But when it comes to the data function’s own dashboard, it is not doing so well. Probably the most-discussed and least-resolved issue for this area of practice is how to demonstrate hard financial measures that result from its activities.
One DataIQ Leaders member who formerly worked at dunnhumby said the chief executive of Tesco put it succinctly: “You’d better know your numbers and they’d better be good.” Pressure to live up to this expectation is growing across the data industry, not least because of the impact on both capex and opex in the light of the Covid-19 crisis.
This may still be tempered by the risk appetite of a given business – in one example, there is a goal to grow fast which allows for more rapid testing and pioneering of data projects ahead of proof points. For the majority, however, there is a regular sense of dread when the chief financial officer (CFO) comes calling to find out what the company is getting for its multi-million pound investment.
The CFO will certainly be aware of the monthly running costs of the data office.
At the top level, adding cash to the business (or saving it on the bottom line) looks like a straightforward metric. For some functions it is – a specific marketing campaign for one product executed across a defined time period will be able to identify incremental sales and revenues, for example, with a good degree of clarity.
When it comes to the role that data may have played within that, things become more complex. To use the simple example of that marketing campaign, a number of models will have been built (channel preference, affinity, propensity to respond) and data pulled from various sources, then checked for permissions, with a personalisation algorithm run across the delivery channel. Each of those activities effectively incurs a cost in terms of human resource and possibly technology expenses (if rented cloud-based services are used). Even if the organisation does not have a culture of cross-charging, the CFO will certainly be aware of the monthly running costs of the data office, for example.
So how would the data office demonstrate what part its input played in generating those incremental sales and revenues? It is not as if marketing will run a control cell of untargeted, unpersonalised messaging – usually test cells look at the impact of a single variable such as price, image, subject line and so on, not good versus bad data.
One starting place identified by multiple members of DataIQ Leaders is to increase the velocity of activity and thereby enhance the revenue stream for the business. An example given was to move from a direct mail loyalty communication with a four-month cycle time into a real-time digital app. As well as savings made on print and production, this data-driven channel closes the loop between contact and action by a customer. There are still many such opportunities within legacy organisations that need the drive of a data champion to be taken.
Risk is a key area in which data can show real impact, bringing more accurate analytical models into play, adjusting customer profiles to bring down costs and increase acceptance rates, for example. It also offers the opportunity to “cry wolf” about fines for non-compliance with regulations, although care must be taken not to overdo this or repeat it too often. As many data offices in large organisations have been incubated inside the risk function as have grown out of marketing.
Data science comes at a high cost and no immediate benefit.
But for some data activities there is no legacy to replace or cost-savings to make because they are innovative and entirely new. Data science is a prime example – it comes at high cost in terms of both human practitioners and supporting technology with no immediate benefits. For that reason, expectations about the level of return it will deliver tend to be high. At the same time, the period over which this is measured can be too short to be realistic.
Tension between data science as a unit and its funding stakeholder is common as a result. The culture which supports start-ups where technology overheads can be very high and revenues non-existent does not seem to have ported across into conventional businesses, even where they are adopting similar solutions.
A further challenge to measuring ROI is which financial metric the data office wants to measure against. What may appear to be a simple measure, such as profit, becomes complex in some sectors, such as financial services, where a sale may not be recognised immediately – for example an insurance policy also comes with the added risk to profitability of an eventual claim.
On top of this, accounting practices need to be applied, such as discounting for net present value, which add a further layer of difficulty, especially if the CDO is not an accredited accountant (which almost none are). Notably, one senior data leader recently attended an INSEAD course on finance for business leaders which offered schooling on exactly these types of practices which are second nature in the finance department.
Data offices will continue to be under pressure as a cost centre.
If these hurdles can be surmounted, the standing of the data office in the organisation will be transformed. As one member put it: “Once you can show you are a revenue generator, conversations are easier.”
Given the challenges which proving ROI presents, it is clear that data offices will continue to be under pressure as a cost centre. Depending on what happens to the economy, some sectors may have to embark on further cost reductions and it will be necessary to make compelling arguments to keep as much of the resource as possible.
There is a third path between revenue generator and cost centre which may be more realistic – operating on a cost-neutral basis. In this approach, costs are acknowledged at the top level and an agreed level of contribution from stakeholder-supported activities is recognised. For the CDO, the key is to balance the portfolio of activities so that significant gains can be used to offset expensive projects. Where major foundational work is being pursued, finding these big wins is critical.
Data is a numbers game. For a long time, however, it has not been able to put its own numbers into play. Advocacy and fear of missing out have fuelled an acceptance that organisations need to pay its table stakes just to keep up with their rivals. With continued economic uncertainty and pressure on revenues, as well as twitchy shareholders and investors, data can not count on this support going forward. But achieving proper accounting of its ROI – that is a different game entirely.