A key theme at the recent Econsultancy Digital Cream roundtable on Personalisation was paralysis: being unsure how to prove the business case to justify investment and start the personalisation journey.
This uncertainty is leading to inertia as digital teams invest in what they know works, such as paid search, rather than take the leap of faith and pursue what they believe will work but don’t have a robust model to validate.
However, there are some simple steps that people can take to test the impact of personalisation before worrying about sophisticated options like using predictive modelling to drive on-site merchandising and geo-personalisation of online advertising.
This blog is a walk-through of what I think is a realistic roadmap for personalisation, starting with the absolute basics (hey basics often work really well so don’t think you’ve got to go all weird science straight away!) and gradually progressing to the sexy wizardry of advanced targeting.
I’m breaking this down to five steps (us consultants like to use numbers to make it look structured and well thought out):
- Early dawn.
- Branching out.
- Future sailor.
Feel free to scoff at my suggestions and pull apart my approach. If you have any suggestions/alternative approaches, please can you explain these by adding a comment to benefit other people reading the blog.
1. Early dawn: laying down the roots of personalisation
The key barrier to adopting personalisation techniques is an inability to justify the investment because there is no existing data model to prove the impact.
So you should start doing things that don’t involve buying in third party software or reengineering the website. Start simple and low cost and learn what works and prove to other people in the business it can have a positive impact.
For example, you can customise your brand value proposition and USP bar to show different content to new and returning visitors.
Free People, a US fashion retailer, does this by tailoring its delivery message for non-US audiences:
Visit intent is likely to vary between these two segments. For example, new visitors are less likely to trust your brand and understand the key benefits to shopping with you.
Perhaps a 365 day returns message is more important to them than a return visitor who already knows this, or a statement about brand heritage to reinforce credibility.
You can use your analytics data to discover which areas of the homepage/site each segment turns to most e.g. which navigation elements do UK visitors click on most? You can use persistent surveys like Qualaroo & Kampyle to overlay qualitative data to help better understand intent and user journey barriers.
Translate this data into hypotheses about which messages each segment is most likely to be influenced by. Then make sure you have tracking in place to measure interaction with these messages e.g. clickable text that can be tracked as an event, or viewed on a clickmap.
Personalising real estate on the home page is often the quick win at an early stage as it can be done relatively simply via the CMS. The Wine Society uses a key content zone under the top navigation to show personalised information to members when they log-in.
It’s a subtle change but demonstrates how you can take small steps.
The good news is that all of this can be tracked and measured easily. You can view page level data within analytics tools if you are creating new webpages, or you can add event tracking to specific content/on-page actions.
You can then segment reports to show data for new vs. returning visitors.
2. Branching out into marketing channels
There is a surprising (and quite frankly disappointing) amount of mass marketing still going on in digital channels, especially via email. The next logical step is to segment the customer set to do some basic personalisation of content & messaging.
For email, personalisation can start at the sign-up stage. Schuh, for example, uses basic segmentation to split visitors into buckets for Men and Women.
From the start, you’ll get email content that is more relevant to you than a generic broadcast trying to cater for both genders. Yes it’s basic but it’s another quick win.
You can then use your CRM programme to gradually add more customer level data (e.g. product interests, average spend, content interests etc.) to help you refine segments and drill down into a greater level of personalisation.
Selfridges sends new email subscribers an invitation to personalise the emails they receive via a preference capture landing page. It’s simple and effective.
And what about social media? Imagine what would happen if you captured Twitter IDs in your database and then ran personalisation campaigns to a select few high value customers.
You could tailor an offer based on the products and content they visit and use Twitter to speak to them personally, not just rely on the next email campaign.
I’ve not seen this in B2C but I’ve personally received targeted tweets from B2B marketers/business owners I follow relating to past interactions e.g. contacting me about a new white paper based on the papers I previously downloaded/enquired about.
3. Awakening: early shoots of automating personalisation
The next stage is to bring automation to the party, which requires development either through bespoke solutions or integrating third party software. In my experience, the cost of building your own algorithm to the level of sophistication of specialist software providers often makes the bespoke route unviable.
A good example is using learning-based algorithms to power on-site search so that search results are personalised based on:
- A visitor’s previous behaviour and the pages/products visited/bought/shared etc.
- The general behaviour of visitors searching on a particular search term – which products were clicked most, converted best etc.
The personalisation of site search goes beyond merchandising (although personalisation and merchandising are closely related), which primarily uses business rules to determine what to show people e.g. we want to prioritise all items that have high stock levels in all variants in search results pages.
The personalised version learns from visitor behaviour, so the results shown are related to their browsing activity and/or the general browsing behaviour of all visitors who searched on that particular search term.
The algorithm is set to learn continuously, so what is shown one week is likely to change the next. This is ideal for factoring in seasonal variations in visitor behaviour e.g. products viewed & bought at Christmas are likely to differ to those during the rest of the year due to the influence of gift purchasing.
There are tools out there that provide this level of personalisation. A good example is SLi Systems, which provides Learning Search and Learning Navigation. Boden is one of its Learning Search customers and you can read a case study of the results.
4. Enlightenment: real-time marketing personalisation
Real-time personalisation helps to customise ad content based on the user, for example using geo-targeting to promote products/services relevant to the user’s location.
Using a core template, the data management platform dynamically creates ads by pulling in the relevant content versions based on the user profile. It’s essentially a large MVT experiment to learn which ad structure and content works best.
This isn’t a distant dream. The technology and data driven companies exist to implement this right now.
For example, Infectious Media has worked with retail brands to create personalised advertising campaigns. It’s worth taking a look at this interesting video on YouTube where Peter Burns, Waitrose Online Marketing Manager, discusses its real-time advertising.
Here’s the low down from Attila Jakab, Client Services Director at Infectious Media:
Advertising can be personalised on many layers, who is targeted, where, at what time, and with what message. For Waitrose we used real-time advertising to target only non-customers who lived within the coverage of retailer’s distribution centres, biasing toward centres with spare capacity.
We delivered advertising at times when consumers were receptive and on sites which were more likely to stimulate engagement. The creative was designed dynamically, pulling in pictures, offers and formats in real-time, based on individual consumer preference.
We also ran advertising for existing Waitrose customers, using data to provide offers that were relevant to individuals as they browsed the web. The adverts were specific to customer experience, for example, echoing offers they had seen in an email; offering money back from a second, third or fourth shop; or encouraging additional purchases on an existing order before a delivery cut off.
Most retailers think you can only increase average order value on site, but marketing off site can have the same effect. If the results are compared to the previous year, average order value and shopper frequency increased dramatically and order amends nearly doubled.
5. Future sailor: seamless personalisation across web & marketing channels
To reach this point, you will need to integrate data sources to help create the single customer view. Consider a B2B company that provides IT software solutions and runs events to promote thought leadership.
If the data sources aren’t aligned, the marketing team won’t know that customer A, who has bought solution B, has also attended event C that was about solution D. They are then likely to miss promoting solution D to the customer in their campaigns because they’re unaware of the interest.
With a central customer view, you can tie-up all data points to create a more detailed view of the customer. For example, a department store like House of Fraser could know that customer X mainly buys from brand Y & Z online but in her local store, she actually buys more from brand A and spends more on average.
Through email marketing you can then use dynamic content to personalise at the customer level with relevant products and even segment the database to split out single channel customers from multichannel and then tailor the template to suit each e.g. multi-channel customer template has a content zone to promote the local store activities and promotions.
This data can be made available to everyone in the business who uses customer data to deliver service. Customer service agents (CSA) can have information surfaced to them when a customer calls in, as soon as they enter their account details.
For example, the customer recently attended a VIP event at the London flagship store for the launch of a new brand and the CSA is able to ask them about their experience and inform them of a new promotion on that brand.
All the data can be surfaced to the CSA at the point of call and use business rules to control what is shown and when e.g. if customer has an active complaint, don’t promote marketing messages.
Where are you on the personalisation evolution scale?
I’d be interested to hear from other people and learn how you or your clients have embraced personalisation and what you have learned along the way.
My personal experience is that many get the basics right but struggle to move on to the next stage, especially tying up the on and off site experience.
More often than not, this is down to data gaps, sometimes created by internal silos, but usually the result of not sitting down and planning data tracking requirements upfront.
OK, over to you now….