PCA Predict has long been the standard against which all other UK addressing packages have been judged – and thanks to the latest version of Capture+, the solution looks set to remain at the leading edge of addressing for the foreseeable future.
PCA Predict has long set the pace in UK addressing. First it pioneered pay-per-click, service-based delivery and then introduced the revolutionary Capture+ in 2012. Now we have the latest version of Capture+, will it push the boundaries even further?
Via the easy-to-use, minimalist interface, Capture+ users simply start typing their address into the search box and pick their address from the drop-down list of suggestions. Working in real time, the service updates that list of suggested addresses with every single additional keystroke.
It’s all very simple and intuitive from a user perspective but, just like Google, behind the scenes lies a lot of server horsepower and some very clever coding. Standing out is its machine learning ability, able to apply knowledge of previous search patterns to improve current results.
Capture+ was a hit from the start, with some of the latest users including JD Williams, Domestic & General, Halfords and Late Rooms. We’ve already seen one update, the version that arrived in 2014 which offered a new look-up engine with a vastly faster response time. It also delivered better data entry error handling – a weak point of the initial release.
Described by the company as a complete rewrite from scratch, this latest release deals with input errors in a far more elegant fashion. It also boasts massively improved international data handling.
The error handling builds on features introduced in the 2014 upgrade, such as the ability to deal with out-of-sequence address elements and commonly confused words and text strings. Where before the error handling was limited to a handful of countries with each possible mistake painstakingly hardcoded manually, it’s now fully international and completely automated.
There are five main areas of improvement, starting with spacing errors. Users might enter extras spaces between two words, splitting up normally single words – Parkview instead of Park View – or unintentional spaces within single words: Man chester or Lon don.
Where that could cause wrong or missing results, Capture+ now copes with unlimited spaces by effectively disregarding them altogether. It handles missing characters much better too: Com versus Colm Avenue or Parkview versus Prkview. It also spots accidentally transposed characters – Park/Prak – not dealt with by the previous version which was mostly restricted to spotting duplicate characters, as in “Ffrance”.
All of these capabilities are particularly relevant to the typing errors generated during the mobile searches that make up an increasing proportion of online address look-ups. The vendors describe Capture+ as now “fully fuzzy”, with a mixture of standard algorithms such as edit distance and many other custom ones applied at database level during the matching process.
International address abbreviations are also taken care of. In the UK, the obvious example is St for Street, in France, Bd for Boulevard or in Germany, Str for Strasse. Capture+ extends this to all the countries it covers, able to recognise that Nm is how Czechs often abbreviate Namesti (square).
Spotting genuine spelling mistakes is another huge strength, supported by the service’s machine learning capabilities. By linking common spelling errors to the right address eventually chosen by users and storing them over time, the service is able to correct mistakes in real time. As noted above, the software has always had this machine learning ability, but it was previously used to only prioritise search results.
All these clever new matching and error handling features are impressive but the most striking one is that Capture+ now does all its matching processing in real time; only the space detection function looks up pre-computed answers. Equally striking is that the server power needed to produce these results has dropped by two-thirds due to the ground-up code rewrite and its use of proprietary PCA algorithms.
While we’re on the technical underpinnings, the new engine’s ability to work across distributed server nodes means PCA can make use of cheap and scalable cloud processing power from the likes of Amazon. This way, they can deploy local nodes to maintain performance in any global region and can instantly spin up extra power to cope with surges in seasonal demand, such as pre-Christmas or Black Friday.
The other major feature in this upgrade is that users can now input search terms in their own language and Capture+ will automatically select the country in which to search for results. In the previous version, the user had to click on a flag beside the search box to manually select their home country. Using the flag is still a client-selectable option.
This automatic country recognition is initially based on IP address but, as IP-derived location can sometimes be misleading, the service widens the search when further input data suggests the initial country choice was wrong. Maybe this ecommerce customer is a Frenchman on holiday in Devon buying a present for his parents back at home?
A single engine that drives the whole international match process is what underpins this automatic detection. Previously PCA used an individual search engine for each country, hence the need for the flag to select it.
The international upgrade also extends to reference data. Rather than the 15 countries previously supported in detail, the unified engine covers the full 240 global countries. National postal providers and UPU files are the first choices for reference data but TomTom provides backup coverage where required. The vendor has spent years sourcing the right files and each individual client can add their own custom data if required.
Enhanced tag technology (also proprietary) now makes Capture+ much quicker and easier to install. Clients simply drag and drop to insert the tag within each web page’s HTML code, with full compatibility with tools like Google Tag Manager. With an account set up, the service runs straight away with the defaults but multiple further configuration options are available.
Here, users can alter everything from interface style to security and reference data or force Capture+ to default to a particular country and language. Simply by ticking a box, clients can pay per click to check against extra reference files such as Eircode or the PAF Not Yet Built and Multiple Residence files. Internationally, a client can specify any number of extra files such as Zipcode-based US state tax rates.
As well as bringing the functionality of Capture+ to the pages it’s embedded in, the tag is a gateway to PCA Predict’s many other SaaS products like mobile and email validation. The vendor’s upcoming Triggar customer experience monitoring tool will also use the data these page tags collect to produce insights for applications like conversion optimisation.
This latest upgrade was still in beta as DBM went to press but is scheduled to go into production in late April. A new mobile app is also on the way that lists clients’ current look-up credit balance, along with other site metrics collected via the tag. PCA Predict has already released an update for the Magento plug-in version of Capture+ and further updates for integrations to other tools like Salesforce will appear in the near future.
Also coming down the line is a full rewrite of the Cleanse+ batch service using the same new Capture+ engine. That engine’s fuzzy matching capabilities will support fully international auto field mapping for file import and it will again be able to access cloud processing horsepower to enhance local performance anywhere in the world.
With its mobile-friendly error correction and automatic global address capture, this latest version of Capture+ proves innovation is still alive and well in UK addressing. It keeps PCA Predict right at the leading edge of technology and gives the company the perfect platform for further international expansion.