Zenefits Tests Decision-Guide Cards vs. Single CTA in Its Hero

The ABWatcher blog

Zenefits Tests Decision-Guide Cards vs. Single CTA in Its Hero

Zenefits is running a live Optimizely hero test pitting a generic search-bar CTA against two comparison cards and an assessment quiz. Here's what the variants reveal about lead segmentation design.

Aiko Tanaka

Design Director · · Updated

TL;DR

  • ABWatcher detected a live hero A/B test on zenefits.com with 85% confidence, flagged via vendor_sdk_active (Optimizely).
  • Variant A (visits 1–4, 7): subhead + search bar + single "Book a Demo" CTA.
  • Variant B (visits 5–6): two comparison cards — "HR, fully handled" vs. "HR that fits you" — with "Explore PEO" / "Explore HR Plus" links plus a "Take our Assessment" CTA.
  • Same browser (Chrome) rendered both variants across different visits, ruling out browser/engine-based rendering as the cause.
  • Likely hypothesis: an interactive decision-guide widget improves lead segmentation quality versus a generic demo-request CTA.
  • Detection logged across 7 tracked visits, with variant split occurring at visits 5–6.

What does each Zenefits hero variant actually show?

Variant A leads with a search bar and one CTA; Variant B replaces that with two side-by-side comparison cards and a quiz-style CTA. In Variant A, the eye has one job: scan the subhead, drop into the search field, or hit "Book a Demo." It's a single-column decision — low cognitive load, but also low information. There's no differentiation of who the visitor is or what HR problem they're trying to solve. It's a generic top-of-funnel net.

Variant B restructures the hero as a fork. Two cards sit in visual parity — same size, same weight, same button treatment — which is a deliberate similarity cue (Gestalt's law of similarity) telling the visitor "these are comparable options, pick one." The headlines do the segmentation work before the visitor reads a single word of body copy: "HR, fully handled" signals full outsourcing (PEO), "HR that fits you" signals a lighter, self-service model (HR Plus). That's a classic binary-choice pattern lifted straight from pricing-page design, just moved up into the hero.

Why put a decision fork above the fold instead of a single CTA?

Because a single generic CTA forces all visitor intent through one narrow funnel, while the card layout pre-qualifies leads before they ever reach a form. A search bar assumes the visitor knows what they're looking for and is willing to type. Most aren't, and most won't. Cards convert that friction into a tap. The "Take our Assessment" CTA raises the segmentation ambition even further — it's not just "pick a product," it's "answer questions so we can route you," which reads as an attempt to capture firmographic and need-state data before a sales conversation ever starts.

From a hierarchy standpoint, this is the more interesting variant to critique. Two cards plus a third assessment CTA is three competing actions in one hero. That's a real contrast risk: if "Take our Assessment" isn't visually subordinate to the two primary cards, the fork gets muddled into a triple-choice, and choice paralysis creeps back in. The test is likely also quietly measuring whether that third path cannibalizes clicks from the two comparison cards.

How does the F-pattern change between the two versions?

Variant A supports a fast top-left-to-CTA scan; Variant B forces a left-to-right comparison before any action. A search bar sits comfortably inside a standard F-pattern: headline, subhead, single action, done — very little horizontal eye travel required past the first line. The two-card layout breaks that pattern deliberately. It asks the visitor to read horizontally across both cards before committing, which is slower but arguably higher-intent. This is the classic CRO trade-off: speed-to-click versus quality-of-click. A generic "Book a Demo" click could come from anyone, including buyers who don't fit Zenefits' ICP. A click on "Explore PEO" after reading two competing headlines is a self-selected, better-qualified signal for sales.

What does the test setup tell us about how it's being run?

ABWatcher's detection method (vendor_sdk_active) confirms this is a live, vendor-run experiment rather than a rendering artifact, geographic redirect, or caching quirk. The tell: the same Chrome browser rendered Variant A on visits 1–4 and 7, then Variant B on visits 5–6 — meaning the split is session- or cookie-based bucketing from an experimentation platform, not a browser-engine difference. That single detail matters more than it looks. Teams debugging inconsistent screenshots often assume it's a caching or device issue; this is the pattern that says "no, it's a live experiment," and it's exactly the signal Split Hero and similar A/B plugins are built to generate and measure.

Here's the variant breakdown as detected:

Attribute Variant A (visits 1–4, 7) Variant B (visits 5–6)
Hero structure Subhead + search bar Two comparison cards
Primary CTA(s) "Book a Demo" (single) "Explore PEO" / "Explore HR Plus"
Secondary CTA None "Take our Assessment"
Segmentation attempt None (generic) Product-line self-selection
Visual pattern Single-column, linear scan Side-by-side, comparative scan

What should product and design teams take from this?

The underlying bet Zenefits appears to be testing — swap a generic CTA for a self-segmenting decision widget — is a well-worn CRO move, and the research on hero-section changes consistently shows that even small copy or structural shifts above the fold can swing conversion meaningfully; one widely cited hero rewrite alone produced a 221% lift in a single test. The craft risk here isn't the hypothesis, it's the execution: three CTAs above the fold demands very deliberate contrast hierarchy, or the fork becomes a maze.

Common questions

What is Zenefits actually A/B testing on its homepage?

Zenefits is testing its hero section: a simple search bar with one "Book a Demo" CTA against two comparison cards ("HR, fully handled" vs. "HR that fits you") plus a "Take our Assessment" CTA. The test was detected live on zenefits.com starting from the tracked visit sequence, with an 85% confidence rating.

How was this test detected as a real experiment and not a rendering bug?

ABWatcher flagged it via vendor_sdk_active, meaning an experimentation SDK (consistent with Optimizely-style tooling) was active on the page. The same Chrome browser showed Variant A on visits 1–4 and 7, then Variant B on visits 5–6, confirming session-based bucketing rather than a browser or device rendering difference.

Why would a company replace a search bar with comparison cards?

A search bar assumes visitors already know what to look for; comparison cards pre-segment visitors by need (full HR outsourcing vs. lighter self-service tools) before they ever submit a form. This typically produces fewer but higher-intent, better-qualified clicks for sales follow-up.

Does adding a third CTA like "Take our Assessment" hurt conversion?

It can, if not visually subordinated to the primary two cards — three competing actions above the fold risks choice paralysis. Design hierarchy (size, color weight, placement) needs to clearly signal that the assessment is a secondary, lower-commitment path relative to the two main product cards.

What should teams do with this detection on their own roadmap?

Audit your own hero for single-path versus segmented-path CTAs this sprint, and if you introduce a decision-fork test, cap it at two primary actions with one clearly subordinate secondary link to avoid diluting the click signal you're trying to measure.

Read next

More from the ABWatcher blog

From the glossary

Vocabulary the article uses. See all terms →

  • Shrinkage Estimator— A statistical adjustment that pulls noisy per-variant or per-segment estimates toward an overall average to reduce overfitting to randomness.
  • Flicker Test— A diagnostic check to confirm whether users briefly see the original page before a variant renders in client-side experiments.
  • Pre-Experiment Bucketing Audit— A review process confirming users are being randomly and evenly assigned to experiment variants before launch.
  • Above-Fold Exclusion Rule— A test-design rule that deliberately prevents changes to above-the-fold elements so an experiment isolates below-the-fold effects.
  • Flicker-Free Testing— A technical setup that prevents users from briefly seeing the original page before a test variant loads.
  • Orthogonal Test Design— A test structure where multiple variables are changed independently so their individual effects can be measured without interference.

See more like this

ABWatcher catches A/B tests like this across 1,000+ brands.

Watch live experiments at 1,000+ high-converting brands, complete with hypothesis and takeaway. 7 days of Pro, free, no card, then Free forever for ten watched companies.