How to Use AB Tasty to Personalize for Returning Visitors

The short answer: keep at least 30% of returning visitors on the original, because AB Tasty's report will not call a goal ready until each version has had 14 days, 5,000 unique visitors and 300 unique conversions, and at the default allocation of 100% to the experience there is no original to compare (readiness rule, traffic allocation). With that settled: install the tag in the head, set the consent mode, build a returning-visitors segment, create a Simple Personalization, make the change with Evi Content drafting and you validating, pick one primary goal, split 70/30, attach the GA4 connector, QA it, launch, and read the chance to win once the report says ready.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

AB Tasty's website now redirects to Wingify after its merger with VWO, but the product's documentation at docs.abtasty.com is still the source for every step below; the AB Tasty glossary entry covers what the merger changes for a buyer. The job: a visitor who came back sees a different hero, and you learn whether that moved a number you care about.

Before you build:

1. Do the readiness arithmetic. For example, 12,000 returning visitors a month at a 70/30 split puts 3,600 a month on the original, so 5,000 unique visitors takes 5,000 / 3,600 = 1.4 months, and 300 conversions at a 3% rate needs 10,000 visitors on the original, so 10,000 / 3,600 = 2.8 months. If that is too long, split 50/50 for the first run and call it a test.
2. Start from the cookie you already set. A returning visitor is anyone whose ABTasty cookie counts more than one session; the cookie lives 13 months, and new and returning are not mutually exclusive, so one person appears in both across sessions (new or returning criterion).
3. Fix the goal before launch. The primary goal cannot be changed once the campaign has started, and action trackers collect nothing retroactively (goals step).

What you need

• An AB Tasty account with Web Experimentation and Personalization; a new buyer is quoted by Wingify on monthly active users, no price published (Wingify pricing, checked on 5 November 2026).
• Access to the head of your pages, or Google Tag Manager.
• A consent decision, taken with your legal counsel.
• A GA4 property, if you want the campaign in your own reports.

1. Install the tag, synchronous, in the head

On the Tag implementation page, enter the domain, select Standard, copy the synchronous tag and paste it into the head "as early as possible", deploy, click I'm done, and confirm with the AB Tasty Chrome extension, which shows the tag version and weight (tag setup). The docs recommend synchronous in most cases, because asynchronous loading can show the original for a moment before the change applies (synchronous or asynchronous).

2. Set the consent mode

Under Settings, Cookies & privacy, Privacy, the Restrict cookie deposit setting has three positions: no restriction, consent managed by your tag manager, or consent delegated to the AB Tasty tag (consent policy). The returning-visitor segment reads the ABTasty cookie, which lives 13 months, and storing it is the kind of access to a device that Article 5(3) of the ePrivacy Directive conditions on consent unless strictly necessary. Which position fits your site is your counsel's call.

3. Build the segment

Open the Segments library, click Create segment, drag New/returning visitors into the working area, name it and save; a saved segment is reusable in later campaigns (segment guide).

4. Create the Simple Personalization

Go to Personalization, click Create and choose Simple Personalization (simple personalization guide). The Main information step asks for a name, a hypothesis and one sample URL with no parameters; Evi Hypothesize keeps a checklist of what the hypothesis still lacks and proposes a complete one you can accept or edit (main information step).

5. Make the change, with Evi drafting and you approving

In the Visual Editor, select the hero element, choose Make a change, and type the request: a new headline for people who have been here before, a different button. Evi Content generates the modification and its code, shows a preview, and waits for one of three answers: refine, leave and edit by hand, or validate (Evi Content guide). Nothing enters the variation until you click validate, which is the human check. Select one element at a time; a whole section trips a context limit error.

6. One primary goal

In the Goals step, pick the metric the change should move most as the primary goal, and add up to 50 secondary goals (goals step). Goals are prefilled from the editor's trackers in order, so check that the first one is the one you mean.

7. Targeting: who, where, how

WHO takes the segment from step 3, which the docs call the most important section; WHERE takes a URL, a saved page or a page type; HOW is an optional trigger, for example the Source criterion with contains and Instagram (source criterion).

8. Hold back the original

By default all targeted traffic goes to the experience. The docs recommend leaving 20 to 30% on the original, and never less than 10% (traffic allocation). Set 70/30 and leave it: a later change applies only to new visitors.

9. Advanced options: GA4 and alerts

Attach the Google Analytics 4 connector, created once in the Integration Hub with your Measurement ID; it sends an abtasty event carrying abtasty_campaign and abtasty_variation, which you register as custom dimensions in GA4, and it is not retroactive (GA4 integration). Leave loading on deferred, the default (loading modes). Sequential testing alerts flag an experience that is losing conversions; they need a conversion or transaction rate as primary goal (sequential alerts).

10. QA, then launch

In the QA & Overview step, enable the IP address or cookie condition; the URL parameter option exists only for A/B tests. That puts the campaign in QA state, visible to you alone, and a creator or viewer role cannot launch it (QA mode). Launch it, or schedule the start.

11. Read it when the report says ready

The report shows Bayesian statistics once a goal is ready: a chance to win, reliable at 95% or above, and a 95% interval with a median gain (statistics guide). A result that was reliable yesterday and is not today had too small a sample; the docs say to wait a few more days so weekdays and weekends are both in (reliability troubleshooting).

For status changes, AB Tasty documents a Zapier app with triggers for Campaign Status Change, SRM Triggered, Sequential Testing Triggered and Learning Created, with guides for Jira, Asana and monday.com (Zapier integration). Wire SRM Triggered to the channel your team reads; the Zap carries the alarm, and a person decides whether to pause.

What the docs warn about

• Flicker comes from an asynchronous tag, a tag placed low, deferred loading or a tag over 120 kb; the fixes are the synchronous tag right after the opening head and Instant loading for that one campaign (flicker troubleshooting).
• GA4 will not match. AB Tasty counts a visitor's sessions from first exposure and GA4's scopes count differently; compare users, not sessions (GA differences FAQ).
• The Reporting Copilot skips you. Evi's analysis applies to A/B and multipage tests with a transaction primary goal and over 1,000 visitors on the original, not to personalizations (reporting copilot).

What the default is for

The allocation default is 100% to the experience, and a default is a decision someone else took for you: a campaign at 100% counts as live the minute it launches, which suits a vendor whose pricing tiers count concurrent personalization campaigns, 3 on Growth and 5 on Pro at Wingify, better than a campaign held against an original (Wingify pricing). The holdback that proves the campaign paid is the number you type in yourself. After the first ready result, write down the chance to win, the median gain and the GA4 split for the same dates.

FAQ

Can I show the personalization before the visitor accepts cookies?

The docs say display does not need consent when it is not based on personal data, but a segment built on the ABTasty cookie needs the cookie, and without consent the report stays empty (consent troubleshooting).

Why do new and returning add up to more than 100%?

Because they are counted per session on the same cookie: a visitor is new in the first session and returning in the next, so one unique visitor sits in both segments (new or returning criterion).

Can an AI agent run this campaign for me?

Not end to end. Evi drafts inside the editor, AB Tasty's MCP server serves Feature Experimentation & Rollout flags rather than web campaigns, and the Zapier triggers send events out rather than taking commands in (MCP server on GitHub).