A CRM manager at a mid-market SaaS company opens her weekly reporting dashboard and notices something odd. Open rates for the newsletter have climbed steadily for three months, but replies, clicks, and trial sign-ups from email have stayed exactly flat. She checks the client breakdown and finds the answer: more than half of her list reads on Apple Mail, and Apple’s Mail Privacy Protection has been quietly inflating every open rate she has tracked since the feature rolled out. The number she used to trust as a proxy for interest has become a proxy for nothing at all.
This is the position most email marketers are in now, whether they have noticed it yet or not. The behavioural signals that used to power segmentation, from third-party cookies to reliable open data, are eroding one browser update and one privacy feature at a time.
The businesses pulling ahead are now the ones asking customers directly, then building automation around the answer.
Activate Your Zero-Party Data
Why Guessing Is Getting More Expensive
For a decade, email marketers built segmentation on inference. Browsing behaviour, purchase history, and third-party cookie data let you guess what someone wanted without asking them. That approach is getting less reliable every year, not because of one dramatic cut-off date, but because of a slow accumulation of restrictions.
Google announced in July 2024 that it would abandon its plan to fully deprecate third-party cookies in Chrome, but the resulting “user choice” model is not the reprieve it sounds like. Chrome now lets users manage cookie preferences directly in Privacy and Security settings rather than through a dedicated prompt, and Google has confirmed that Privacy Sandbox APIs continue to operate alongside cookies rather than replacing them. Safari and Firefox never waited for Google to decide.
Apple’s Intelligent Tracking Prevention has blocked third-party cookies by default since 2020, and Firefox’s Enhanced Tracking Protection has done the same since 2019, which means a large share of web traffic has already been effectively cookieless for years regardless of what Chrome eventually settles on.
Email has its own version of the same problem. Litmus’s ongoing email client market share data shows that more than half of all email opens now happen on a device with Apple’s Mail Privacy Protection activated, which pre-loads tracking pixels the moment an email lands in the inbox rather than when someone actually reads it.
Every automation trigger, re-engagement flow, and open-based segment built on Apple Mail data is now working from a number that no longer means what it used to.
The honest position: third-party cookies did not disappear on schedule, but they became unreliable enough that building a segmentation strategy around them is a bet against a trend that is only moving one direction.
None of this means personalisation stops working. It means the data has to come from somewhere else, and the most reliable somewhere else is the customer.
What Zero-Party Data Actually Is (And Why It Beats Guessing)
The term zero-party data was coined by Forrester Research to describe something specific. Forrester defines it as data that a customer intentionally and proactively shares with a brand, including preference centre selections, purchase intentions, personal context, and how the individual wants to be recognised. That is a meaningfully different category from first-party data, which is what you observe from a customer’s behaviour on your own site and in your own emails, and third-party data, which is purchased or inferred from sources outside your direct relationship with the customer.
The distinction matters because of what each type of data actually tells you. Behavioural data tells you what someone did. Zero-party data tells you what someone wants.
Fatemeh Khatibloo, formerly a VP principal analyst at Forrester, has described zero-party data as valuable precisely because a brand does not have to infer intent when a customer has already stated it directly.
A subscriber who clicks on three running shoe emails might be researching a gift for someone else entirely. A subscriber who tells you in a preference centre that they run and want shoe recommendations has removed the guesswork.
- Cookie networks, data brokers, ad platforms
- No direct consent from the individual
- Increasingly blocked by Safari and Firefox by default
- Opens, clicks, purchases, page views
- Reliable, but tells you what someone did
- Degraded on Apple Mail by MPP since 2021
- Preference centres, micro-surveys, stated intent
- Tells you what someone wants, not just what they did
- Immune to browser and app tracking restrictions
The practical upside for a lean marketing team is that zero-party data does not require a data science function to use.
A preference centre selection can trigger a segment the same day it is captured. There is no model training, no waiting for six months of behavioural history to accumulate confidence, and no dependency on tracking infrastructure that a browser update could break next quarter.
Three Ways to Get Customers to Volunteer What You Need
Collecting zero-party data well is a design problem before it is a technical one. Ask for too much, too soon, and the completion rate collapses. Ask for too little, and the resulting data is too generic to act on. Three mechanisms, used together, tend to produce a data set that is both accurate and genuinely usable.
Preference Centres
A preference centre is the foundational mechanism, and most businesses that have one are using it far below its potential. The common failure is treating it as a frequency toggle, essentially asking subscribers to choose between more emails or fewer emails and calling that personalisation.
A preference centre that earns its place asks about content categories, product interests, and communication channel, and it asks in language that matches how the subscriber actually thinks about your business rather than how your internal team structures its email calendar.
The best placement is not a single link buried in the footer. Surface the preference centre inside the welcome sequence, inside any email where a subscriber shows an unusually low engagement signal, and immediately after any transactional event that reveals a new interest, such as a product return or a support enquiry about a specific feature. Each of these moments gives the subscriber some motivation to engage, and a two-minute update is a low enough cost that reasonable completion rates are achievable.
AI-Prompted Micro-Surveys
Where a preference centre is a static form the subscriber visits, a micro-survey is a single, contextual question dropped directly into an email or a post-purchase flow.
The AI component sits in the sequencing and the follow-up, not in the question itself. A well-built automation can look at what a subscriber has already told you, identify the single highest-value gap in that profile, and ask exactly one question to close it, rather than presenting a long form that asks about everything at once.
This works because it respects the actual psychology of the exchange. Nobody wants to fill in a twelve-field form for a brand they subscribed to three weeks ago, but most people will answer one relevant question embedded naturally in an email they were already reading. The AI layer’s job is deciding which single question is worth asking this particular subscriber this month, based on what is still unknown and what would unlock the most useful segmentation if answered.
Progressive Profiling and Confirmation Loops
The third mechanism is the one most businesses skip entirely: closing the loop on data they already have.
If your platform captured a product interest at sign-up eighteen months ago, it is worth confirming that interest is still current before you keep segmenting on it.
A short, low-friction confirmation, such as asking whether a stated interest is still accurate with a simple yes or update option, does two things at once. It refreshes ageing data before it quietly becomes wrong, and it reinforces to the subscriber that their input actually changes what they receive, which is what keeps completion rates healthy the next time you ask for something.
Zero-party data is not a one-time collection exercise. Treat preferences as data with a shelf life, and build the confirmation loop into your calendar the same way you would a re-permission campaign.
Using Preference Data Without Feeling Like a Data Grab
Collecting this data is only half the framework. Using it in a way that respects the reason it was volunteered is what keeps subscribers willing to keep giving it. McKinsey’s research on personalisation shows that companies excelling at it generate around 40% more revenue from those efforts than average performers, but the same body of research is clear that trust is the constraint that determines whether personalisation works or backfires. McKinsey’s broader analysis finds personalisation lifts revenue by 5 to 15% and improves marketing ROI by 10 to 30% when it is built on data the customer actually consented to, rather than data that feels inferred or purchased.
- ✓ Tells you explicit preference and intent, not an inference
- ✓ Survives browser cookie restrictions and Apple Mail Privacy Protection
- ✓ Builds consent-based trust because the customer chose to share it
- ✓ Works from day one, with no behavioural history required
- − Won't grow your list, it depends on people already subscribed
- − Doesn't replace behavioural data entirely, you still need opens, clicks, and purchases
- − Goes stale without a refresh cadence, preferences drift over time
- − Fails if the ask feels like a data grab, framing and follow-through matter
The practical rule that keeps this honest is simple: never ask for a piece of information you are not prepared to visibly act on within the next two sends.
If a subscriber tells you their preferred product category and the next three emails they receive are generic broadcasts, the preference centre has taught them their input does not matter, and the next request will be ignored. AI-driven segmentation makes this achievable at scale precisely because it can route a preference update into a live segment automatically, rather than waiting for someone on the team to manually rebuild a list.
What This Looks Like in Practice
As an illustration, consider a lean marketing team at a mid-market hospitality brand running one undifferentiated monthly newsletter to twenty thousand subscribers.
If they introduced a preference centre asking three questions, covering destination interest, travel frequency, and preferred contact channel, surfaced in the welcome sequence and in a single dedicated campaign, and paired it with an AI-prompted micro-survey sent to anyone who had not engaged with the preference centre after ninety days, a reasonable expectation would be a completion rate somewhere in the 15 to 25% range for an actively engaged list, in line with typical preference centre benchmarks reported across the email platform industry.
That is enough of the list to build meaningfully differentiated segments without waiting for a full behavioural history to accumulate.
This is a hypothetical scenario, not a reported result, and it is worth treating any completion-rate figure this way until you have your own data to compare against. The mechanism, however, is consistent across the accounts sendXmail has worked with: preference data collected through low-friction, well-placed prompts and then activated immediately through automation produces segments that behave more predictably than behavioural inference does, because segment membership reflects a stated preference rather than a guess.
For teams already running behaviour-triggered win-back sequences, layering in a preference confirmation step before the first win-back send is often the single highest-leverage addition to an existing programme.
Measuring Whether It’s Actually Working
The metrics that matter here are different from the ones most teams default to. Preference centre completion rate is the obvious starting point, but it needs a denominator that makes sense: completions as a percentage of subscribers who were actually shown the prompt, not as a percentage of the full list. A low completion rate against total list size might just mean most subscribers never saw the invitation.
More useful is the first-send match rate, meaning the percentage of the first campaign sent to a newly segmented subscriber that aligns with the preference they stated.
If someone tells you they want e-commerce deals and their next three emails are all about a different product line, something in the automation logic is broken, and it will show up as declining completion rates on future prompts before it shows up anywhere else. Teams already tracking the kind of granular cost breakdown described in our piece on the hidden cost of manual email marketing will recognise this as the same category of execution-consistency problem, just applied to preference data instead of send cadence.
Finally, track engagement delta between preference-based segments and the equivalent behaviourally-inferred segment, where you have both available.
This is the number that justifies the investment to anyone questioning whether the preference centre build was worth the engineering time. If a stated-preference segment consistently outperforms an inferred segment on click rate or conversion, that is the evidence the strategy is compounding rather than just adding administrative overhead.
Is Your List Ready for a Preference Centre?
Not every list needs this investment right now, and building a preference centre before the underlying list and automation infrastructure can support it just adds an unused feature to the tech stack.
- A few thousand active subscribers or more to justify the build
- Segmentation currently relies mostly on guesswork or generic demographics
- Noticeable decline in open-rate reliability from Apple Mail or MPP
- Multiple content types or product lines currently lumped into one stream
- Pre-launch or fewer than a few hundred subscribers
- One consistent newsletter with nothing to meaningfully personalise
- No automation platform connected to act on preference data yet
- Main challenge is deliverability infrastructure, not personalisation
If the checklist above points toward “not yet,” the more urgent work is usually deliverability and list hygiene rather than preference collection. Personalisation built on top of a list with weak sender reputation just gets more targeted messages into fewer inboxes.
If the checklist points toward “good fit,” the sequencing matters: get the preference centre live and the automation routing built before investing heavily in the AI-prompted micro-survey layer, since the second mechanism depends on knowing what gaps in the first one are worth asking about.