Here are a couple of possibly unrelated thoughts:

First, over the last year my Chairman and I have been asking groups of people two questions: How many people regularly buy things recommended to them by Amazon? And what sort of recommendation would they trust? Whilst this isn’t empirical research, between us we have spoken to over 1,000 people and this is what our polling has told us:

  • Very few of them regularly bought an Amazon ‘recommended for you’ product.
  • Many more engaged with the ‘other customers also bought’ suggestion ("It’s how I know it needs batteries", said quite a few) .
  • The vast majority were far more likely to find other customer ratings/reviews influential than anything any retailer suggested.

If this is right (and it’s quite a large sample across many different demographic groupings) then investing in ‘often bought with’ and customer reviews/ratings may well help influence customer behaviour. 

Secondly, in a very well argued piece in the FT Tim Harford explains the vast investment made by retailers such as Amazon, Tesco and Target in algorithms that drive their recommendations may well only really be successful through the application of the law of averages. In other words, they have a customer base of such a scale that even a small percentage increase in sales can be worth the cost and effort. If he’s right (and as one of the UK’s leading statisticians, there’s a fair chance he is) then personalised recommendations is possibly a very expensive undertaking with, unless you have significant scale, a slim chance of a sensible payback.

Funnily enough, looking at our clients’ experiences we don’t think these two pieces of data are unrelated, but we do think they can get overlooked in the rush to embrace personalisation as an online sales driver. 

Segmentation vs. Individual Treatment

Where retailers employ an ‘individual history’ approach, virtually all of it is based on automated guesswork informed by a combination of past purchase/account history and/or cookie history (if it’s not wiped – and over 30% of internet users now wipe their history clean regularly). Tim Harford argues that only scale can deliver value from this approach.  Our unempirical research would back that up. Why?  Well it’s your best guess, and in a world of fast-changing tastes and fashions, I may well have moved on from my last purchase, or made a mistake or bought for someone else and not for me or many other reasons that mean all that processing will deliver nothing. And the problem is unless you talk to me at every stage of my purchase journey, you’ll never know enough about the 'why' to make tracking and processing the 'what' really worthwhile. Unless of course you are enormous, so a small increase that comes with the statistical probability that you’ll get it right for a small number of people, pays back the large sums involved.

So, is there any other sort of personalisation that works for the ‘average’ retailer?  A sceptic might argue that much of what is talked about as personalisation, is in reality clever segmentation, as opposed to something tailored to reflect individual needs. They would be right of course and getting this right may well deliver an increase in sales.

The key strategies are:

  • Enable the customer – This is done through navigation filters such that you only offer up what they want and enables customers to look only at products and services that meet their criteria. Simplest, least costly and still likely to be one of the most effective drivers of better sales engagement.
  • Inform the customer – This is done through reporting what other customers with their needs have done and how they rated the outcome.
  • Engage the customer – This is done through early stage engagement that helps them establish the relationship they want with you. Great examples here are the FT’s daily email selections and Waitrose’s ‘customised offers’ proposition. Here the customer is in control and has set the parameters of their interest. It is easier then to segment customers and follow-up with ‘customised’ communications.
  • Follow customer interest – This is easily done with cookies and enables holiday companies to link your searching for hotels in Rome to adding in suggestions for car hire for example. This is automated linkage at a meta level where a trip to Rome is automatically associated with flights, hotels, car-hire, events etc.
  • Respond to customer intent – Once something is in ‘the basket’ help the customer by providing information on ‘what others did’.  This strategy is reputed to have added billions to Amazon’s top line.
  • Relate to the customer – Actually far fewer of us want relationships with companies than we would like to think. We don’t want them as our new BFFs nor do we want them in our inboxes everyday but effective, thoughtful, segmented emails are still one of the most effective ways of growing sales and one way of getting people to agree to hear from you is to let them regulate how often they want to hear from you.

There are some smart things that technology can help you do without spending a fortune:

  • Interest based segmentation - A major publisher in Australia adjusts their subscription offer to reflect the sections of the paper they know are of interest to a user and reflect this back in both imagery and offer.
  • Adjusting services offered at a B2B level to the company type the user represents.
  • Seeing intent on a site without purchase and highlighting that product as a carousel banner when a user returns.
  • Weather based real-time imagery / adjusting offers.
  • Location based segmentation - highlighting local offers / opportunities based on a known IP address of the user.

The elephant traps are:

  • Over-reliance on guessing – The more you are relying on automation to make a judgement on motivations, the more you are risking failure. Chasing customers around the web with automated display ads after they leave your site without buying, doesn’t just waste money, it can irritate or worse ensure you have a detractor for life.
  • Investing in suggesting – Don’t offer what you want to sell as a personal recommendation as it just makes you look silly.
  • Refusing to ‘personalise’ – Just selling what you want to sell in the way that works for you won’t help either.
  • Pestering the customer – Emailing too often with a tone or content that suggests you know what they want is more likely to ensure they get deleted and unsubscribed, than to encourage them to engage.
  • Swallowing the pseudo-psychology – Relationships can only happen between animate, living things and emotional relationships more so. What generates an emotional connection is inter-personal. This is different from an emotional reaction: pleasure, pain, anger etc. which can indeed be generated by a good or bad shopping experience.

Why would people keep shopping online with you? It probably won’t be because you’ve personalised their experience. It will be because you’ve made what they want easy to find and easy to buy at a price point, that is within their acceptable range. It will be because you can get it to them in the timescale they need it and when it arrives it isn’t damaged and it does what you said it would do. It’s as simple as that.