The world of fundraising does not, by and large, like surprises. When faced with a kink in the road ahead, it likes to be able to plan for it – to know which way to turn before it gets there.
Well, the charity landscape in recent years has changed, with a proliferation of marketing channels used to solicit donations from supporters. As the methods of communication become more and more diverse so have the ways of responding, and while this has presented many new fundraising opportunities, it has also made it much harder to decipher the messages contained in the results – to separate the good stuff from the not so good, and plan accordingly for the future. Surprises still abound, and as the sector is well aware, more uncertainties lie round the corner in the form of legislation.
Multiple marketing channels are hastening the inevitable end of the era of direct measurement from stimulus to response. For example, most charities are seeing an increasing proportion of donations being made online, often without any link to a reason for the donation. While there are ways of helping the donor to provide some useful information (drop-down menus at the point of donation, for example), it’s possible that without a direct link between cause and effect, as much as half of all donations could be coming in from an unknown origin.
How does econometric modelling help?
By using econometric modelling techniques we can begin to understand the relative importance of the various factors that may be simultaneously influencing donations, and so identify the key drivers behind these ‘unknown’ donations.
Econometric modelling can also help to identify the full effect of channels such as TV, where although some responses may be directly attributable, the medium will undoubtedly generate a halo effect acting upon other forms or channels of donation behaviour.
In a nutshell econometrics can help you:
- In a nutshell using econometrics can help to:
- Prove the effectiveness of marketing spend (including halo effects)
- Show impact of spend on donations and ROI
- Assess impact and return for separate marketing channels
- Identify factors outside client’s control (interest rate changes, weather, major events) and factor their impact into response
- Measure the impact of competitor spend
By understanding these issues, it is then possible to begin planning for the future, including:
- Optimising marketing budgets
- Forecasting income based on drivers
- Creating scenario plans
- Identifying and quantifying unusual events
What does econometric modelling do?
Econometrics – or Market Mix Modelling – provides a useful means of identifying and quantifying the effects of the various drivers affecting a key performance indicator (or dependent variable).
The key performance indicator (dependent variable) would usually be sales value (donations) on a weekly or monthly basis, but could equally be sales volume, new sign-ups, web visits, campaigning actions…
The technique uses historical data to understand why sales vary from one period to the next and explains the sales variation in terms of proposed explanatory factors. The aim is to discover both the factors and the weights of those factors which explain as much as possible of historical sales. A good model would expect to explain over 95% of sales variations over the period being looked at.
The process involves trying many different explanatory factors and weights, to find a combination that a) provides a robust statistical explanation of most of the recent donation variations and b) fits marketing common sense (we have seen examples of models that, although statistically robust, suggested that one’s own advertising has a negative impact on sales!).
What goes in?
As we’ve established, the first step is to determine the dependent variable – the factor which we are trying to explain. The second step is working out what might influence that variable.
As with any analysis project, the key to a successful econometrics model is to ensure the robustness and validity of the data going in to it. Great care should be taken in collating all the potential drivers of donations for your charity (own and competitive marketing, seasonality, the weather, movements in economic indicators, etc) and identifying robust and measurable data sources for these factors. Marketing activity is usually straightforward and can be measured by spend or by impact/reach figures. Other variables such as PR and brand awareness might appear trickier to quantify but there are resources out there which can help to apply a measurable dataset to these factors.
In separating the effects of different variables it is essential that these variables follow different patterns or occur at different times. If all marketing activity takes place at the same time every month and to the same amount or degree, then it will be difficult to isolate the relative importance of each individual factor. It is fairly uncommon to see this – most marketing programmes have an innate variation built in to them, but for example if TV ads are on air constantly at the same levels then the model would have no experience of alternative levels to learn from.
A further important element of the project involves building a warehouse of various manipulations of the data which will help model interactions where the relationship is non-linear.
Adstock – this is a common way of treating advertising factors whereby they are given a weight depending on the recency of the advert, thereby allowing us to determine the rate of decay of advertising – people see advertisements and remember them but they are forgotten at a certain rate.
Lagged Variables – for many variables it is appropriate to assume the effect upon actual sales materialises one or more weeks after the actual event.
Logged Variables – this enables the identification of diminishing returns whereby higher levels of any given factor may impact donation levels less than lower levels – for example the incremental benefit of press advertising may start to diminish after a certain point.
Combined Variables – in some cases the presence of two (or more) specific factors is required together to influence donation levels – this allows for such relationships to be identified.
What comes out?
The outputs of your econometric model should reveal the relative contributions of each sales driver over the period of the model; weekly contributions of each sales driver; elasticities and coefficients of each sales driver; base level sales. Examples of typical output are shown below.
The elasticity will indicate the likely percentage change in sales for a 10% change in the driver. For example an elasticity of 0.1 for TV advertising would indicate an additional 1% sales from an additional 10% investment.
The co-efficient shows the unit gain in sales from a unit gain in the driver. So if TV had a coefficient of 44,000 this might indicate that each additional advert would provide an additional £44,000 in sales.
Knowing the value of the co-efficient enables us readily to calculate the ROI or profit from each marketing activity
Another useful output can be to show the base level of donations if all explanatory factors were set at zero – in other words if there was no advertising.
Looking to the future
Once the past performances have been understood, the model factors can be used to optimise marketing spend in the future. Identifying the true ROIs of all your activity will enable to you to increase spend in your best contributors, or to cut or reduce spend where returns are diminishing.
The knowledge you acquire from your model can help you to create scenario plans for different investment levels or different external environmental conditions. Knowing seasonal variations can help you build plan for different times of year, or change marketing spend to try to address any particular low points.
Here’s a straightforward example of how an econometric model can feed into strategic planning. In a recent model built by Wood for Trees for a major art museum, we were able to show that their direct mail expenditure was actually having a 25% higher impact on membership sign-ups than their direct measurement had suggested. This knowledge enabled the museum to increase spend on direct mail to counteract a predicted reduction in walk-in sign-ups as a result of a forecast decline in visitor numbers to the museum.