Using the right data for the right purpose
Earlier this month, Nepa had the good fortune of attending the Marketing Analytics and Data Science Conference in San Francisco (April 3-5), an annual rite of passage for marketers who are already sold or hoping to be sold on – as the conference itself boasts – the “superpowers of data science and analytics.”
As a proud benefactor of these superpowers, the tone of the conference was both familiar and validating, especially as Nepa strives to create value for its clients at the intersection of big data and data-drive consumer insights.
The conference was chaired by David Boyle, Executive VP of BBC Worldwide, and featured esteemed speakers from Academia (Harvard, Carnegie Mellon), a variety of verticals (Netflix, Politico, AMEX, Schwan Food Company) and – as San Francisco is among the most popular cities for new businesses – the start-up community (Sam Yagan, Bloomberg Beta).
Nepa, of course, took copious notes throughout MADS17, so it’s our pleasure to now reveal some of the bigger takeaways from the conference:
Prepare and then ask the right questions
- Alistair Croll, of Harvard Business School, spoke eloquently about how marketers once asked a question before they acquired data to (hopefully) answer it. The new answer is not to build a machine prepared for all questions, but to build a data infrastructure by answering all known questions the business has.
- Or, as Bill Franks, CTO of Teradata, succinctly put it in 2012: “To succeed with big data, start small.” Croll also shared research from Microsoft researcher Cormac Herley about the infamous Nigerian prince email scam, which we’re certain no attendees ever fell for. What everyone wants to know is why anyone would fall for it, when it’s proffered with such ridiculously unintelligible language. Sure, if the English were at all coherent, click rates would improve, but as recipients got closer to making a transaction, they would realize the scam and, have wasted the time of the crooks. The implausible story, by design, only attracts the most gullible people, those most willing to see it through for the entire scheme.
The lesson is to never tacitly accept the metrics. We see this in the focus on CPM for ad sales or focus on channel sales at omni-channel retailers, question everything so your efforts are achieving maximum impact.
AI and Big Data are not always correct. In fact, often they’re way off.
- Susan Etlinger, an analyst at Altimeter Group, shared some wisdom from the audience on the topic of algorithmic bias. At its core, this means the inputs into algorithms can contain bias that then have unfortunate results in the output. An example of this comes in the literary world: A database of prose will have predictive models believing there are many more black sheep than white sheep because the former are discussed more in literature, even though they’re significantly rarer in reality. Those living in rural areas know what we’re talking about.
Lesson here: Be mindful of the data sources used and their potential biases. Side note here – with speakers from BBC, Politico, Civis Analytics, and more – there was no lack of chatter about the 2016 election – where there were many biases in data that led to flawed prediction models.
Optimizing without customer experience is dangerous.
- Nepa embraces the power of data science and analytics, but get concerned when the focus is too much about operational efficiency. David Rogier, founder of MasterClass, shared an example from Tesco where the grocery chain was able to optimize a store so efficiently that it could run with four employees per shift compared to the usual 30 plus for the same sized store. They achieved this largely by having only self-checkout options and laying out the store for fastest restocking. The problem was that, over time, there were no customers!
But why? As Nepa has learned using experience data, reducing the cost of labor, lowering inventory and driving more efficient use of space can improve operational performance, but it will take its toll on sales and traffic if drivers of customer experience are not accounted for.
Presentation and education matter.
- Rebecca Haller, Director of Audience Insights at POLITICO, shared some words of wisdom learned from her early writing days: “Journalists don’t ride the bus enough.” It’s a nice way of saying reporters are out of touch with the people they’re serving.
- Josh Hendler, CTO of Purpose, shared that when he was working with field teams he didn’t really need more data analysts, only those who could educate people on the available tools and how to use them.
These two points are very applicable to the research and analytics industry, where insights are organized and curated with headquarters or the c-suite in mind, not the people on the front lines that can act on them.
And, to wrap
I think one of the key themes about data science and marketing analytics was on display in a presentation delivered by Haile Owusu, Chief Data Scientist at Mashable. When it comes to data there are “haves” and “have nots” – i.e. companies that have first party data (retailers, OTT apps, etc.) and those that don’t (CPGs, content creators, etc.). For the “haves” – this can be used to significant competitive advantage (e.g. Facebook, Amazon, Walmart, etc.) – though, unfortunately, many companies do not take advantage of this opportunity (omni-channel retailers come to mind).
For the “have nots” – Haile Owusu reminded us that not all hope is lost – his team uses data fusion and robust analytics – to predict which pieces of content will go viral and where. These insights help them to optimize their investments to support the content with greatest potential. Through advancements in data engineering and data science fields, the data “have nots” are able to bridge data sources and use them for market advantage.
Thanks to Knect365, David Boyle, the presenters, and attendees. “MADS17” was a great conference and we look forward to MADS18. Hope to see you there.
What about Nepa?
If you’re not keen on spending valuable time refining methodologies, sitting behind one-way mirrors, and waiting on crosstabs, all before it’s all analyzed into insights with fleeting impact, then Nepa is someone to speak to.