Black Friday Blues: How to dodge a thankless day's online sales crush
Peaks are not unpredictable. Peaks are NOT unpredictable
Black Friday is the day after Thanksgiving, the day when – having celebrated the Pilgrim Fathers' survival of their first winter – the public celebrate by going shopping.
Indeed, so enthusiastic was their celebration, retailers would see their balances slide into the black.
Increasingly, Black Friday has moved into the digital world – and into other countries (the UK) as sales and marketing people have cottoned on to its potential. Then of course we have the out-and-out online shopping experience that is Cyber Monday.
Inevitably this has caused technology headaches for even large retailers due to massive spikes in sales relative to the days and weeks beforehand.
Black Friday 2014 was arguably the biggest yet to hit Blightly’s retailers – and not everybody’s IT was as prepared as their company’s marketing and sales push.
Websites at Boots, PC World, Argos and even Amazon went down with the inevitable adverse publicity and, more troubling for them, lost business.
So, how can one avoid the disaster of numerous online credit card authorisations giving way to a white screen proclaiming “503 Service Unavailable”?
It's science, not witchcraft
Companies have been having sales spikes for years, and so unless you're a brand new retailer with no history to refer to, you should have some idea of the likely increase in sales during busy periods. For instance, I used to provide consulting for a small travel company that averaged a handful of bookings a day during December, and then hit 1,200 on the first day back in January.
It was perfectly normal, and it took it in its stride – and planned for it. It knew the pattern, and though it was never possible to predict the precise skew of bookings it knew enough to cope.
Granted, there are sometimes bigger-than-normal spikes. But the average company that does its homework and analyses both its own historical figures and the trends around the industry will be able to predict things with a decent level of confidence.
Think end to end
The biggest mistake in any system is to fail to consider all of the individual components of the system, the individual steps in the browsing, searching and purchasing process, and the interactions between all of the above.
There's no point having a website that can take 5,000 purchases an hour, for instance, if your merchant services provider can only handle 4,000. And err on the side of caution: if you're predicting 8,000 purchases an hour and your capacity is 10,000, that's simply not a sufficient margin of error – you should be looking to cope with 15,000 or maybe even 20,000.
End to end means the business, not the systems.
Equally important, though, is that the sales and marketing teams work with the IT teams to ensure that each knows what the other is doing. I wish I had a fiver for each time I'd seen the fulfillment team running around like a bunch of nutters wondering why the effluent had just hit the fan, only to discover that a big (and clearly successful) mailshot had gone out from the marketing team a week earlier.
Is it fair for the IT department to dictate what the sales and marketing teams do?
Of course not. It is, however, incredibly dumb for the company to invite people to come and crash its online trading world. In the offline world you can get away with selling something that you don't necessarily have in stock, or which you haven't made yet; but if you open a shop that can fit in 50 customers and invite a thousand, you'll fall on your face and your goodwill will have evaporated.
Be prepared to compromise, then. You want a thousand customers to come and see you in your 50-person gift shop? Invite them in shifts on different nights, and perhaps hire the pub next door so you can do them 150 at a time over fewer nights.
You want to promote your online store to 100,000 people but you can't handle them all at once? Stage the promotion so that they don't all land at once.
Taking it right to the end of the process, think of the customer service teams as well. If you sell stuff online, people will call your customer service teams about their goods – to arrange returns, perhaps, or to chase a missing delivery.
The customer service call patterns will largely match the sales patterns – so if you sell ten times more stuff on Black Friday than on the previous day, expect to have a similar skew a few days later in the customer service team's call centre.
Utility companies don't send all their bills on the same day, for precisely this reason – so have the right end-to-end discussions to make sure your peak load doesn't go above the limit of what you can cope with.
Put it somewhere sensible
This is probably where I'm meant to start talking about using “the cloud” to absorb the additional load of the busy days so you don't have the waste of running vast amounts of processing power throughout the year. So here goes.
First off, please tell me you're not hosting your website on-premises. If you care about your online presence staying up during times of adversity, then unless your in-house IT and internet setup is unusually robust and resilient it's going to be a whole lot easier to house the online world with a hosting provider. So should you host it in the cloud? Is it viable and safe?
Yes, of course it is. Some big companies to whom security matters hugely are doing precisely that. And not only this, but they're doing so securely and successfully, and most importantly they're doing it robustly.
An excellent read on the subject is Netflix's description of how it uses AWS to host its services – and remember, its survived Amazon outages with no service interruption.