Upgrading your ecommerce website can be a dangerous move. It’s not just the appearance of the website that changes but also the underlying technology that might connect to your accounts, stock-control, EPOS, supply chains and, most importantly, to the search engines that help bring buying customers to your website.The easiest way to calculate the value of your SEO footprint is to look at the analytics for the site and calculate the contribution made by SEO to sales. With a mature ecommerce business, a not-uncommon split would be around a third from SEO, a third from paid marketing and a third from other sources such as email newsletters, affiliates or direct traffic. If SEO is around a third of all sales then it’s probably also worth having a more detailed analysis of the sales funnels for a broad mix of sales where a buying customer might bounce between a variety of touching points that include SEO prior to buying. This will increase the importance of SEO to somewhere between a third and a half.A ‘perfect’ migration would take place where every ‘old’ URL maps perfectly to every ‘new’ URL. In practice, this is never going to happen. Sometimes, clients don’t fully realise the importance of SEO migration and skip the boring and labour-intensive task of mapping every URL correctly. It’s not always easy to work out what product from the old site maps to what product on the new site. The web address URL might look completely different with the new site. The categorisation may well have changed and the software that’s automatically creating the product URL could be making lots of changes so that simple string matching is not clear-cut. Sometimes, it can take several minutes to identify what matches to what. With possibly tens of thousands of matches it’s no surprise that people then take short-cuts such as simply directing to a general category or, for things that don’t easily match, to the home page. The only thing worse than that is to not match at all which then creates 404 errors once the site was switched over.To be fair, most web design and internet marketing companies know of this problem and will tell the client that this is something they understand and have a solution for. Unfortunately, in our experience, while they may be aware of the problem, having a good technical solution almost never exists. We’ve come to that conclusion after having to pick up the pieces following several third-party migrations that went horribly wrong.With most digital agencies the solutions offered involve setting up ‘301 redirects’ to just a few of the key landing pages on the site. These 301 redirects might map an old URL for a key page to  the URL of the new landing page.Setting this up for all the main landing pages may take say a few minutes each to check the old page, check the new page, copy the two URLs, paste the results into the 301 redirects file and then implement the redirect. For all the product pages, another approach used by agencies is to simply provide the client with a large spreadsheet and get them to do the mapping manually. Given the huge task, people take shortcuts and simply map to categories. The agency doesn’t mind because they can blame the client. Nobody checks and does detail and the disaster only becomes clear post launch when you get a tsunami of redirected links. After several tries, Google eventually downgrades and eventually deletes the links. In parallel, the site gets a low quality score and business goes sharply backwards. Worse still, the overall low-quality score can then go on to impact your paid advertising. The crash gets worse.

The business owner has the expectation that the shiny new website can only improve business and is horrified when sales collapse anywhere between 10% and 40%.It’s not perfect (yet) and it still needs a bit of manual hand-holding but INDEZ has developed an AI solution that goes a long way to solving the matching problem between old and new web pages. At its core lies an advanced string matching algorithm that looks for similarity measures in combination with splitting URLs into semantic chunks. Added to this is phonetic matching allied with a small library of background knowledge about the structure and categorisation of the data. Early results for two large ecommerce migration projects have looked encouraging with an automated matching accuracy of over 99%. Even when you do this, SEO performance will still dip - usually due to the category changes. If it’s done right then the dip should only be by just a few percent from which the site quickly recovers.As with many areas of ecommerce, INDEZ is expecting AI to provide a growing range of technical solutions that will underpin many of the marginal gains in ecommerce that will take place during 2017.