Barrava's Blog

Conversion rate optimisation: an evidence-first model | Barra

Rich Tank — Mon, 21 Sep 2026 09:00:00 GMT

A practical operating model for conversion rate optimisation that turns customer evidence into prioritised tests, useful decisions and better website journeys.

Quick answer

Conversion rate optimisation works best as a learning system: define the customer job, diagnose the journey, choose one evidence-backed hypothesis, test it with an appropriate comparison and judge both the immediate action and its downstream quality. The objective is not more clicks at any cost; it is more useful progress with fewer avoidable barriers.

CRO is often reduced to button changes and landing-page ideas. That makes activity easy to produce but learning difficult to accumulate. A stronger programme begins with the decisions visitors are trying to make and the evidence that explains where those decisions break down.

This operating model is deliberately channel-neutral. It can be applied to organic, paid, lifecycle and direct journeys, while keeping the website experience and the quality of the resulting customer action in view.

Define the customer job and the business outcome

Start with the progress the visitor needs to make, then name the business outcome that would represent genuine value.

A pricing-page visitor may need to understand plan fit; a campaign visitor may need proof that the advertised promise is credible. Those are clearer starting points than a generic instruction to increase conversion rate.

Pair the immediate event with a downstream quality signal. A form submission can be useful, but qualification, attendance, activation or purchase may reveal whether the apparent improvement reached the right people.

Customer job in plain language
Primary outcome and downstream quality signal
Experience guardrail such as errors, exits or support demand

Diagnose before adding ideas to the backlog

Use behavioural, qualitative and technical evidence to explain the barrier before proposing a treatment.

Review query or ad intent, page hierarchy, analytics events, recordings or feedback, mobile states and the hand-off after the primary action. One signal rarely proves the cause, but several aligned signals can justify a focused hypothesis.

Separate visibility, comprehension, motivation and completion. A CTA may underperform because people never see it, because the promise is unclear, because the offer is weak or because the next step fails. Each diagnosis needs a different change.

Run the smallest test that can change a decision

A good experiment isolates a meaningful uncertainty and states in advance what the team will do with each plausible result.

Google Ads recommends a clear hypothesis tied to a business goal and warns that changing multiple factors can make the result hard to interpret. The same discipline applies on-site: keep the treatment coherent and avoid bundling unrelated redesign work into one comparison.

Not every question needs an A/B test. Usability defects, broken links and inaccessible controls should be fixed directly. Controlled tests are most useful when reasonable people disagree and enough eligible traffic exists to inform the choice.

Turn every result into reusable knowledge

The value of a CRO programme is the quality of its next decision, not the volume of completed experiments.

Record the audience, evidence, hypothesis, treatment, primary metric, guardrail, result and limitations. A neutral result may still show that the assumed barrier was not important or that the test could not detect a useful change.

Review learning across page types and journeys. Patterns can guide prioritisation, but avoid turning one local result into a universal rule. Context remains part of the evidence.

Sources

Test with confidence with the Experiments page — Google Ads Help
Set up events in Google Analytics — Google for Developers
Web Content Accessibility Guidelines 2.2 — W3C

Make the next action easier to find

Barra helps teams organise persistent website actions in one configurable layer while keeping measurement and visitor context in view.

See how Barra works

Install a website toolbar with GTM or a direct script | Barra

Rich Tank — Thu, 27 Aug 2026 09:45:00 GMT

Use a direct script when the site release owns third-party code; use GTM when an established container and publishing process owns deployment.

Quick answer

Install Barra directly when the website codebase and release process should own the script. Use Google Tag Manager when the organisation already governs site scripts through a web container. In both cases, publish the saved domain configuration first and verify the live hostname, loading and rendered behaviour.

A website toolbar can use the same hosted script through two deployment paths. The decision is operational: who owns the change, how it is reviewed and how the team proves that the production page—not only a preview—receives valid configuration.

Barra's public install surface is one script. Interaction helpers may load from the same origin when the selected modules need them, so teams do not add separate customer snippets for chat, feedback, search or experiments.

Choose the ownership model first

Direct installation belongs to the site release; GTM installation belongs to the container's governance and publishing workflow.

Use the direct route when developers manage the site template, consent sequencing and production deployment. Use GTM when an established web container already controls approved third-party scripts and the right team can review, preview and publish changes.

Do not choose GTM only to bypass a release or security review. Google recommends clear account administration and container ownership because one published container can affect every domain it serves.

Install the hosted Barra script directly

Add the canonical Barra script to the site's shared template and keep the loading attributes intact.

The current install snippet is `<script src="https://app.barra.ai/Barra.js" async defer></script>`. The browser hostname is used to resolve runtime configuration, so publish the intended domain and configuration in Barra Admin before testing.

MDN documents how `async` and `defer` affect classic scripts. Barra's snippet includes both as the supported install contract; retain the canonical form rather than inventing a new loader.

Publish the website domain and toolbar configuration in Barra Admin.
Add the canonical script to the shared site template.
Deploy the site through its normal review process.
Load the public production URL and verify the network request, toolbar and selected actions.

Install through Google Tag Manager

Use a Custom HTML tag, fire it on the intended production pages and publish the container after Preview testing.

Google's web-container guidance requires an installed GTM container and access to the website. Barra then runs as a tag inside that governed container; GTM Preview is a diagnostic step, not the production release.

Preview hosts can expose a different hostname from the saved customer domain. Barra fails closed when runtime config cannot be resolved, so a preview-domain failure does not prove that the production domain is misconfigured. Test the canonical live URL after publishing the container.

Confirm the site's GTM web container is installed and owned by the right team.
Create a Custom HTML tag containing the canonical Barra script.
Choose the required production trigger and review consent dependencies.
Use GTM Preview, then submit and publish a named container version.
Verify the public production URL after the version is live.

Run the same production checks for both paths

A successful installation proves config resolution, rendering and interaction on the live hostname.

Check that `Barra.js` returns successfully, the runtime config request uses the expected domain and the toolbar matches the published module order. Exercise links, actions, responsive visibility, docked spacing or floating overlay behaviour, and any selected analytics path.

If the toolbar does not render, check that the domain is published, the snippet or GTM version is actually live, the request is not blocked by Content Security Policy and the page is not an unsupported preview hostname. Barra intentionally does not display demo configuration when runtime config is missing.

Sources

Install a web container — Google Tag Manager Help
Publishing, versions, and approvals — Google Tag Manager Help
The script element — MDN Web Docs
Install Barra — Barra

Frequently asked questions

Is GTM required to install Barra?

No. Barra supports the canonical direct script and a GTM Custom HTML installation. Choose the route that fits your site's ownership and review process.

Choose the install path your team can own

Publish your Barra configuration, install the single hosted script directly or through GTM, then verify it on the real production domain.

Get started with Barra

When persistent CTAs help—and when they hurt | Barra

Rich Tank — Mon, 24 Aug 2026 09:40:00 GMT

Persistent CTAs help when context, relevance and operability are strong. They hurt when urgency or visibility replaces a clear customer journey.

Quick answer

A persistent CTA helps when it keeps a relevant next step available after the page has earned the visitor's interest. It hurts when it arrives before context, covers content or controls, competes with another primary action, or is measured only by clicks.

Keeping an action visible can shorten a journey. It can also turn every scroll into a repeated sales request. The same visual pattern can support one page and weaken another because persistence amplifies both relevance and mismatch.

The useful question is not whether sticky CTAs work. It is whether this action, for this audience and page state, improves a meaningful task without creating a larger cost elsewhere.

Persistent CTAs help after the value is clear

Visibility is most useful when the visitor already understands the offer and may need a convenient route back to it.

A pricing or comparison page can keep a proportionate signup or demo route available while the visitor reviews evidence. A long support journey can retain a relevant contact action without forcing people to abandon the content.

The action should describe what happens next. Generic urgency is a weak substitute for a clear label, credible proof and an appropriate level of commitment.

They hurt when they compete with the task

A persistent action becomes friction when it displaces the content or control the visitor needs right now.

Google Search Central recommends unintrusive alternatives to promotional interstitials that obstruct content. A narrow banner is not automatically safe: chat, consent, feedback and campaign prompts can combine into a crowded edge state.

Test forms, validation, keyboard focus, zoom, the on-screen keyboard and the final page action. If the CTA covers any of them, visibility has overtaken usability.

Match persistence to audience and timing

The same CTA should not appear everywhere merely because it is important to the business.

Choose the pages, devices and visitor states where the action supports a real job. Suppress it where the page already has a stronger primary route, where the visitor lacks context or where a required consent or service control needs priority.

Add an owner and expiry date. Campaign relevance changes, and a once-useful persistent action can become permanent clutter when nobody is responsible for removing it.

Use a four-part test before rollout

Check relevance, geometry, comprehension and evidence before expanding reach.

Relevance asks whether the action fits the visitor's current job. Geometry checks reserve versus overlay and collision states. Comprehension checks the label, consequence and accessible name. Evidence connects interaction to a useful downstream outcome.

Roll out in the smallest representative scope and retain a guardrail. If clicks rise while qualified completion, content use or task success falls, the persistent CTA has not proved its value.

Sources

Avoid intrusive interstitials and dialogs — Google Search Central
What's New in WCAG 2.2 — W3C Web Accessibility Initiative
Understanding Success Criterion 2.5.8: Target Size (Minimum) — W3C Web Accessibility Initiative

Frequently asked questions

Do persistent CTAs improve conversion rates?

They can improve access to a relevant next step, but there is no universal lift. Test the real journey and include downstream quality and usability guardrails.

Keep useful actions available without taking over

Barra gives teams a configurable place for selected website actions, with docked and floating behaviour that can be tested against the real journey.

See Barra in action

Floating action button vs floating action bar | Barra

Rich Tank — Thu, 20 Aug 2026 09:35:00 GMT

Use a floating action button for one dominant action and a floating action bar when several persistent actions genuinely need equal access.

Quick answer

Choose a floating action button when one compact, familiar action clearly outranks every alternative. Choose a floating action bar when a small set of distinct actions must remain available together and the page can support their width, labels and overlap cost.

A floating action button and a floating action bar both sit above ordinary page flow, but they solve different prioritisation problems. Treating the bar as a larger button—or the button as a collapsed menu by default—can hide the real decision about how many actions deserve persistence.

This guide compares the patterns without claiming that either one universally improves conversion. The right choice depends on the visitor job, page context and the interface already occupying the viewport.

Start with the number of legitimate jobs

One dominant job points towards a button; several distinct recurring jobs may justify a bar.

A single contact or checkout action can work as a floating button when the icon or concise label is understandable and the route is appropriate across the page. Adding a hidden tray of unrelated actions turns a simple control into a menu that needs its own discovery and accessibility review.

A floating bar is more honest when search, contact and feedback genuinely need to remain available together. It also consumes more width and attention, so every position needs a stronger justification.

Compare visibility with comprehension

Compact controls save space only if people can understand what they do.

Familiar icons can support a button, but icon-only design is not an excuse to remove the accessible name. A bar creates more room for labels, yet long or translated labels can force truncation or crowd adjacent targets.

Test the pattern at the narrowest supported width and with larger text. If clarity depends on hover, it is not ready for a touch-first context.

Account for overlay and collision

Both patterns float above content, so smaller does not mean consequence-free.

Check whether the control overlaps form buttons, consent choices, help launchers, browser interface or the final page content. A button occupies less area but can still sit exactly where another high-value control needs to appear.

Barra's floating toolbar deliberately does not reserve page spacing. That makes it suitable only where the overlay is accepted and tested. Use a docked, space-reserving alternative when the actions form part of the continuing page structure.

Measure the job, not the shape

Compare patterns through the quality of the resulting journey rather than click rate alone.

Define exposure, interaction, completion and a downstream quality signal. A larger bar may collect more accidental attention; a compact button may be overlooked. Neither result explains value without the next step.

Review by device and page type, and include a guardrail such as content engagement, task completion or error rate. The choice should remain reversible when evidence contradicts the design hypothesis.

Sources

Understanding Success Criterion 2.5.8: Target Size (Minimum) — W3C Web Accessibility Initiative
Avoid intrusive interstitials and dialogs — Google Search Central

Frequently asked questions

Is a floating action button always better on mobile?

No. It uses less space, but it can be unclear, collide with other controls or hide important content. Test the real mobile journey and choose by job, not convention.

Compare the patterns in context

Barra supports docked and floating website action layers so teams can choose the placement that fits the real page.

Explore Barra

Which website actions deserve persistent visibility? | Barra

Rich Tank — Mon, 17 Aug 2026 09:30:00 GMT

Persistent visibility is scarce. Choose actions by visitor relevance, timing, consequence, evidence and the cost of displacing the page.

Quick answer

An action deserves persistent visibility when it serves a recurring visitor job, remains relevant across the pages and states where it appears, has a clear owner and outcome, and creates less friction than the content it displaces.

Teams rarely run out of website actions. They run out of clear places to put them. Contact, search, chat, feedback, offers, consent and campaign calls to action can all be useful, but usefulness in isolation is not enough to justify a permanent place on screen.

The Barra PLACE matrix helps teams make the trade-off together: Purpose, Location, Audience, Consequence and Evidence. It is a decision aid, not a benchmark, and it should be applied to Barra as critically as any other persistent interface.

Purpose: name the visitor job

Start with what the visitor is trying to do, not which team requested the placement.

A persistent search action can help on a large information site where people repeatedly need another route. A webinar promotion may be valuable only on a small set of related pages and for a limited period. Both can have business value, but they do not deserve the same reach.

Write the purpose as a visitor outcome. If the best available wording is merely 'increase engagement', the request is not yet specific enough to earn scarce screen space.

Location and audience: prove repeated relevance

An action should persist only where the same job remains useful for the same audience.

Review page type, journey stage, device and known visitor state. A high-intent pricing page may justify a visible demo route that would be premature on an early educational article.

Barra supports ordered modules and device visibility, so teams can choose a smaller action set rather than copying one global arrangement everywhere. The order itself is placement and should reflect priority.

Name the pages where the action helps.
Name the pages where it should be absent.
Confirm an equivalent route remains when an action is hidden by device.

Consequence: measure the displacement cost

Every persistent action changes the page even when it does not cover content.

A docked bar uses vertical space. A floating action overlays part of the viewport. A large label can compete with the page's own primary action. The right comparison is not action versus no action; it is the action's value versus the attention and space it consumes.

Google advises against intrusive promotional dialogs that obstruct access to content. Several individually modest prompts can still create a combined interruption, so evaluate the complete edge state rather than approving one component at a time.

Evidence: define success and expiry

Persistent placement needs an owner, a measurable outcome and a decision date.

Record the intended progression, not only the nearest click. A contact action might be judged by qualified starts and completed enquiries; a feedback action by usable responses and the issues they help resolve.

Set an expiry or review date before launch. If evidence is weak, test a smaller audience, change the page route or return the action to ordinary content. Persistence is not a permanent entitlement.

Sources

Avoid intrusive interstitials and dialogs — Google Search Central
Understanding Success Criterion 2.5.8: Target Size (Minimum) — W3C Web Accessibility Initiative

Frequently asked questions

How many persistent website actions should a page have?

There is no universal number. Keep the smallest set that serves distinct visitor jobs, remains operable on the narrowest supported viewport and does not compete with the page's primary route.

Give the right actions one deliberate place

Barra lets teams order selected website actions in a configurable bottom layer instead of stacking disconnected prompts.

See how Barra works

Accessible sticky and floating controls checklist | Barra

Rich Tank — Thu, 13 Aug 2026 09:25:00 GMT

A practical accessibility checklist for persistent website controls, covering visible focus, target size, content overlap, labels and real responsive states.

Quick answer

Accessible persistent controls keep keyboard focus visible, provide usable targets and labels, avoid trapping or covering page content, and remain operable when the viewport, text size or competing interface changes.

Sticky bars and floating actions can make an important route easier to reach. They can also create a new obstruction that follows people through the page. The difference is not the visual pattern alone; it is how the control behaves across input methods and page states.

This checklist turns a polished screenshot into a release test. It is not an accessibility certification. Use it alongside the applicable WCAG requirements, assistive-technology testing and the real content around the control.

Keep focused controls visible

A keyboard user must be able to see the control that currently has focus, including page controls behind or beside persistent interface.

WCAG 2.2 added Focus Not Obscured (Minimum): when a user-interface component receives keyboard focus, author-created content must not entirely hide it. A sticky footer, floating panel or expanded action tray can cause exactly that failure.

Tab forwards and backwards through the page with the persistent control open, collapsed and expanded. Include form errors, skip links, anchored headings, consent controls and the final links in the footer. A visible focus ring inside the toolbar does not prove that the rest of the page remains visible.

Test focus before, inside and after the persistent control.
Check both collapsed and expanded states.
Retest when validation messages or cookie controls appear.

Make targets comfortably operable

Size and spacing need to support accurate activation without forcing people to aim between crowded actions.

WCAG 2.2 Target Size (Minimum) sets a 24 by 24 CSS pixel minimum with defined exceptions, including sufficient spacing. Treat that as a conformance floor, not a reason to compress a busy action rail.

Barra's design guidance aims for 44px touch targets where layout allows. The practical test is whether actions remain distinguishable with larger text, on a narrow screen and when a nearby control is also available.

Do not rely on colour alone to distinguish actions.
Keep the accessible name when visible text is shortened.
Remove lower-priority actions before shrinking every target.

Choose reserve space or overlay deliberately

Persistent interface should either reserve layout space or accept and test the consequences of overlaying the page.

Barra's docked mode can reserve the measured rendered height while preserving a larger existing bottom padding. Its floating mode intentionally overlays content and clears that reserve. This makes the placement trade-off explicit rather than hiding it behind styling.

For either model, test the last paragraph, final action, browser zoom, safe areas, orientation changes and the on-screen keyboard. Do not count safe-area space twice, and do not assume a fixed desktop height describes the mobile component.

Review the combined edge state

A control can pass on its own and still fail when chat, consent, campaigns or browser interface occupy the same edge.

List every fixed, sticky and floating element that can appear on the target journey. Decide which element yields, moves, collapses or waits. Raising z-index only chooses which obstruction sits on top.

Finish with real-device testing and a named owner for each exception. Persistent controls remain part of the page, so accessibility review must follow them whenever their actions, labels, dimensions or surrounding interface change.

Sources

What's New in WCAG 2.2 — W3C Web Accessibility Initiative
Understanding Success Criterion 2.5.8: Target Size (Minimum) — W3C Web Accessibility Initiative
Avoid intrusive interstitials and dialogs — Google Search Central

Frequently asked questions

Does passing this checklist prove WCAG conformance?

No. It is a focused release aid for persistent controls, not a complete WCAG audit or certification.

Test the action layer as part of the page

Use Barra's docked and floating controls deliberately, then validate them against the real journey and surrounding interface.

Explore Barra

Design for attention, not interruption | Barra

Rich Tank — Mon, 10 Aug 2026 09:20:00 GMT

The best website prompts do not win by demanding attention. They earn it by appearing in the right context, making a useful offer and respecting the visitor's task.

Quick answer

Design for attention by making the next useful action visible in context, easy to understand and easy to dismiss. Interruption should be the exception, not the operating model for website growth.

Digital teams have more ways than ever to ask for attention. That abundance can create a quiet failure mode: each request may be defensible on its own, while the combined experience tells visitors that every task is secondary to the company's next campaign.

A better standard is simple: make useful actions discoverable without making the visitor negotiate the page.

Attention is a relationship

Visitors lend attention when a page helps them make progress.

A page earns that attention through clear language, relevant proof, predictable behaviour and a visible path forward. It loses attention when prompts arrive before context, cover content or demand decisions that the visitor is not ready to make.

This is not an argument against conversion. It is an argument for conversion that is grounded in the visitor's task.

Context beats urgency theatre

The strongest prompt is often the one that appears when its value is already clear.

A comparison guide can offer a relevant next step after it has explained the trade-offs. A pricing page can make a demo path visible beside the information needed to evaluate it. A persistent action can remain available without repeatedly interrupting reading.

Urgency may occasionally be real, but it should be legible and time-bound. Manufactured pressure makes measurement noisier because it obscures whether people acted from understanding or from confusion.

Show the benefit before asking for commitment.
Use language that explains the action, not only the brand's objective.
Let people continue their task without punishment or repeated obstruction.

Make restraint a growth capability

Restraint is an operational choice that needs shared rules.

Teams can set a default: no new website prompt without an owner, audience, expiry date, measurement plan and a decision about what it displaces. That turns a stream of requests into a considered portfolio.

The result is not a quieter site for its own sake. It is a site where the actions that remain have a better chance of being noticed, understood and chosen.

Sources

Nielsen Norman Group research — Nielsen Norman Group
GOV.UK Design System — GOV.UK
Web Vitals — web.dev

Frequently asked questions

Can a persistent CTA respect the visitor's attention?

Yes, when it is relevant, does not cover content or controls, remains easy to dismiss and is measured against the quality of the visitor journey rather than clicks alone.

Give useful actions a better place

Barra helps teams manage persistent website actions with a more deliberate attention strategy.

Explore Barra

Why 90% confidence can be enough for marketing experiments | Barra

Rich Tank — Mon, 10 Aug 2026 09:00:00 GMT

90% confidence should be the default for most marketing decisions because marketing activity should be reversible and should not create serious impacts. Pair it with effect size, guardrails and a clear decision rule.

Quick answer

Use 90% confidence as the default for most marketing experiments. Marketing activity should be reversible and should not create serious customer, accessibility, legal or trust impacts; use a stricter pre-set threshold when it does.

Marketing teams often inherit a simple rule: do not act until a test reaches 95% confidence. The more useful question is how much evidence is enough for this specific decision.

For most marketing work, 90% should be the default threshold. Marketing activity should be reversible and should not create serious impacts, so the cost of a wrong small decision is usually lower than the cost of slowing every learning cycle. The threshold still needs to be chosen before the data is inspected, with a practical-effect and guardrail plan.

Start with the decision, not the percentage

A significance threshold is a tolerance for false-positive risk, not a universal badge of truth.

In a conventional frequentist test, alpha is set before the experiment. A two-sided alpha of 0.10 corresponds to a 90% confidence interval; alpha of 0.05 corresponds to a 95% interval. A p-value is compared with that pre-set threshold. It is not the probability that the hypothesis is true, and it does not tell you whether the lift is commercially worthwhile.

For a low-cost landing-page message test, acting at 90% may be reasonable if the alternative can be rolled back quickly and the expected upside is material. For pricing, brand claims, a major media reallocation or a change that can harm users, choose a stricter rule.

Set alpha, the primary metric, minimum worthwhile effect and stopping rule before launch.
Record what a positive, negative or inconclusive result will change.
Keep a reversal path where the downside of a wrong decision is meaningful.

Why 90% should be the marketing default

A 90% threshold helps teams learn faster when marketing decisions are reversible and should not create serious impacts.

Marketing is often a sequence of small bets: an ad angle, landing-page hierarchy, CTA description or on-site prompt. Waiting for 95% or 99% on every small decision can mean fewer learning cycles and less useful evidence overall.

That default comes with a responsibility: marketing must not be treated as permission to create serious customer, accessibility, legal or trust impacts. A 90% result with a tiny effect, a noisy metric, repeated peeking or a biased audience is not a reason to declare success. Treat it as a decision signal with stated uncertainty, then monitor the outcome after rollout.

Use expected value and reversibility

Ask what a false positive costs and how easily it can be undone. If a change is easy to reverse and has bounded downside, accepting more uncertainty in exchange for speed can be rational. If it affects trust, compliance, accessibility or a large budget, raise the evidence bar.

Pair confidence with practical significance

Set a minimum effect that would justify the work. A statistically detectable improvement that will not change revenue, qualified demand or visitor experience is not automatically a win. Review the estimated effect and its interval, not only a pass/fail label.

Psychology and medicine do not map neatly to 95% and 99%

The common 95% and 99% shorthand is too simple: thresholds vary by design, consequences, multiplicity and the decision being made.

Psychology has commonly used p < .05 and sometimes p < .01; the American Psychological Association defines alpha as a choice that depends on the consequence of a Type I error. That is a risk decision, not a rule that makes psychology intrinsically a 95% discipline.

UK medical guidance does not make 99% the universal rule. MHRA guidance for clinical investigations describes 5% as the conventional level for two-sided tests, matching two-sided 95% confidence intervals; a typical one-sided non-inferiority test uses 2.5%. NICE likewise describes p < 0.05 as a convention and separates statistical significance from clinical significance. Stricter control can be appropriate for multiple endpoints, safety-critical claims or a particular pre-specified design, but 99% is not the default shorthand for UK medical evidence.

The case for 90% is not that evidence can be relaxed casually. It is that the threshold should match the harm of being wrong, while the analysis still reports uncertainty and protects people from avoidable downside.

A practical operating rule for marketing teams

Use 90% as the default, then raise the bar only when the decision carries serious or hard-to-reverse risk.

Classify the decision as a routine reversible optimisation, a material commercial change or a high-risk customer/trust impact. Pre-set alpha, minimum worthwhile effect, primary metric, guardrails, sample rule and the date or size at which the test will be read.

For routine marketing optimisation, use 90% confidence with a measured rollout and post-launch monitoring. For material or hard-to-reverse decisions, use a stricter threshold and seek replication, holdouts or additional evidence. Report the estimated effect, interval, decision, limitation and next checkpoint.

Default to 90% for routine, reversible marketing tests.
Raise the threshold where a false positive could cause serious impact.
Set the threshold and guardrails before looking at results.
Check practical effect, not only statistical significance.

Sources

MHRA: Statistical considerations for clinical investigations — Medicines and Healthcare products Regulatory Agency
NICE glossary: p value — National Institute for Health and Care Excellence
GOV.UK: Analyse your data for digital health products — Department of Health and Social Care
APA Dictionary: significance level — American Psychological Association
ICH E9: Statistical Principles for Clinical Trials — International Council for Harmonisation

Frequently asked questions

Does 90% confidence mean there is a 90% chance the result is true?

No. In the usual frequentist interpretation, it describes the long-run performance of the interval procedure, not the probability that this specific hypothesis is true.

When should marketers use 95% or 99% instead?

Use a stricter pre-set bar when a false positive could create substantial spend, customer harm, accessibility or trust risk, or when the change is hard to reverse.

Run cleaner website experiments

Barra helps teams manage persistent website actions without adding uncontrolled variables to the visitor journey.

Explore Barra

Measure attention before optimising CTAs | Barra

Rich Tank — Thu, 06 Aug 2026 09:15:00 GMT

CTA clicks are incomplete evidence. Measure whether people had a chance to see, understand and choose the action before deciding what to optimise.

Quick answer

Measure attention as a sequence: exposure to the relevant content, meaningful engagement, CTA interaction and the quality of the next outcome. This helps teams distinguish a weak proposition from a poorly placed or poorly measured CTA.

A call to action can underperform for several reasons: the visitor did not see it, did not understand the value, was not ready to act or encountered friction after clicking. A single click rate cannot separate those explanations.

A useful measurement plan treats attention as evidence, not as a vanity metric. It helps teams decide whether to improve the message, placement, journey or event model.

Define the attention sequence

Attention is the opportunity to make an informed choice, not merely an impression in a browser.

Start with a page view and add context that matters to the decision: reaching a relevant section, interacting with product proof, opening a comparison or viewing a persistent action without it obscuring core content.

Do not turn every movement into an event. Choose a small number of interactions that explain why someone might reasonably choose the next step.

Connect attention to intent

The event model should show progression, not just activity.

Map events to a simple funnel: encountered value, explored evidence, started an action, completed an action and qualified downstream outcome. Different page types will need different evidence, but the language should remain understandable to the people making decisions from the report.

Use consistent naming and document what each event means. Google Analytics provides the mechanics; the business definition is the part the team must agree.

Record the content or component context with the interaction.
Separate a repeated click from a completed next step.
Review mobile and desktop patterns separately when the layout differs.

Turn the result into a decision

A measurement plan should make the next action more obvious.

If people engage with the evidence but do not start the action, clarify the offer or reduce the next-step cost. If they never reach the evidence, revise hierarchy or traffic message match. If they start but do not complete, inspect the hand-off.

This is more productive than endlessly changing button colour because it ties a site change to a specific explanation of visitor behaviour.

Sources

Google Analytics event measurement — Google
Web Vitals — web.dev
Nielsen Norman Group research — Nielsen Norman Group

Frequently asked questions

Should every CTA have its own event?

Track the actions that matter to decisions. Group truly equivalent actions, but retain enough context to know where and why an interaction occurred.

Keep CTAs useful, not intrusive

Barra helps teams present persistent actions without letting them take over the page.

See Barra in action

A conversion rate is not a growth strategy | Barra

Rich Tank — Mon, 03 Aug 2026 09:10:00 GMT

Conversion rate is a useful signal, not a strategy. Sustainable growth requires a clear customer job, credible value, quality measurement and learning across the whole journey.

Quick answer

Optimising conversion rate without defining customer quality and the journey behind the metric can produce activity without durable growth. Treat conversion as one signal in a system that includes value, trust, retention and learning.

Conversion rate is attractive because it is easy to calculate and easy to compare. It can also encourage teams to optimise the nearest measurable action while overlooking whether that action creates value for the customer or the business.

The alternative is not to abandon conversion work. It is to put conversion in its proper place: as feedback from a broader growth system.

Start with the customer job

A better conversion path begins with what the visitor is trying to accomplish, not which component the team wants clicked.

Name the job in plain language. A visitor evaluating Barra may be trying to reduce clutter without losing useful website actions. That is different from a visitor looking for a generic pop-up tool.

When the job is clear, the page can explain trade-offs, show proof and offer a next step that matches the visitor's confidence.

Define quality before optimisation

A high-volume event is not necessarily a high-quality outcome.

Set a primary outcome and a guardrail. For example, an increase in demo requests is meaningful only if the team can later assess fit, attendance or progression. A short-term lift paired with poorer quality is a signal to investigate, not celebrate.

This also protects the team from optimising through pressure, obscurity or accidental clicks. Good growth reduces uncertainty for the visitor as well as for the dashboard.

What customer benefit does this action represent?
What downstream signal would show it was a good outcome?
What experience or trust metric would warn that the change is harmful?

Build a learning loop

The value of an experiment is the decision it improves next.

Capture the premise, audience, change, result and limitation. Share it across media, product and web teams so the next test starts with more context than the last one.

A smaller number of well-instrumented experiments often creates more progress than a long backlog of isolated conversion ideas. The goal is cumulative understanding, not a permanent hunt for a headline metric.

Sources

Google Analytics event measurement — Google
Nielsen Norman Group research — Nielsen Norman Group
GOV.UK Design System — GOV.UK

Frequently asked questions

Is conversion rate still worth tracking?

Yes. Track it alongside the quality and downstream outcomes that make the rate meaningful for the customer and the business.

Create a clearer customer journey

Use Barra to manage persistent website actions with more control and less clutter.

Explore Barra

Barrava's Blog

Conversion rate optimisation: an evidence-first model | Barra

Quick answer

Define the customer job and the business outcome

Diagnose before adding ideas to the backlog

Run the smallest test that can change a decision

Turn every result into reusable knowledge

Further reading

Sources

Make the next action easier to find

Install a website toolbar with GTM or a direct script | Barra

Quick answer

Choose the ownership model first

Install the hosted Barra script directly

Install through Google Tag Manager

Run the same production checks for both paths

Further reading

Sources

Frequently asked questions

Is GTM required to install Barra?

Choose the install path your team can own

When persistent CTAs help—and when they hurt | Barra

Quick answer

Persistent CTAs help after the value is clear

They hurt when they compete with the task

Match persistence to audience and timing

Use a four-part test before rollout

Further reading

Sources

Frequently asked questions

Do persistent CTAs improve conversion rates?

Keep useful actions available without taking over

Floating action button vs floating action bar | Barra

Quick answer

Start with the number of legitimate jobs

Compare visibility with comprehension

Account for overlay and collision

Measure the job, not the shape

Further reading

Sources

Frequently asked questions

Is a floating action button always better on mobile?

Compare the patterns in context

Which website actions deserve persistent visibility? | Barra

Quick answer

Purpose: name the visitor job

Location and audience: prove repeated relevance

Consequence: measure the displacement cost

Evidence: define success and expiry

Further reading

Sources

Frequently asked questions

How many persistent website actions should a page have?

Give the right actions one deliberate place

Accessible sticky and floating controls checklist | Barra

Quick answer

Keep focused controls visible

Make targets comfortably operable

Choose reserve space or overlay deliberately

Review the combined edge state

Further reading

Sources

Frequently asked questions

Does passing this checklist prove WCAG conformance?

Test the action layer as part of the page

Design for attention, not interruption | Barra

Quick answer

Attention is a relationship

Context beats urgency theatre

Make restraint a growth capability

Further reading

Sources

Frequently asked questions

Can a persistent CTA respect the visitor's attention?

Give useful actions a better place

Why 90% confidence can be enough for marketing experiments | Barra

Quick answer

Start with the decision, not the percentage

Why 90% should be the marketing default

Use expected value and reversibility