonths at 20,000 visits (two groups)
+10% (2.0% → 2.2%)2.2%about 80,700about 8.1
+25% (2.0% → 2.5%)2.5%about 13,800about 1.4
+50% (2.0% → 3.0%)3.0%about 3,800about 0.4

For example, the +25% row: 7.85 × (0.02 × 0.98 + 0.025 × 0.975) ÷ 0.005² = 13,806 per group, so 27,612 visits in total, or 1.4 months of traffic.

What this means: a small site cannot see a 10% improvement in any reasonable time. Make bigger changes (fix the worst problem rather than polish the best page), compare the same weeks before and after, and treat the result as direction, not proof. The two-proportion test used here is described in the NIST engineering statistics handbook.

app.tapcub.com

Visits needed per group to detect a lift

sample data, baseline 2.0%
+10% (2.0% → 2.2%)80,700
+25% (2.0% → 2.5%)13,800
+50% (2.0% → 3.0%)3,800
Sample data
Sample data: with a 2.0% baseline, 5% significance and 80% power, a +10% lift needs about 80,700 visits per group (8.1 months at 20,000 visits), +25% about 13,800 (1.4 months) and +50% about 3,800.

What each tip looks like in practice

A tip becomes a change request when you write the before and after:

  1. Show costs early. Before: shipping appears on step 3 of checkout. After: "Delivery from 4.90, free over 50" under the price on the product page, and an estimated total in the cart. Verify: the step-3 rate and the share of sessions that visit the shipping-policy page.
  2. Cut form fields. Before: 14 fields including company size and phone. After: email and password only, with the rest asked after signup. Verify: started-to-submitted rate, and the field where most people stop.
  3. Guest checkout. Before: account required before payment. After: "Continue as guest" as the primary button, with an option to save details after paying. Verify: contact-step rate and sign-in clicks.
  4. Faster pages. Before: hero video loads first. After: compress the largest image and delay the video. Verify: LCP at the 75th percentile against the 2.5-second line.
  5. Match page to source. Before: every ad goes to the home page. After: each campaign goes to a page that repeats its promise in the headline. Verify: signup rate by campaign.
  6. One primary call to action. Before: three equal buttons. After: one filled button and two text links. Verify: primary click share.
  7. Answer objections. Before: pricing page has no answers to "can I cancel?" After: a three-line answer next to the plans. Verify: chat questions about cancelling on that page.
  8. Fix errors. Before: payment form shows a generic error for some cards. After: a clear message and a retry. Verify: error rate and orders after payment errors.

Change one thing at a time when traffic allows it. When it does not (see the sample-size table), batch related changes and say so in your notes, so the later reading is honest about what moved. A batch of three changes that lifts the rate by 15% cannot be split into three separate 5% effects.

Which tip first: expected extra conversions

Rank tips by a rough expected gain: baseline conversions × share of conversions that pass through the change × lift you expect. The lifts below are assumptions for illustration; replace them with your own estimates.

Sample data: 400 conversions a month.

TipShare of conversions affectedAssumed relative liftExpected extra conversions
Show costs early100%5%20.0
Cut form fields100%4%16.0
Guest checkout100%3%12.0
Match page to ad (paid traffic is 25% of conversions)25%8%8.0
Faster pages (mobile is 60% of conversions)60%3%7.2
Single primary call to action60%2%4.8

The top three affect everyone at checkout. Ordering by reach times lift is what makes the list useful; a tip with a high lift on a small share of traffic can lose to a small lift on everyone.

How to check a change worked

  1. Write the claim first: "Showing shipping on the product page will raise the cost-step rate from 70% to 74% within 4 weeks."
  2. Use the same weekdays before and after, and mark the change date on your charts.
  3. Check a second metric that should not move (for example traffic from the same sources), to rule out a change in who visited.
  4. Look at segments: desktop and mobile, new and returning. A gain on one with a loss on the other averages out to nothing.
  5. Decide the stopping rule beforehand: a fixed number of weeks or visits. Stopping when the number looks good is how false wins happen.

How to do it in TapCub

Take tip 1 (show costs early) and verify it in TapCub:

  1. Build the funnel. In All-platform analytics › Behavior › Behavior analysis, create cart → shipping step → payment → order. Each step shows the rate against the previous step and the median time; the largest drop is marked.
  2. Compare before and after. Add an annotation in Configuration › Annotations on the change date, then read the funnel for the four weeks either side.
  3. Break down by device or source. Use breakdown to see whether mobile and desktop moved the same way, and Web analytics › Sources to see rates by campaign for tip 5.
  4. Check speed and errors. Monitoring › Performance shows page performance from real users (Core Web Vitals) and Monitoring › Errors shows the error rate and affected sessions for tip 8.
  5. See what people ask. For tip 7, Live chat › Inbox shows the page each conversation started on, and Live chat › Reports › Conversion attribution lists the pages chats start on.

TapCub does not run A/B tests or split traffic, so use your own experiment tool or a before-and-after comparison and read the result here.

FAQ

What are the best CRO tips?

The ones that fix a measured problem. Start with the biggest drop in your funnel: costs shown late, long forms, forced accounts, slow pages or errors. Each has a metric you can check before and after.

How long should a CRO test run?

Until it reaches the sample you planned. At a 2.0% rate, detecting a 25% lift takes about 13,800 visits per group; a 10% lift takes about 80,700.

Do faster pages increase conversions?

A Google and Deloitte study found a 0.1-second mobile speed gain was associated with 8.4% more retail conversions. It is a retail finding; check your own step rates by page speed.

How many form fields should a checkout have?

Baymard's review found an average of 23.48 form elements by default in US checkouts and suggests 12–14 as ideal. Remove fields you do not need.

Can I do CRO without A/B testing?

Yes: use before-and-after comparisons on matched weeks, but treat the result as direction, not proof, and make large changes.

Sources

Was this article helpful?

We use your answers to fix and expand this page.

Sophia Bennett

Product Adoption Consultant

Explores customer onboarding, live conversations, and self-service, turning common usage questions into practical guidance and improvements.

Check each change in your funnel

Step-by-step funnels, annotations and page performance in one place.

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics