Compare Two Versions of a Page or Post Each Week

You rewrote a blog title because it felt stronger. Did it help? Most small businesses never find out. The page just sits there, and the next change is another guess. Comparing two versions properly is the way out, but it sounds like work for a team with a data analyst. It is not, if the rules are set in advance. You need a clear measure, a minimum amount of data, and the patience to let a result settle before you act on it. This page explains how a comparison like that works for blog pages and social updates, what numbers matter, where small sites run into limits, and how ConveyLoop runs the routine for you and shows what it concluded.

What a steady comparison routine gives you

Why small sites need ground rules before comparing anything

A site with a few hundred visits a month cannot read much into one good day. Random swings are large at low volume. If a title change is followed by a spike, the spike may have nothing to do with the title.

So the rules come first. Pick one measure per content type. For a page in search, click-through rate is the honest one. For a social update, engagement rate. Set a minimum run time and a minimum amount of data. Decide how much better a version must be before it counts.

What counts as enough data

ConveyLoop uses fixed gates. A page comparison runs at least 14 days and needs 300 search impressions, or 500 page views if Search Console is not connected. A social comparison runs at least 72 hours and needs 500 impressions.

A change is kept only if it passes a significance test at 95% confidence and beats a minimum gain, such as 8% better click-through for a page or 12% better engagement for a post. If the result is still unclear after three times the usual length, it is marked inconclusive.

Page comparisons look at the four weeks before the change against the weeks after. Only one comparison runs at a time on each page and on each social network, so two changes never muddy each other.

What to change, and when to stop

For pages, the change is the title, the description and the FAQs, tested together, or a new section added. For social, it is a new hook or tone against posts written the usual way. Changing one kind of thing at a time keeps the result readable.

Stopping early matters too. A comparison ends if the main number falls 15% or more, or conversions fall 8% or more. In Auto mode the earlier page version is restored. In Semi and Review modes you get a recommendation note instead.

Video is not yet covered by the Experiments feature, so do not expect video results there.

How ConveyLoop Helps

Every Monday ConveyLoop scores each live page and post, picks the ones falling behind, and starts a comparison on them. You do not set up the test. The Experiments page lists what was changed, what won, the confidence status and the lift, and the weekly email sums it up. Winners are kept and patterns from them shape what gets written next. Losers are undone or flagged, depending on your approval mode. Measurement uses ConveyLoop's own cookie-free script, so Search Console and GA4 are optional.

Frequently Asked Questions

How long should a comparison run before I trust it?

Long enough to pass a minimum of time and data. ConveyLoop uses 14 days and 300 search impressions for pages, and 72 hours and 500 impressions for social updates.

Can a small site do this without Search Console?

Yes. ConveyLoop measures pages with its own script. Without Search Console, a page needs 500 page views before a comparison can be judged. One catch: Markdown and MDX files lack the script, so they need the Metrics snippet or Search Console.

What happens if a change makes things worse?

The comparison stops if the main number drops 15% or more, or conversions drop 8% or more. In Auto mode the earlier version comes back.

Do I have to approve every change?

No. You choose Auto, Semi or Review mode per project. Review sends everything to you first.

What if there is no clear winner?

The test ends as inconclusive after three times its usual length. Nothing is promoted on thin evidence.

Does it need WordPress?

No. It publishes to WordPress or a Git repository on GitHub.com, GitLab.com or Bitbucket Cloud.

Does this cost extra?

No. Every feature is on every plan, starting at $19 a month plus sales tax for one site. Plans differ by volume.