Quantcast
Channel: Lost Sales monitoring – Tribe
Viewing all articles
Browse latest Browse all 27

Black Friday 2016 – today’s failures of CX, page errors and lost revenue

$
0
0

see:

What is different in today’s Black Friday 2016 web performance and CX errors, over last year?

The action begins: Black Friday 2016

With a predicted £1.3 billion set to be spent in Black Friday sales and online shopping likely to account for around 70% of purchases, the stakes are high for ecommerce retailers.

Last year we tracked lots of performance issues – the most common of which related to slow performance and timing-out errors – 2015 Black Friday analysis of website failures. What went wrong?  The challenge for most sites is, of course, handling the massive spikes in traffic so that customer experience isn’t negatively impacted.

The point of no return:  Retailers websites in Overload

When retailers are planning for a surge in demand, they have to calculate at which point the site has reached maximum load.  Load testing based on a fully realistic model can predict what that is.

But in planning for the worst: what should a retail store website do when it is overloaded?  There is a need to block further users by either delivering an error message or using a queuing system to prevent the site from slowing to unacceptable levels for all shoppers, (which hits sales conversions badly).

A story we heard today, concerned a data centre that offered their retail client ‘Overload Protection’ :  the retailer was asked to ‘name the number’ at which it should kick in.

But the metric offered was Concurrent Users might seem like the best and simplest measure but because it’s based on cookie usage, it’s a notoriously unreliable metric. A cookie may be allocated for up to an hour, whether or not the user is online, so there’s no way of knowing how many people are actually shopping – and how many have wandered off for a cup of tea.

It certainly helps to establish desired outcomes well before preparing for load testing.

( We’ve blogged about the pitfalls of the wrong metric, when load testing  for peaks:  Climb Every Mountain Or The Importance of Load Testing All The Different Kinds Of Peak

To queue or not to queue?

If you’re expecting an overload at some point, it makes sense to have a plan in place. You could simply throw out an error message but this runs the risk of creating a greater load as the frustrated shopper continually clicks again and again:  trying to force themselves back into the site! A queuing system offers the customer a timed access point which is arguably less frustrating for all.

If you do have to resort to an error message, make sure it kicks in before the site slows to a standstill and create one that’s specifically designed for customers on the day, rather than a generic ‘oops’: for example, QuidCo had problems this morning; but the root cause was probably not due to ‘updating the site’:

Spreading the load

Because many retailers have been offering Black Friday deals in the run-up to this year’s event, it may well be that shopping activity isn’t as concentrated today as it has been in previous years – we’ll see.

The Guardian reported that Currys/PC World had a short outage in the early hours of the morning but that this was quickly rectified. A spokesman for the retailer said that they’d had over half a million visitors to the site between midnight and 6am this morning and that orders were up 40% on the same period last year.

Users experienced some problems with the Argos site – particularly regarding its business logic.

Currys PC World is reporting its highest ever number of orders, up 40% on 2015, with over half a million visitors to the website before 6am alone.

Argos had it appears 500,000 visits in its first hour of trading – though some errors occurred, around the search and filtering: eg a search here supposedly returned 25 results

but only five products were actually displayed.

Some pages were also missing ‘next page’ buttons, making it difficult to advance through the site. It is worth checking data against business logic when load testing.

High street shopping?:  Online was bigger again this year

It does seem that most people are tracking down Black Friday deals online rather than on the high street. Early reports show that stores were relatively calm, with only a few bargain hunters prepared to grab their booty in person.

So this year no  videos of fighting in-store that happened back on Black Friday 2014 in the UK, which also had a range of interesting web failure lessons: Black Friday website crashes: Boots, Tesco, ASOS, Net-a-Porter, TopShop

Update: 13:30 Black Friday 25 November

There is no such thing  as a retailers’ website that does not throw weird errors and glitches now and again:  to some users if not all.

It seems Boots.com had a brief glitch today   – which was attributed to site maintenance , according to Twitter:  not sure if it coincided with some less  than friendly error pages like this one: (as above the errors you show your users do matter) – maybe it was earlier.

This glitch for Barbour was not just a  few users impacted:

Convert Black Friday problems, into client capture !

Dr Martens have been clever today: they had the foresight to use their Error mechanism to capture email addresses.

As we said above, it’s always best to plan for failure, to minimise  the business impact.  But Dr Martens have created a silver lining to this lost revenue  cloud !

Other Brands merely let Twitter keep users informed: e.g. Dell

The post Black Friday 2016 – today’s failures of CX, page errors and lost revenue appeared first on Tribe.


Viewing all articles
Browse latest Browse all 27

Latest Images

Trending Articles





Latest Images