Data analysis: Staying in or going out?
Data analysis is now a vital business function but deciding whether to do it in-house of externally is a dilemma facing many marketers.
Above: Pizza chain Domino’s says sales data is its most valuable data
“Even in a small company you should have some means of understanding the potential of the data you’re working with in-house,” says Gemma Carver, group marketing director of online restaurant reservations provider Livebookings. “Any business worth its salt should be trying to develop some sort of internal data analysis capability, no matter how modest.”
Her view addresses a conundrum now facing more marketers every day, as data, the ‘new oil’, becomes the fuel for almost every business: should you prioritise investment on in-house skills and technology, or on identifying external specialists to add value where internal resources cannot? Many brands are still working out how to strike the balance.
Legal & General Investments (L&GI) has made a significant change in the management of its marketing data recently. The company used to task its media agencies with providing a weekly report detailing channel sales, but digital manager Shaun Ashdown says one of the big frustrations of working in this way was lots of ‘double counting’. All the agencies “seemed to claim every single sale”, he argues.
L&GI also wanted to migrate to a more sophisticated model of sales attribution, away from ‘last click wins’, and it saw an opportunity to employ a new system and gain more ownership of its data. It now works with Qubit, which provides it with behavioural modelling data, which the company subsequently gives to its agencies.
“It means we now own the data instead of agencies owning it, and we tell them about their performance,” says Ashdown. “There is no hiding behind anything.”
The shift in ownership has also made the company more productive. Its peak season is a six-week period at the end of the tax year, during which it makes about 70 per cent of its annual sales.
“We have very concentrated sales meetings during that time - before, during and after,” says Ashdown. “These used to involve people going through all the numbers already provided to us, but not actually discussing what was different, what was changing.” By giving the data to the agencies – not the other way round – L&GI has ensured these meetings now focus on insights and actions.
“We’re not allowed to talk about numbers. That was a big shift this year and it worked very well.” It means the agencies, together with the in-house marketers, can discuss trends that have effected the industry - such as cash ISA rates - and how these have affected the future search terms used, as well as how this informs activity.
While deep data analysis demands specific external skills in order to present the data in a useful format for the company to interrogate, Ashdown acknowledges that having a management information (MI) analyst in-house is still critical in order to query the data and apply specific product information to gain useful insights.
“You have to apply the business intelligence on top of [externally produced] data. There is no point giving that to an agency - they know a fair amount about our business, but they don’t have the bigger picture of what we’re selling or our product strategy, and that is where the MI forms a critical link.
“Our analyst can look at the data and say ‘the FTSE dropped then so that’s why that happened’. He pulls the data together to give to the agency and they put it in their system and link it up. That is when it gets quite complicated and labour intensive, and you want an agency doing that bit.”
Wherever business data generated and subsequently analysed, applying specific business intelligence to findings is of course as important as capturing, merging and segmenting data. Pizza chain Domino’s says the most valuable data to the company is its sales data.
“That tells us how were doing, what health the business is in and whether we’re moving in the right or wrong direction,” says Nick Dutch, head of digital marketing. The majority of this information is created through the company’s own systems – and then merged with external data sources – meaning much of its commercial and customer performance analysis is in-house.
One reason Dutch gives for working this way is company knowledge. “We would like to think we know our business better than most so we’re best placed to mine our data and interrogate it for insights.” He believes that having the data in-house also enables the company to move faster and react quicker than it would if working with external suppliers. It is also a pragmatic decision. “We generate about 90 per cent of the data in-house so it’s better to bring that final 10 per cent in than it is to outsource 90 per cent.”
Although Domino’s ’day to day’ data processing is better done internally, the company does use agencies for in-depth projects. As Dutch says, “We would generally work with our agency teams on more complex pieces of analysis that we’re not resourced specifically to deal with in-house or where a large proportion of the data being used sits outside of our data environment; it’s what best drives most efficiencies in cost and time.”
This last point is one that Carver at Livebookings also reiterates: “We don’t set out saying certain stuff is going to stay in-house and certain bits will be outsourced – it is a question of need and what is most appropriate.”
For Livebookings, data is the lifeblood of the organisation, and the company manages the vast majority of its data in-house. But it has recently outsourced its customer relationship management operation, its most successful marketing activity, to an external platform provided by eBay Enterprise - not that Carver likes the term ‘outsourced’, believing it sounds like it is putting its CRM at arm’s length.
She is clear that the company is not distancing itself from the data in any way. “On the contrary, I think working with the right partner will bring you closer to the data because it will give you the capability to work with it in a very dynamic way.” The company has made the decision in order to give itself better email capability, as well as being able to tap into eBay Enterprise’s experience of big data. “We have a lot of capability in-house but as you start to become more sophisticated, it is about having the right tools that allow you to analyse the data. It is about understanding the balance between drawing on external companies whether for infrastructure, technical or strategic support, and deciding what you want your internal capabilities to be.”
She says trying to replicate eBay Enterprise’s tools in-house would not make sense. “They have 10 or 15 years’ experience of building and refining and iterating on tools that help you analyse data quickly and effectively, so why would we go and build all that ourselves? It would be a distraction for us as a business, but having those tools is critical for running our marketing programme.”
Conversely, classified ad website Gumtree, which is owned by eBay, has decided against using its parent company’s proprietary technology to manage its web analytics data, opting instead to use Google’s BigQuery tool. Gumtree senior business analyst Duncan Mckie explains that it would “take a lot of time” to ensure the website was running it competently and correctly.
”Most of the time would be diverted away from actual analysis and value creation of the data, to just managing the data in the first place. You don’t necessarily get return on investment if you have to do everything yourself.”
Gumtree has therefore decided to take what Mckie describes as a “hybrid” approach: “We have kind of outsourced the web analytics [technology platform] in order to allow us to do more analysis in-house.”
Gumtree was one of the first brands to pilot Google’s latest iteration of BigQuery, which integrates Google’s Analytics product with its data query tool, allowing customers to perform analysis on large data sets.
“We can now understand when a person views a page, replies to an ad or interacts with one of our forums, and then break this down by user to understand the sequence of hits and the journey prior to the conversion across devices – something we weren’t able to do previously,” says Mckie.
He adds that using the analytics platform has allowed the company to explore its data on a more granular level – giving rise to potentially valuable customer insights – without a huge up-front investment or expert knowledge. “It gives easy access to data without any special skills beyond standard data analysis.”
Konica Minolta has taken a similar approach, bringing in specialist technology to afford it greater control over its social media and news coverage data in-house. The company works with online media monitoring specialist Meltwater to distribute its news and keep track of coverage, competitor activity and market trends.
“Instead of waiting for [return on investment] reports from agencies or catch-up meetings and conference calls, we can log-in and see any coverage and news in real-time,” says Ben Leeson, marketing communications manager at Konica Minolta. “We use one product to measure everything, which allows our campaign monitoring to be much stronger too. It’s a more consistent and consolidated approach.”
Leeson believes the benefits of doing it this way are huge, offering control without the need for technical know-how. “We have full ownership over the entire process – from content construction right through to distribution and tracking. In terms of report generation, it is a lot easier because the data is there and it enables us to monitor the success of our campaigns easily.”
The role of external agencies, consultancies and technology providers in marketing data is as necessary as it has ever been – perhaps even more so, given that businesses are processing more data every year – but it is clear that marketers also need greater capabilities in-house than before. This is far from being a zero sum game, and investment will be necessary both internally and externally if brands are to realise the full value of their business data.
Case study: Powering publicity
In July 2013, Paddy Power generated over 1,000 pieces of global coverage within two days following its activity around the birth of Prince George, when it speculated on the odds of the sex as well as the name of the baby. The bookmaker received a similar amount of worldwide publicity – both online and offline – following the re-election of Barack Obama as US president and the naming of the new Pope.
“There is absolutely no way you could begin to track all that data and get a sense of where your messaging has gone without good technology,” says UK PR manager Rory Scott. The company works with technology provider Meltwater, using its software to monitor the coverage so it can access the data in-house at the touch of a button.
“We can do a quick search and you can pull all the information up, arrange it by country, get instant access to all the relevant links, and you can overlay that with competitor analysis – it makes life much easier.”
The company also uses the system to access daily reports of industry-related news around the UK, as well as competitor analysis. “Stories come through on a daily alert depending on what search parameters we have set up, meaning we can go back to a particular journalist or put out a broad release to capitalise on a story, or react to a story with certain odds,” says Scott.
He gives the example of the announcement that singer Jessie J was leaving BBC talent show The Voice. “We put out a release about Cheryl Cole being 3/1 favourite to replace her in the hot seat and that led to global coverage in minutes, so the ability to see what is going on in the news agenda, track it and react to it quickly is priceless,” claims Scott.
The solution also requires few specialist skills. “The dashboard is easy to use. It tracks sentiment, so if we go in one day and [media coverage] is right down the scale, we can tell at a glance and find out what is going on. It also brings up flashy graphs and word clouds, which make it easy to report back to the business. When I was collating the ‘royal baby’ coverage, I knew exactly how much publicity we had achieved in just two clicks.”
He could also drill down further, identifying where in the world stories had appeared and comparing it to competitor coverage. “We knew our [royal baby coverage] was a 10% increase on our competitors for example, and that we did really well in Australia, Canada, the US and the Commonwealth countries,” he claims.
Scott says working with a specialist is more cost-effective for the brand than trying to build the capability in-house, as well as giving Paddy Power access to specialist expertise. “When we had a particularly bumper crop of royal baby coverage, I could speak to my account manager at Meltwater and ask him to put a report together. He is effectively our analyst, so when we do need that specialist expertise and we want to drill down further, he steps in. I don’t see any need for an in-house analyst just yet.”
Advanced analytics manager
Telefónica O2 Ireland
O2 uses a variety of systems, with data feeds coming from both internal and outsourced systems. Our insights and analytics teams can not function efficiently if they have to access multiple systems each time a request comes in, so we instead created a central data warehouse, which aggregates the data and provides the information for 90 per cent of our analytics requests. We proactively workshop what new data feeds could be worth centralising, considering potential trends and future requirements.
Managing data in-house gives more control and speed. Our people do not tend to rotate as often as in outsourced companies, and this provides us with more continuity and specialist skills.
We tend to do most of our data analytics in-house with highly skilled data mining resources, but we are also experimenting with outsourcing certain tasks for which objectives are well defined and outcomes are clearly measurable.
The benefits of outsourcing can include cost management - having long-term, fixed-cost contracts can provide certainty and security. Also, in terms of capacity, we may simply not have the resources to organise a particular data requirement in-house in a tight time window.
Of course there are also challenges. On the outsourcing side, they include lack of transparency, data security risks, quality control and high costs. If we choose to go in-house, there is a significant challenge in maintaining infrastructure and personnel.
One aspect that isn’t dependent on outsourcing or in-sourcing is measuring campaign effectiveness, where the primary principle should be independence. If the same organisation that is building campaigns is also responsible for their evaluation, it is naturally biased towards positive assessment. Organising accurate and scientifically valid campaign performance measurement is a very challenging task, involving both statistics and finance.
O2 took steps recently to address this, moving responsibility for final assessment of campaign performance from a campaign operations department to an advanced analytics team. This enabled us to assess our campaigns more impartially, and has radically reshaped the way we run below-the-line campaigns.
From this experience, I would argue that campaign performance measurement should stay in-house and be protected from influence, but there is value in skilled external consultants helping to put in place proper processes and best practices.
More available data means more opportunities for businesses to outsmart their competitors, but the success depends on good leadership in both in-house and outsourced scenarios. Plenty of companies try to build predictive models and advanced segmentations, but most projects fail to yield positive results due to a lack of analytical leadership and lack of continuity.
For large and data-rich companies it makes sense to build their own analytical team. For smaller ones, outsourcing may look like a better option. There is a school of thought that bringing in consultants on short contracts is the answer, but this rarely works. If a company does not have an analytically savvy leader when outsourcing data analytics, it is planning for failure.
Maciej Wasiak will speak at Crunch on 10 October, part of the Festival of Marketing organised by Marketing Week and Econsultancy.