For online retailers with large product ranges, the inability to create relevant merchandising rules for each customer is costing time, money and sales.
The rules can't cope…
For online merchandisers aiming to personalise experiences for their customers, the approach, to date, has been to create personalisation rules that dictate what a customer might see on a page. In simple terms this might be ‘if this product is displayed, show these other promotions’ or ‘if we know this about the shopper, do that’ and so on.
This is well and good with a limited range of products; a site with a skincare range might suggest cleanser and moisturiser to someone looking at toner - logical product partners that form part of a skincare routine, and easy rules to write.
But when you scale up to a thousand products across different departments, the whole process becomes significantly harder to manage. In fact if becomes unmanageable. Think about it, if a visitor is broadly segmented five ways, that’s five thousand rules to write and maintain. Add to the mix colour ways and sizes and that multiplies yet again.
And there's more
It’s not just the volume of product and options; rules cannot be written for unpredictable buyer journeys. Rules cannot flex to different contexts - a product line that is out of stock, a new bestseller or an emerging trend – without having to be rewritten, written or deactivated, which will then have a knock on effect on other dependent rules.
What’s more, it is clear retailers are not coping. A search for ‘online merchandiser’ vacancies reveals 300 vacancies and ‘website merchandiser’ 466 (Indeed.co.uk). And while they may not all be looking for rule writers, there are some significant brands looking for online merchandisers: MandM Direct, Furniture Village, Dixons Carphone, Amazon, George.com, Fitflop, Asos and Harrods all had vacancies at the beginning of October.
If enough rules cannot be written and enough staff cannot be employed to write those rules, or carry out other online merchandising tasks, it is clear that opportunities are being missed, sales lost - with employees doing the equivalent of draining an ocean with an egg cup.
So if the rules based approach doesn’t work for larger retailers, what can they do?.
Predictive machine-learning - Automated product exposure
This is a challenging concept for many retailers; the in store visual merchandiser employs artistic flair with a degree of science and in some organisations is ordained with demi-god status.
But with the infinite possibilities online brings, conventional wisdom breaks down. The only way to deal with the mammoth task and heavy lifting is through machine learning, described variously as giving “computers the ability to learn without being explicitly programmed [rules]”, “producing reliable, repeatable decisions and results” and able to uncover “hidden insights”.
This has several clear benefits:
- Real-time learning – decisions based on new data can be made instantly with instant outcomes. No need to pore over spread sheets of historical data before writing an out of date rule.
- Fewer, but more efficient staff – rather than doing the ‘grunt work’, online merchandisers can apply their skills and experience to more creative and productive tasks such as visual representation of products and promotions and responding to external needs such as how to shift the 10,000 Sam Allardyce coffee mugs in the warehouse.
- Increased conversion rate and AOV – by exposing more relevant products to all visitors
- Costs reduced – as there is no longer a need to write more and more rules as the business develops.
Time to break away from rules
Predictive machine-learning is being used, successfully, in financial markets, and marketing automation is moving over to using predictive machine-learning. These are trends driven by the same issues facing retailers – masses of information and not enough people or time to act intelligently on it instantly.
With that in mind it is hard to escape the conclusion that now is the time for retailers to stop struggling with rules based personalisation and move to the next phase of online merchandising – predictive machine-learning or automated merchandising.