What went wrong?!
It has been reported by the BBC that Argos have had issues with their new roll out of their redesigned eCommerce site. And for the second most popular eCommerce destination to be effectively down for 2 days this has a huge impact for online revenue and also the Argos brand. So what could have gone wrong with the roll out and what lessons can other online retailers learn from this.
The first item to look at is the time of year that this occurred, the end of October. Like many retailers Argos will have a code freeze in November to ensure that the platform is stable for the pre-Christmas sales and also for the Christmas trading period. Code freezes normally last until after the January sales and it is a good way to ensure that no changes happen that could effect sales. So this major release looks like it had to go in before a code freeze starts. Because of the issues that Argos saw over the last couple of days, it is safe to assume that the deployment was not fully tested and was rushed to put live just so that it can go in before the code freeze.
The second item to look at, is when the issues occurred on production, why did they not roll back to the previous version of the code? With every new release that a retailer makes the roll back strategy has to be the most important part of the deployment. As this provides the insurance that when things go wrong, and with software they will at some point go wrong, that there is a backup plan that allows minimal distribution to sells. The fact that they did not go back suggests that either this was too big a release with too many changes to other components that they could not go back, or they did not have a roll back plan.
The third item to look at, is now that the site is stable and is live, is it performing at the most optimal for a retailer that is number 2 in the IMRG and Experian top 100 list. Looking at Google Page speed, it is currently getting a B grade and for YSlow it is getting a D grade. We would expect them to be getting an A grade on both of these. Also the page load time for the homepage is taking over 5 seconds to load and has just over 1.4MB of data. Both of these are not good for mobile and tablet users. The normal rule of thumb is that if a page takes over 4 seconds to load, the shopper will normally switch off and look at another site. Checking for W3C errors, on the homepage there are 7 errors; W3C errors will effect how the page is rendered on certain browser and can stop the purchase path.
So what can other retailers learn from this:
1. Test, Test and Test; Perform User Acceptance Tests, Load testing and full regression testing before you go-live and if there are any problems that will effect trading, don’t push it live to meet a deadline. In Argos case they should have pushed it back a couple of weeks and have gone live during code freeze. It is a big strong decision to make by a eCommerce Director, but if it saves bad press and more importantly loss of revenue then it would be the best decision to make.
2. Plan every release so that it has a clearly defined and tested roll back strategy. For smaller releases this is easy, however for large releases this requires lots of planning and testing. And when things start to go wrong was a release, don’t be afraid to roll back and try again. Valuable lessons can be leant when a release goes wrong and can be used to improve future releases.
3. If you are releasing something new to your website, make sure that it is the best quality with A grades on both Page Speed and YSlow, with 0 W3C errors and that all pages load in under 4 seconds.
InventCommerce are a Solution Integrator for Hybris, Magento and Demandware, we have a wealth of experience in eCommerce projects, releases and international expansion.