by James Lawson, consulting editor
If today’s B2B marketers genuinely aspire to true insight and turning those aspirations into workable realities, it takes more than simply splashing out on a new analytics system
Simon Lawrence | Founder & CEO, Uncommon Knowledge
Nigel Magson | Managing Director, Adroit Data & Insight
Jon Clarke | CEO, Cyance
Adam Erbert | Business Development Director, Market Location
Applying data-driven insight helps B2B marketers learn from past customer and prospect behaviour to improve today’s B2B campaigns. That might involve profiling current customers before finding lookalike cold targets or building sophisticated multivariate predictive models to inform targeting, timing and creative. But simply buying a new analytics system isn’t enough.
“Most of the companies we speak to are much richer in data than they were ten years ago, but much poorer in information or insight,” says Simon Lawrence, Founder & CEO of consultancy Uncommon Knowledge. “Despite lots of talk and interest in advanced analytics, it still remains an aspiration rather than reality.”
If marketers aspire to true insight, first they need to put the foundations in place. That means going back to data basics to build a clean, well-ordered view of customers and prospects across channels. The challenges are familiar: siloed source systems with incomplete, inconsistent and out-of-date records.
“We often have to do a lot of pure data engineering work first,” says Nigel Magson, Managing Director at Adroit Data & Insight, citing the Supply Point Identification Numbers (SPIDs) used by the water industry.
“You have to link to the operational business address of the water meter to the business’s billing point or HQ. Organisations are often unaware that multiple supply points are all in fact one customer and so misunderstand the value they represent.”
Forging bespoke linkages to construct a practical operational view of parent and subsidiary companies is a common task for B2B experts. The trick is to paint a full picture of each customer and prospect while avoiding unnecessary detail in the database.
“There’s no point in linking to Tata Group over in Mumbai if you want to target Jaguar Land Rover,” Magson says. “We automate heavily but very often it does come back to bespoke manual research and eyeballing.”
One current Adroit database supports the British Medical Journal’s subscription email marketing. A bespoke data model lets it accept vast quantities of international cold lists, such as Ringgold’s incredibly detailed academic taxonomy.
Calor’s database is another Adroit example of how B2B throws up unusual hybrid combinations. Calor might initially sell to a property developer on a building site but, when those properties are built and sold, the database has to be updated with the names and freshly minted addresses of the new consumer owners using sources such as the Royal Mail’s Just Built and Not Yet Built files.
The right questions
Discovering this kind of data complexity means asking the right questions in the first place. At one publishing client, Uncommon Knowledge had to pull data from multiple silos to form a joined-up view of marketing: salesforce, two different marketing automation tools, a research survey platform and two systems for tracking online activity.
“Online activity is usually heavily analysed using tools and dashboards within the various execution and automation platforms used,” says Lawrence. “But these analytics are micro in nature – informing the specific activity – rather than macro, like asking whether the audience is the right one to target in the first place – and whether we’re targeting them with the right product.”
That comes back to the need for a single view, where identifying the best existing customers is the usual first analytical step. RFM scoring is a great way to start, highlighting high value buyers whose profile might be used to select new targets.
“You always start with the customer data,” says Adam Herbert, Business Development Director at Market Location. “Are the client’s customers over-represented in certain sectors? Profiling can quickly indicate opportunities they may have overlooked.”
Running multivariate analysis to identify customer profiles helped Uncommon Knowledge gauge existing penetration and the attractiveness, volume and potential of remaining prospects for its publishing client. The SCV also underpinned further work to understand the customer journey using historical campaign data.
That showed the path prospects took to become customers and which “touches” prompted those journeys. However, Lawrence notes B2B data limitations make it difficult to analyse in detail and then apply those learnings universally.
“We still really only have SIC and company size to play with,” he says. “These don’t provide sufficient insight to explain decision making behaviour or inform targeting. What’s needed is new data or new ways of thinking about businesses and decision making behaviour.”
He points to the popularity of personas as an encouraging sign, now a standard way to inform user experience and content development. However, personas are more anecdotal than based on hard data.
Linking bespoke market research to standard business demographics is a common solution here, giving a wide-ranging, consistent universe that also offers detailed insight into the motivations behind business actions. For his publishing client, Lawrence used the results of several surveys to develop a model of business sector maturity.
Imputing those research results to the SCV helped further segment the prospects based on attributes like procurement maturity: did companies take a “big stick” approach to pushing down suppliers’ prices or form more sophisticated partnerships? That in turn informed the choice and tone of content used in all channels.
Uncommon Knowledge is itself now developing a new segmentation focusing on the owners and directors of small businesses. With echoes of Lawrence’s previous work in this area, it classifies individual values, preferences and prejudices and their influence on their company’s purchase behaviour.
“Using primary data that can be attributed to all businesses,” says Lawrence, “we are in a much stronger position to serve content that is directly relevant and create richer personalisation and engagement.”
Content personalisation, web engagement: those digital applications are where much of B2B insight investment is now focused. Allied to marketing automation technology, bringing together online behaviour with offline demographics is helping target and prioritise prospects and companies.
Lead scoring is a particularly fertile area. By allocating different scores to prospect activities like website visits, content downloads or response to an outbound email, marketers can find out how close a prospect is to conversion and so allocate the type of marketing treatment to employ.
IP and multiple other digital tracking methods are central to lead scoring, helping link email recipients to web visits and linking anonymous website browsers to a recognisable business. As part of the email campaigns Adroit runs for one client, it built a points-based lead scoring system to gauge the implications of subsequent activity by those clicking through.
Adroit weighted visits to different website pages to reflect their engagement importance. Recency and number of site visits also feed into the final metric, along with negative inputs that reflect unwanted behaviour.
“We knit this together with scores derived from other channels,” explains Magson. “If there’s no response from email, then we can switch to a different channel.”
Adroit uses Communigator’s GatorLeads to manage the IP tracking, with its data one of the feeds to a SQL database that gives a full single view of each company and its associated contacts. A selected base from a UK B2B universe is another input, which enhances the client data.
“We’ve built lots of other linkages to digital platforms like Dotmailer and Adestra through their APIs,” says Magson.
Account-Based Marketing vendor Demandbase is a leader in this kind of IP linking, supporting an account-level view of prospects and also helping cut waste in paid search and RTB display ad buying. Knowing who is on a web page means marketers can choose to serve ads only to prospects from companies they are targeting.
However Market Location’s Herbert warns IP tracking and automated scoring methods may not be as effective as software vendors would like us to think. “Lead scoring can be very misleading at times and may not accurately predict conversion,” he says.
That’s fair comment. Though we might know that IBM have checked out our product information, it’s hard to know who the person browsing our website really is and their real intention. Is their activity indicative of what the business they work for really wants – or are they a relatively lowly employee rather than a senior decision maker?
“If a company keeps visiting your website, maybe they are looking to sell you something rather than the other way round,” says Herbert. “IP tracking is absolutely brilliant in consumer marketing where you can track the individual but in B2B it’s not so effective, especially at larger companies. Also, you only have around 800,000 business IPs which is less than a third of the businesses out there.”
Cyance’s Nexus platform is the latest to take this digitally-driven cross-channel approach, linking online tracking to offline activity to build an integrated view of channel activity. Using techniques like machine learning to detect patterns and to gather reliable data, Nexus supports automated predictive model building to prioritise the most likely prospects for contact.
While partnering with established vendors to handle tasks like ETL, Cyance has put a lot of effort into developing Nexus itself albeit with some input from predictive experts Warwick Analytics.
“We looked at various different solutions but wanted to build our own IP,” says Cyance’s CEO Jon Clarke. “It uses a mixture of open source algorithms which we tune to suit our own uses. With automation, you can crunch through vast data sets with little human input but we also overlay decades of experience to provide human context.”
The other vital part of the Nexus solution comes in matching prospects to buying signals identified from the same individual or business elsewhere on the web, again using techniques like IP matching. This kind of background data gathering is another hot area for B2B innovation with the likes of Lattice Engines, Mintigo and Infer attracting a lot of investment over the last few years.
Cyance collects this online data itself and also takes in feeds from multiple data partners, adding it to the hundreds of millions of conventional records on its Global Multiverse database.
“Today, pretty much every buying journey starts online and, increasingly, it ends there too,” says Clarke. “We capture data on web buying behaviour, determine its relevance for our client and what point that person is at in their buying journey.”
This intelligence adds another layer of “behavioural intent” to predictive models. “That combination gives us the reliable and effective scoring logic for model building,” explains Clarke.
That means more accurate ranking of target lists: those that show additional buying signals along with a high propensity to buy based on standard business demographics or previous purchase go to the top. Cyance claims dramatic improvements in targeting based on subsequent conversion rates.
One client used Nexus to rank its outbound telesales list. “That gave a 300% increase in appointment setting success based on cold leads,” says Clarke. “That’s largely because we knew what they were looking at on third party websites. In a further 50-50 AB split test, 80% of appointments came from our modelled data.”
Those figures are compelling but Herbert ends by reminding us that there are always going to be limits to analytics in B2B and that certain sectors, companies or products can defy analysis. Factoring (selling payable invoices to a third party to aid cashflow) is a good example.
“Either they need it or they don’t,” he says. “It’s not related to company, size, age or anything else. Analysis can’t tell you everything – so use it as a guide rather than a rule.”