Everything you need to know about autofill

Fillr Autofill.png

Somewhere in the not too distant past, there would have been a business owner or two who scoffed at the notion of creating a website. ‘What’s the point?’ you can just imagine them saying. ‘We’re doing just fine without it, so why should we change?’

Fast forward to 2017, and experts claim that you are either a digital business, or a dead business.

And so is very likely going to be the case for many digital unicorns and tech of our era, including autofill. But what is autofill, really? It might very well be the answer to mobile shopping cart abandonment,  the magic ingredient aggregator apps need to thrive,  and the secret to successful shopping coupons,  but how does it even work?

Turns out autofill is a bit more complex and a lot more interesting than you ever thought.

For the novice and for the expert, here’s everything you need to know about this vitally important, up-and-coming service.

What is autofill, really?

On its surface, autofill, the act of auto-filling in anywhere between one and hundreds of form fields, seems like a relatively simple task. Your autofill program simply stores your info and retrieves it every time it sees a form.

You get to fill in said forms up to 536% faster than you otherwise would.

You save a truck load of time, you don’t have to fat-finger your way around form filling on your mobile, and you get to get on with your life.

Yet, the majority of our experiences with autofill don’t even vaguely resemble this.

Firstly, there are huge accuracy problems with most autofills. Research has showed that browser autofills are woefully inaccurate, with some filling form fields accurately only 53% of the time.

Secondly, not only are most autofills inaccurate, they also present a security risk. With some autofill add-ons that don’t have a cloud back-up or sync, the browser itself needs to save your information and it’s usually synchronised with your Google accounts, making you vulnerable to attacks. That, and you really never know what forms your autofill is filling for you. You might think your autofill has just popped in your name and address, but hidden bank detail forms could be lurking in the background, and usually you’ll be none the wiser.

So therein lies the actual experience of autofill for most users – an inaccurate, non-secure service where you’ll most likely have to heavily edit your information, and you’ll be putting your bank details (and a swathe of other highly personal information) at risk.

So why doesn’t autofill work like it should?

The reason that most of our experiences of autofill are so disappointing is that autofill is a far more complex process than meets the eye.

Autofill could work well, in theory, if all web forms were exactly the same, and all data stored in our browsers was secure. Sounds comical, doesn’t it? But unfortunately, those are the very assumptions that most autofills are built on.

As anyone who has ever autofilled a form will know, web forms vary wildly in their presentation and makeup. This is because web developers vary wildly in how they approach creating a web form, meaning that the HTML code used in forms can be almost as unique as fingerprints. For the autofill program, this means that what might seem like a simple name or address field could be very hard to read and understand. And if the autofill doesn’t understand the field it has, of course, only a small chance of autofilling it correctly.

Even if the autofill does understand a field, knowing what to put in there can be equally challenging. With any given website, there is an almost unlimited set of plugins that decorate and validate different form elements, meaning it’s hard to know what should go where. Particular JavaScript UI elements can also wreak havoc on autofill’s ability to do even a simple assignment to a field value. Think about it – how many times has your autofill got your address, the date, or other common fields wrong because the page wanted them in a format your autofill didn’t recognise?

Then there’s the issue of security, which is something that everyone, quite frankly, can’t afford to ignore. Any Google search on ‘browser autofill security’ reveal literally thousands of chilling stories of hackers obtaining personal data, including bank details. Basically, it simply isn’t safe to store personal data on your browser, and browser-based autofills don’t offer an ‘option b,’ unfortunately.

How did we get here?

So how is it that some autofills have come to lag so severely behind the requirements of the modern user?

One of the main reasons is that most autofills were built on a ‘standards’ approach, in that web developers are supposed to follow certain standards when building forms. However, as we know this rarely happens. The reality is that form field creation doesn’t (and cannot) follow standards. For example, how can we have a ‘standard’ when the Australian and US way of writing the date is different? How can we have standards, when some forms are hidden? How can we have standards, when the requirements of online forms are becoming more complex every day?

Another reason is that most autofills predict a field as simply a field as part of a group, as opposed to groups of fields. What this means is that while your standard autofill may be able to fill out your address, it will get inherently confused it if you have, for example, a different billing and shipping address as many of us do. This means more inaccuracy which has numerous real world consequences. When ordering Christmas gifts for your friends and family, who wants them sent to your home address?

On the surface, autofills work.

But in reality, they don’t.

The very premise on which autofills were built was, unfortunately, never realistic and is becoming less so by the day.

Autofill that sees what the eye sees, and fills like the brain would

What the autofill industry really needs is an autofill that sees form fields like the eye would, and processes and fills information like the brain would. Basically, a service that sees the forms fields as they appear (as opposed to trying to interpret what a form is based on complex code), and fills them, understanding the complexities of different formats, presentations, and form field types (like billing and shipping addresses). A human-like autofill, if you will.  

Fortunately, one new autofill-as-a-service offering, Fillr, has come very far in making this reality.

Doubling accuracy

Fillr has literally created a more human autofill by building unique algorithms into their autofill-as-a-service that scans web forms like the human eye would. This is enabled by what Fillr calls a ‘middle layer’ in the autofilling process – basically a processing engine that collects page data and deconstructs it. This middle layer, which contains Advanced Textual Heuristics,™  is powered by a proprietary mapping engine, algorithms and machine learning technology.

Fillr’s middle layer works by harvesting as much textual context and metadata from the form fields page, and filling user profile values into a HTML form field. By doing so, Fillr is able to read the page in a similar way to what a human would, as they’re able to determine contextual information, such as the difference between billing and shipping addresses, along with country, industry and data transaction type. What’s more, Fillr is as quick (or possibly even quicker) than a human in its reading capability. The Fillr middle layer can detect, collect and analyse metadata on dozens of fields in less than half a second!

Fillr’s technology also has the ability to fix form inaccuracies and variations. By using an internal scoring system, Fillr can determine which attributes best describe a field, and decode incorrectly named fields or typos. In addition to this, it can interpret form differences in different locales; a feature which has been enabled with the help of language experts. So Fillr is truly gathering and understanding context in a way that no other program could.

And the result? Fillr autofills forms with an incredible 97% accuracy – nearly double that of the average autofill.

Harnessing the power of CPU processing power

Not only is Fillr’s autofill-as-a-service intelligent, it is in fact continually in the process of becoming more intelligent, through machine learning which is enabled via cloud-based CPU processing.

Unlike other autofills that use around 200 text pattern matches to generate form mapping predictions (for example Chromium, the project on which the Chrome autofill is based), Fillr evaluates over 1,500, just for English. What this means is that Fillr's technology has significantly more chance of understanding a form and being able to fill it with the right information.

None of this would be possible without cloud-based CPU processing, which Fillr takes full advantage of. Basically, the ability to read and understand a form accurately takes a lot of brain power – that’s for a human – and similarly it takes a lot of machine power. Mobile devices are simply unable to process this vast amount of information at speed.

What’s more, Fillr’s technology is constantly ‘learning.’ What this means is that as forms and web development updates, so too does Fillr’s autofill. In effect, the algorithms are constantly improving and enhancing their intelligence, another great feature enabled by cloud-based CPU processing.

Data security

Fillr may have come along way with autofilling accuracy, but have they solved autofill’s security problem?

It certainly appears so.

Unlike other autofills where data is stored in the browser, with Fillr, your data is only stored on your device and then encrypted using military-grade 256-bit AES encryption. By keeping it off browsers and out of the cloud, it is far more protected than it would be for other autofills.

What’s more, Fillr never actually sees your data – Fillr’s tech provides instructions on what each form field is looking for, and then assists with putting it into the right form fields. Your data is also stored behind a PIN this PIN isn’t stored anywhere except for your head, so if you forget it, Fillr is unable to help you, you’ll need to reinstall Fillr and set it up again. This may seem inconvenient, but would you prefer the alternative – someone stealing all of your most personal data? These types of measures are simply what is required to keep your data safe in the era of widespread cyber crime.

Where to from here?

Just like other technologies that have come and gone, autofill ‘as we know it’ may be fundamentally flawed. The reality is that the autofill is infinitely harder than it looks, and the issues that exist, for example, programmer errors (such as misnamed fields), ignored HTML standards, constantly changing forms, multiple languages and local data nuances will not go away – in fact, they’ll only get worse over time.

And if they don’t, which they won’t, let’s be honest – who wants an inaccurate and non-secure service?

Yet, autofill as Fillr knows it certainly seems like it will stand the test of time. And in coming years, who knows?

Experts may very well claim that your forms are either autofilled, or they’re as good as dead.

Fillr has developed ‘Autofill as a Service’, the world’s most intelligent and accurate autofill that seamlessly integrates into your social or shopping app.  Contact us today to find out how our technology can help your customers to transact faster and more effectively across millions of merchants sites, boosting your conversions and revenue.