Data in the new normal

We explore why data-driven marketing needs a rethink as brands start to recover from the COVID-19 crisis

Data enables marketers to better understand the true state of the world within which they operate. But what do you do when the data being used to make those predictions no longer describes reality?

The COVID-19 crisis has introduced a chaotic external factor element into data analysis and planning, blowing accepted norms of customer behaviour out of the water and replacing them with volatility, uncertainly, complexity and ambiguity. Regular expectations of consumer confidence and buying behaviour no longer apply, nor do assumptions on patterns of human movement.

And with the situation changing every day, any plan may not hold up for long once it makes contact with the real world. It’s a lesson Victorian retailers and hospitality companies in particular are coming to learn very quickly.

COVID-19 has left many marketers and data modellers scrambling to both better understand the world as it now is and to find more accurate representations and proxies for modelling how it might be.

The change has been felt acutely in the retail sector. At multifaceted online retailer, Catch Group, chief marketing officer, Ryan Gracie, says the COVID-19 lockdown delivered a massive boost in new customers and increased activity from existing ones.

“Most of the difference in the data is more for the buying team around the products we should be buying more of for the future,” he tells CMO. “As there has been a significant channel shift to online, we’ve had to fill the warehouse with relevant product.

“The products we were showing were definitely geared around the search terms we were seeing them come in from, and the propensity for stuff to sell. So fashion took a dip, and pantry grew. Sportswear and home gyms – those items that people needed to stay fit and healthy at home - all skyrocketed.”

Changing consumer behaviour has also led to redrawing the baseline for daily product sales and forecasts for long-term growth to avoid both over-buying and under-buying.

“The feeling is there is definitely a new baseline,” Gracie says. “We are doing Christmas figures every day as the norm now.”

The changes in behaviour have also been clearly noted at online marketplace, Gumtree Australia. Head of marketing, Amanda Behre, says the onset of COVID-19 led to a massive spike in home office and fitness categories.

“We saw a big increase in people browsing things like sport and fitness equipment, with weights growing by 66 per cent,” she says. “We also saw a lost more people browsing electronics and computers, up about 40 per cent. Desks grew by 218 per cent.”

As the lockdown progressed, behaviour changed quickly.

“We started to see people looking for more things around home improvement and exercise equipment like bicycles and mountain bikes, home décor and DIY tools, and even musical instruments and games and electronics,” she says. “As restrictions have started to ease, we have seen more people browsing things like caravans.”

As most states come out of lockdown, Behre is seeing some activity bounce back to normal, with spikes in specific categories such as wedding and party services, which are up by 38 per cent.

Data insight: Optimise versus observe

Knowing what customers will want today and into the future is a question ThoughtWorks’ principle consultant, David Colls, has been pondering on behalf of his clients. He says a different approach to data gathering might be essential now.

“The old patterns of consumption aren’t really relevant for making decisions anymore,” Colls says. “We have been operating in a model of optimising with data for some time. But at a high level, it is about going back to a more exploratory mode with data, to understand what new signals might be relevant in the current situation.

“That means stepping back from a world of clearly defined KPIs and objectives and metrics and an established decision-making process on that data, and looking at the data with fresh eyes, or looking for fresh sources of data.”

That’s leading many clients to begin looking at what other sources of data they might already own that they aren’t utilising effectively. Colls says that often leads to their contact centre, and its vast troves of unstructured customer data.

“That is quite unstructured natural language, so they are asking how do we get more insight out of this natural language data,” he says.

The new world is also placing greater emphasis on domain experience to ensure models being produced reflect reality, and to help determine where the inputs might need to be tuned.

“The data science hasn’t changed, but we always advocate that for any model that exists, you have to have the business and category context in decision-making,” says Dentsu Analytics executive director, John Price. “What we have seen in the last couple of months is we have really had to dial out outcomes of the model with what clients are seeing. And we would not advocate someone rely solely on the data science until they have input their business and category context.”

That also means at times turning back to tried and tested methods that lie outside of the data domain – such as talking to people.

Cameron O'Riordan is director of sales and marketing at Stack Sports, a business that helps sporting clubs and associations make better decisions using data. While his company thrives on data, it is the human connections helping him navigate the path ahead.

“It is really about speaking to people in the industry, speaking to colleagues, and even our competitors, to see what they are seeing,” O’Riordan says. “And we are seeing the community come together more to share information to help each other out.

“We’ve pivoted to online webinars and online education sessions to provide content to our customers and potential customers. We recently ran a three-hour online session with 350 people. And I know that of those 350 people, 70 per cent aren’t customers, so we now have a source of lead generation and key data we can use strategically to keep conversations going and potentially generate new leads.”

Up next: The how and where of data-driven decision making in the new normal, bolstering agility, plus tapping new data sets

Page Break

Data how and where

But while COVID-19 has clearly wrought significant changes, there are many behaviours proving more resilient to the unprecedented circumstances. According to general manager for data quality and targeting at Experian in A/NZ, Steve Philpotts, many data points remain constant.

“The key to understanding the extent of the impact of a change in data is to understand what data is used to make decisions. For instance, is it stable or volatile, and to what extent does your business model rely on human behaviour?” he asks.

“If we take clothing retail as an example: People will still have the value of taking pride in their appearance. A children's clothing retailer is still going to target parents with children, female fashion labels are still going to resonate with females of certain age groups. The data around who their customers are remains fundamentally unchanged.”

What businesses do need to take into consideration now is the ‘how’ and ‘where’.

“Such as, how will people shop in a 'new normal', or where should our products be located - a boutique holiday location, CBD location, or in suburban shopping centre,” Philpotts says.

That places huge emphasis on data quality, and the need to analyse what data might be incorrect or missing. It means validating contact records, incorporating data collection as an engagement strategy, and leveraging third-party data providers to back fill information.

“With the latter, this will incur a cost but will quickly fill the gaps and any cost should be outweighed by a strong customer return,” Philpotts says. “The key here is to make decisions that will future proof your database.

“If you are filling gaps in behavioural or attitudinal data due to COVID-19, it is important to note there hasn't been enough time to fully assess if these behaviours have been embedded. Also, the collection of data may still be changing and could be problematic, as you could amplify one-off changes, which could lead to incorrect decision making in the future.”

Philpotts says one lesson emerging quickly is the need for organisations to have the agility to work with data to reach decisions quickly, and be able to work with data from multiple systems, or of unknown accuracy. That, in turn, means models need to be sufficiently robust and flexible to handle these new requirements.

“Driving cultures of innovation and learning are also great strategies to test and learn, in turn providing a more holistic view of how to engage with customers,” Phillpotts says. “Ultimately, the successful organisations that come out of this environment are the ones that have the agility, and the ability, to drive authentic engagement with their customers to foster stronger relationships.”

Agility in data-driven decision making

That concept of agility has also become more important for director of strategic consulting for APAC at data technology company Silverbullet, Tim Beveridge.

“A fast-moving situation requires an approach to data science that can give you a fast-moving read on those signals,” he says. “Most of the modelling done in marketing is on the assumption things are slow moving and predictable and follow patterns. Rather than using the approach used in the past and try to make that work, I would be getting my data science team to explore other methods for using faster moving signals.”

One very obvious area for change has been in out-of-home advertising, especially that connected to public transport.

“That public transport signal would add more value whereas previously it would have flatlined,” Beveridge says. “It is about resetting how those models are built, resetting what feeds into those models, and building more than one model to figure out the best way of moving forward.

“Some of the variables in models marketers are using that would have been excluded for not being very helpful should now be included. Our recommendation is to go back to your assumptions and your models with your data scientists to rerun things to see if those models describe the reality of the world in a different way.”

Organisations that have geared themselves towards agility are going to be able to recut models a lot faster, implement and operationalise those models more easily, and communicate about those models a lot better, Beveridge continues. “Agility is a key facet to all of this.”

So while the world might be different, and the data it generates along with it, the fundamentals of good modelling and analytics should hold true. For the team at Dentsu Analytics, what is required now is exactly what good modellers do all the time.

“We are always trying to account for the sensitivity in the market to any given variable,” Price says. “The conversation we are having with clients is that this is business as usual for your model, which means if you are doing it right, you are continuously updating those models so you always have the latest data feeding into that.

“Models are built by looking at a longer history over time, understanding all those sensitivities, and doing the best to predict what happens if demand drops or stimulus increases. For most of the brands we are working with, there is the knowledge that their category hasn’t disappeared, but consumers are demonstrating a heightened sensitivity to what has happened through COVID-19.

“So if you think about cars and energy and insurance, there is a tightening of the wallet, so from a modelling point of view, we need to update the models to better understand that sensitivity.”

Check out more of CMO's features into how brands navigate the COVID-19 crisis:

Follow CMO on Twitter: @CMOAustralia, take part in the CMO conversation on LinkedIn: CMO ANZ, follow our regular updates via CMO Australia's Linkedin company page, or join us on Facebook: