What a monthly testing cadence taught me about conversion rate, failed experiments, and staying visible to AI search.

Silhouette of a person running a marathon — *Source:* *Jeremy Lapak, Unsplash.*

Most teams treat a website launch as the finish line. The site goes live, everyone celebrates, and attention moves on to the next shiny object. But for a high-traffic site in the tech industry, launch day is really the starting point of a much longer discipline — the ongoing work of making sure the site keeps performing as the product, the market, and search in general keeps changing.

I lead that work for a B2B SaaS marketing site in the real estate technology space. Here's the process behind it, and some of what a real half-year of testing looked like — good and bad.

1. Get a baseline and build a cadence.

It's incredibly easy to fall into the trap of being reactive. Someone outside your organization notices a problem > you panic > you patch the site, rinse and repeat. The problem with being reactive is that your to-do list becomes infinite without you ever fully understanding why things are broken on your site. You're leaving so much on the table by not taking the time to think through the user experience on your site and test against a hypothesis. You officially need a baseline.

So how do you even begin to tackle a project of this magnitude? Easy — go back to basics. Run your site through a free tool like SEO Wallet, Google Search Console, or Screaming Frog to get cold, hard data. Review your site and take stock of any branding inconsistencies, overall conversion and engagement rate, and any technical elements that need improvement. Document it somewhere that's easy for you to reference — this is your baseline, the thing you're trying to improve upon.

For us, that baseline looked strong: hundreds of thousands of sessions, thousands of form fills, and a site-wide conversion rate that climbed roughly a third over the prior half's numbers. None of that came from one redesign. It came from a handful of optimization tests run through our testing platform and a steady stream of SEO content published and tracked against that same baseline, month over month.

Now for the frequency…

Here's what to do: set up a regular, recurring review cycle rather than addressing a running list of fixes. In the case of BoldTrail, that meant monthly optimization meetings where stakeholders across the team align on priorities, review analytics together to identify where the site is underperforming, and leave with assigned next steps. Depending on how much time and resources you can dedicate to this initiative, increase or decrease the frequency of meetings accordingly. Keep track of the project in some kind of project management software, and take detailed notes on what's discussed in each meeting.

Keep in mind, though, that cadence is what actually matters. Without it, any optimization work turns reactive — a scramble whenever a certain metric drops, rather than a steady practice of continuous improvement. For us, that monthly rhythm keeps the site (and the team) accountable to its own numbers.

2. Test before you commit.

You’ve invited your stakeholders to a recurring meeting. Now what?

Here’s a tip: the data you collected from your baseline step should inform which page needs the most attention. During your sync, review what you’ve found and come up with some hypotheses and experiment ideas. Figure out what your metrics are, what you’d like to improve. Think about your childhood science class, where you first learned about the scientific method. If you need a refresher, the process goes like this:

Make an observation > ask a question > form a hypothesis > conduct an experiment > analyze data > form a conclusion.

Here's the golden rule: nothing significant ships without first testing the hypothesis. "Significant" means anything that could move the needle on conversion rate — a button style, CTA copy, an image treatment, a new headline. And I'll be honest — not every test wins. Here's what a real quarter of testing looked like for us:

Page	Test	Result
Demo request form	Form headline	Small lift
Homepage	Hero CTA colors	Small dip
Product page	CTA copy variation	Sharp drop
Homepage	Social proof on demo form	Modest lift
Comparison landing page	New pop-up modal	Strong lift
Pricing page	New waterfall form	Notable dip

Losses like the CTA copy test are exactly why testing matters — they're cheap to catch early and expensive to miss. Across the quarter, the net effect of all this testing was a real, measurable increase in overall site-wide conversion rate, quarter over quarter. It wasn't a dramatic before-and-after screenshot. It was a slow, compounding gain built out of a handful of small experiments, several of which didn't work.

Watercolor wireframe sketches of website layouts — *Photo by* *Hal Gatewood* on *Unsplash*

Getting a test live cleanly takes real coordination — design and development have to be in lockstep, and the demand gen team needs a say in where testing effort actually pays off. That's usually the landing pages carrying the most traffic and pipeline pressure, where even a modest lift in conversion rate has an outsized downstream effect. A recent pricing page redesign is a good example: since launch, it's converted well above where the old page performed, with strong engagement and a real, if early, trickle of closed business — numbers we wouldn't have without a page built specifically to be tested against.

At its core, this is about how decisions get made. Instead of leaning on instinct or whoever argues loudest in a meeting, a change has to earn its place on the live site by proving it works. That's slower than pushing a redesign straight to production, but it's what keeps the site from collecting regressions dressed up as improvements — and it's the only way I've found to reliably tie design decisions to revenue.

3. Treat search as a moving target — including AI search.

The third piece of the process is technical SEO, and it's evolving faster than most teams' playbooks have caught up with. Alongside traditional search engine optimization, this case study describes a deliberate effort to structure the site for AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) — making sure content is structured so it can be surfaced and cited correctly as more prospects discover products through AI-assisted search rather than a traditional results page.

Magnifying glass beside the corner of a laptop on a marble surface — *Photo by* *MJ Duford* on *Unsplash*

In practice, that's less about chasing a new set of keywords and more about restructuring content so a language model can actually read it: clear headers that state a claim rather than tease one, direct answers near the top instead of buried under three paragraphs of preamble, and data points and comparisons that stand on their own if they get quoted out of context. It's a shift from writing for a ranking algorithm to writing for something closer to a very literal-minded reader. Traditional SEO isn't going away, but ignoring this layer means slowly becoming invisible to a growing share of how people actually search.

Why this process matters more than any single redesign.

It's tempting to think of a site's performance as something a redesign fixes once and for all. The case study pushes back on that idea directly, and the line is worth sitting with:

“A launch is a moment. Optimization is a discipline."

That distinction is the real takeaway for any SaaS team maintaining a marketing site. A launch is visible — it has a date, a moment of celebration. Optimization is mostly invisible. It's monthly meetings, a losing test on a product page, a steady drip of blog posts nobody throws a launch party for — the kind of work that rarely produces a dramatic before-and-after screenshot, but steadily determines whether the site keeps earning the results it launched with.

A repeatable framework for other teams.

Distilled into a general process, this approach looks like:

Set a fixed review cadence. Monthly (or another consistent interval) beats reactive fixes, because it forces regular attention instead of crisis response.
Ground priorities in analytics, reviewed together. Cross-functional alignment on what the data shows prevents optimization from becoming any one team's opinion.
Test before permanent changes ship, especially on high-traffic, high-intent pages where conversion lift compounds.
Pair technical SEO work with content strategy. Structure and content need to move together, not in separate workstreams.
Design for AI-assisted discovery alongside traditional search. AEO and GEO aren't replacements for SEO — they're an additional layer of how a site needs to be found now.

A site is never really "done." Treating optimization as an ongoing operating rhythm, rather than a project with an end date, is what keeps a launch from becoming a slowly depreciating asset.

Web Optimization Is a Marathon, Not a Sprint.