What is PAF and why is it vital to UK businesses?
The Postcode Address File, or PAF, is the UK’s definitive address database. Owned and maintained by the Royal Mail, the data is updated every day, as approximately 90,000 postal workers go about their deliveries and make note of any changes. These daily updates, which Fetchify include automatically in our address search and validation databases, will involve newly built houses, demolished houses, older buildings split into multiple homes or offices, etc.
The first PAF database, containing every residential and business address in the UK, was put together in the 1980s. Although the address information was being collected digitally, the filing system was still on paper, making updates a laborious and infrequent process. Of course, as technology has advanced, the system has been steadily upgraded and these days, the Royal Mail gathers and disseminates 4-5,000 data updates every day. In recent years, the biggest change to PAF has been a collaboration with the British Forces Post Office to reformat BFPO addresses to match the standard PAF address layout, and to include these addresses in the PAF dataset.
Why is PAF important? Gathering accurate customer data is vital to so many businesses, whether ecommerce, government organisation, financial services, etc. An inaccurate address can mean your deliveries go missing, a bill could go to the wrong house, you could waste money on paper marketing that doesn’t find its target. Any of these, or a plethora of other issues, will cost your business time in fixing the problem and speaking with angry customers on the phone, money in lost goods, lost packaging and wasted postage, and ultimately, lost customer loyalty. With Fetchify’s simple address lookup and validation API, you can avoid these pains before they even happen.
Our API allows you to search for any UK address by typing in just a few characters. There are two different PAF-powered options for finding and validating UK addresses: Postcode Lookup and Address Auto-Complete. How do they work?
With UK Postcode Lookup you type in the postcode and then select the correct address from the drop-down menu.
Address Auto-Complete is our global address search and validation solution and PAF is only one of many different data sources that we use. With Address Auto-Complete, you can start typing from any point in the address, whether street name or postal code. The type-ahead address search tool delivers address results from as little as four characters. Simply select the correct address when it appears. Another benefit to Address Auto-Complete is that you can find the right address even if you don’t know the postcode, for example when sending a gift.
The Postcode Address File lists every deliverable point in the UK. For buildings that are subdivided into multiple homes or businesses which share a single letterbox, PAF can be enhanced by Multiple Residence data from the Royal Mail. The Mutiple Residence Enriched dataset is available as an addition to the Address Auto-Complete search. If you think this additional data could be useful for your business, get in touch at support@fetchify.com and our team will be happy to help.
Our best in class address lookup and validation API, uses the most up to date and definitive UK and international data sources. This simple addition to your checkout increases conversion rates by substantially reducing the hassle of long forms, increasing your customer’s trust in your website and providing a smoother user experience. The validated address data also reduces failed deliveries caused by customer errors, saving you time and money, providing exceptional return on investment, and making sure you reach the right address every time.
How data decay is quietly removing your best customers before they ever decide to leave. Somewhere in your CRM right now, there is a customer you think you lost. They stopped buying about eighteen months ago. They went into a lapsed segment, got a couple of reactivation emails, did not respond, and were eventually written off. The assumption was that they moved on. What actually happened, in a surprising number of cases, is much simpler. They moved house. The reactivation emails went to an inbox they no longer check. The direct mail went to a flat that has a different tenant. The customer was not gone. They were just unreachable. And because the database had no way of flagging the difference, they were counted as churn. This is how data decay works. Not in dramatic failures, but in a steady accumulation of records that have quietly stopped being accurate. Around 30% of customer data goes stale every year, not because anything went wrong, but because people move, change jobs, switch email addresses, or get married. Left unaddressed, that figure compounds. A database that has not been properly maintained for three years may have a third of its records either partially or wholly unreachable. The problem is that it is almost invisible until it is already significant. A handful of bounced emails does not raise an alarm. Neither does a slightly elevated returns rate. The metrics look broadly normal because the volume of bad data is not yet high enough to distort them. By the time it is, the damage is done. The churn you cannot account for Most businesses have a reasonable handle on the customers they actively lose. Cancellations are tracked. Lapsed accounts are flagged. Retention programmes exist precisely to address the customers who stop buying. What those programmes cannot reach is the customer who never formally left. They sit in the CRM as a lapsed record. They count toward the database size. They get included in reactivation segments. They cannot receive the communication because the address on their record is no longer valid. The downstream effect is real. A repeat customer whose address changed after a house move never receives the offer that would have brought them back. A lapsed member does not see the renewal reminder and lets the subscription quietly expire. In both cases, the organisation records an attrition event. In neither case did the customer actually decide to leave. A customer who moved house is not the same as a customer who left. That distinction tends to matter quite a lot when you are trying to work out where your retention budget should go. Why reactivation campaigns underperform When a win-back campaign comes back with poor results, the instinct is to interrogate the campaign. The subject line gets tested. The offer gets more aggressive. The timing gets adjusted. All of that is reasonable. None of it helps if a meaningful share of the list cannot receive the email in the first place. A lapsed customer segment typically contains three types of contact: people who genuinely disengaged and are unlikely to respond, regardless, people who might respond to the right message, and people who would respond, but the email never arrives because the address has changed. The frustrating thing is that you cannot easily tell these groups apart from the outside. Low open rates and low click-through rates look the same whether the cause is disengagement or data decay. Email is only part of it. Physical address decay affects direct mail and delivery. Phone number decay affects SMS and outbound calling. Each channel erodes at its own rate, and most organisations are not tracking the accuracy of their data across all of them. 30% of customer database records become inaccurate within 12 months, without any action by the customer. What changes when the data is clean A data cleanse does not just improve deliverability, though it does that. It changes what the numbers actually mean. When ghost records are removed from a lapsed segment, the remaining file is smaller but more meaningful. Reactivation revenue from that cleaned list is real revenue, not a percentage improvement calculated against contacts who were never going to respond. The churn figure, once recalculated without the unreachable records, is often more positive than expected. Some of what looked like permanent attrition turns out to be recoverable. There is a GDPR dimension too. Article 5(1)(d) requires that personal data be kept accurate and, where necessary, up to date. The ICO can issue fines of up to £17.5 million for data accuracy failures. Most organisations are not at serious risk of enforcement, but most organisations also have not checked how their database holds up against a standard they are legally required to meet. The more common consequence is commercial rather than regulatory. Marketing budgets applied to an inaccurate list simply do less than they should. The same spend, against a validated file, produces measurably better results. Not because the campaigns improved, but because the contacts can actually receive them. The practical starting point Addressing data decay does not require a significant IT project. For most organisations, the starting point is a cleanse of the existing CRM: matching records against current address databases, identifying email addresses with persistent bounce history, removing duplicates, and flagging phone numbers that are no longer in service. Done once, it resets the foundation. Done regularly, and combined with validation at the point of data capture, it prevents the drift from accumulating again. The customers in those unreachable records did not all decide to leave. Some of them are still out there, still buying in your category. They just moved. Improve your data health and protect your business today. Reach out to our team below for a free data health check.
Jay’s career has never followed a straight line. Electronics engineering. Automotive systems. A social app for hostels that was about to launch just as COVID closed every hostel in the world. A pivot into web development. And eventually, Fetchify - where he now leads the team building the technology that keeps millions of data lookups running accurately every day. Looking back, the route makes perfect sense. Jay has always been drawn to what’s next. To faster feedback. To building things that work and seeing them work quickly. Software gave him all of that in a way that automotive engineering, for all its complexity, eventually stopped doing. The long way round Jay studied electronics engineering and came out of university specialising in embedded systems. By 2015, he was working on automated parking systems - the kind built on sensors and split-second decisions - and for a while, he found it genuinely interesting. But something was missing. “I wanted to see results faster,” he says. “With embedded systems and automotive work, the feedback loops are long. I wanted to build something and see it working.” So, he pivoted. He taught himself mobile development and from there, a startup building a social app for hostels and hotels - a platform that matched guests by shared interests, so someone travelling alone could find other guests up for the same activities. It was a genuinely good idea, with a handful of places trialling the beta version. Then 2020 arrived, the hospitality industry stopped overnight, and the timing simply couldn’t have been worse. Most people would have counted it as a setback. Jay counts it as part of the story. Finding something that fits He joined ClearCourse, initially working on the membership CRM side of the business. When a role came up at Fetchify, he knew it was the one. Tech Lead. A team to run. Real scope to build, improve and innovate - and enough space to do it properly. “What I love most about my job is the chance to be innovative and improve the quality of the software - and the opportunity to keep learning. There’s always something new.” His approach to leading the team reflects the same values. He talks about trust a lot - giving people the space to do things the way they think makes sense, rather than prescribing the path. The team checks in daily, whether that’s to swap ideas, talk through a problem, or join a scrum call. It’s not just his immediate team either: the wider Fetchify team, and within the ClearCourse group, there’s a culture of helping out. Of people being willing to lend a hand when it’s needed. “Software development can feel like a solo job, but actually the team here is solid, and we enjoy working together.” The thing he's most excited about Ask Jay what he’s most passionate about right now, and the answer is immediate: AI. Not in an abstract, trend-chasing way - but with a specific and considered view of what it actually means for software developers and the organisations they build for. “AI is raising the bar for what developers can produce. But I see it as a two-way collaboration - a helping hand to do the grunt work, while the ideas, the creativity, the innovation still come from people. It should help people achieve more in less time. Not replace the thinking.” His long-term goal is to help other ClearCourse businesses integrate AI into their products - starting, naturally, with Fetchify. For a company built on data accuracy, the intersection of clean data and AI capability is not an abstract future conversation. It’s already the direction of travel. Beyond the screen Jay grew up in Egypt, and travel is still one of the things he values most. He heads home to family a couple of times a year, and fits in city breaks wherever he can - somewhere new, with good food and different people and things to explore. His ideal off-duty scenario involves a beach, good conversation, and absolutely no particular agenda. The gym, friends and music round it off - time away from the screen that, for someone whose working life involves building technology that processes millions of data points a day, seems like a fairly sensible skill. When he imagines the distant future - the looking-back version - he pictures a career of creation, innovation and the willingness to embrace whatever comes next. That, and a beach somewhere warm. We’re very glad the winding road brought him to Fetchify.
“Fetchify turned what felt like a crisis into a straightforward fix - and in just a couple of days. We went from not being able to contact anyone to generating four new client applications from a single send. The data cleanse didn't just fix a problem - it opened the door again.” – Marcel Stirling, Phoenix Insolvency
There is a lot of enthusiasm right now about what AI can do for ecommerce and CRM teams. Personalisation at scale. Predictive analytics. Automated outreach that learns and adapts. The pitch is compelling, and much of it is real. But there is a foundational question that almost nobody is asking loudly enough: what happens when you run AI on bad data? The answer is not that the AI fails gracefully. The answer is that it fails at scale, confidently, and in ways that are harder to trace than a simple spreadsheet error. This is not a theoretical risk. It is already happening inside the organisations that have moved fastest to adopt AI-driven tools without first addressing the quality of the data those tools run on The assumption nobody questions Most organisations treat AI as a layer that sits on top of their existing data. Feed in the CRM, connect the customer database, and point the model at the transaction history. The assumption is that AI is smart enough to work around imperfections. It is not. AI systems are pattern recognition engines. They find what is consistent in the data and treat it as a signal. If your data consistently contains errors - outdated addresses, duplicate records, lapsed contacts still marked as active - the AI learns those patterns as the truth. It bases its predictions, segments, and recommendations on a foundation that does not reflect reality. B2B contact data decays at 30% per year. For a database of 100,000 records, that means 30,000 entries become inaccurate every 12 months. When an AI personalisation engine is drawing on that data to decide who to target, when to contact them, and what to offer, it is working with a picture of your customer base that is one-third wrong AI doesn't fix bad data. It amplifies it. What this looks like in practice The problems that emerge are not dramatic. They are quiet and cumulative, which makes them harder to catch. Automated email sequences reach the wrong people or the wrong addresses, generating hard bounces that damage your sender reputation and, in serious cases, trigger blocks from email service providers. Personalisation that references a customer's last purchase or location draws on a record that has not been updated in two years. Predictive models identify high-value customers to target for retention campaigns - but a portion of those customers moved, changed roles, or lapsed long ago. Each of these is a cost. Collectively, they represent a significant drag on the performance of tools that were supposed to be driving efficiency. The irony is that AI makes these problems less visible, not more. A human reviewing a list might notice that an address looks wrong. An AI processes it at speed and acts on it. A case study: what happens when AI meets dirty data A professional services firm recently experienced this directly, who work with our sister company FLG for lead management. The team began bulk emailing an existing database through their email marketing system - a reasonable use of automation for a business trying to re-engage contacts at scale. The data, however, was old. Hard bounces accumulated quickly, and their account was flagged and blocked from sending. Fetchify cleansed the data. Contact information was standardised, and inactive or undeliverable entries were identified and removed. When they resumed outreach, the results were immediate - higher engagement, no delivery issues, and the kind of performance the automation was always supposed to deliver. The AI-driven outreach did not fail because of the tools. It failed because the data had not been maintained. Once the data was clean, everything else worked as intended. The AI readiness question organisations should be asking As AI becomes a standard component of ecommerce and CRM operations, the conversation around data quality needs to change. It is no longer just a compliance issue or an operational nicety. It is a prerequisite for AI to function as intended. Before deploying any AI-driven personalisation, automated outreach, or predictive analytics tool, the right question is not 'which AI platform should we use?' It is 'is our data clean enough for AI to learn from?' For most organisations, the honest answer is no - not without first running a data cleanse. The good news is that this is not a complex or expensive process. It is a one-time exercise that resets the foundation, followed by ongoing validation to prevent decay from accumulating again. What clean data actually enables Organisations that address data quality before deploying AI achieve fundamentally different outcomes. Personalisation engines draw on accurate records and produce recommendations that reflect the real customer base. Automated outreach reaches real inboxes and generates real responses. Predictive models identify genuine opportunities rather than ghost records. The regulatory dimension is worth noting, too. The ICO can issue fines of up to £17.5 million or four per cent of global annual turnover under UK GDPR for data governance failures. AI that acts on inaccurate or out-of-date data does not protect organisations from that exposure - it amplifies it, at speed and scale. Clean data is not an enhancer of an AI strategy. It is the essential prerequisite that makes an AI strategy viable. The organisations seeing the best results from AI aren't necessarily the ones with the best tools. They're the ones with the cleanest data. Start with a free data health check and find out where you stand.
