More and more our clients tell us that identifying insights is a key role for consumer affairs. It is exciting to be asked to represent the voice of the consumer but it can also be a substantial responsibility – you want to be sure you don’t “miss anything” and you want to be very confident in your findings. Pointing out ‘opportunities’ owned by other groups can be a controversial exercise!
So we see two challenges to overcome:
- Identifying insights with limited resources
- Having objective ‘proof’ that your observations are ‘real’
I want to share a few approaches to this problem that we have converged on as useful. These three ideas have come from our work with brands and we have been told they are helpful. (I would be remiss if I didn’t note that all of these are features of our INCITE analytics software, which we would be happy to discuss! You can also do all of this in Excel if you are so inclined.) Collectively these ideas can help take your insight generation from 'panning for gold' - a manual, slow process dependant on luck to a more systematic, automated infomation factory!
1. Make the Math do the Work!
If you work with more than one brand, and have segmented consumer responses into hundreds of reason codes you have a huge number of potential combinations of issue, product, and even location that may be of interest every month (or week, or day…)! The simplest approach to offering actionable insight to your business is to be asked specifically for input on a particular issue or decision.
But that is not always the case – it is often up to consumer care to identify relevant topics to offer perspective on. How do you sort through all of the possibilities to hone in on the items that may be meaningful?
One way is simply to send data and report on the ‘top 10’ – this is fairly easy (and safe) but likely will miss many potentially interesting issues – the top 10 is often fairly static. We have found that a great way to quickly identify shareable trends in consumer response is to focus in on what has changed the most! To do this we recommend using some fairly simple statistics – specifically z-score. (No need to do the actual math – luckily we have computers!)
There are three key statistical building blocks to this approach:
- Average or mean – the central point in a set of data
- Standard Deviation – a measure of how much the data varies from the mean, the larger the standard deviation the more the data varies
- Z-Score – the number of standard deviations a particular data point is above or below the mean
Calculating the z-score of the current period (day, week, month) for a product and/or subject compared to the prior twelve months gives you a quick way to find the items that are most different from the norm. This helps you zoom in on items that may be well below the ‘top 10’ but have changed substantially up or down.
It also helps give you objective support when you report issues that may cause heartburn in other groups. Saying ‘your promotion complaints are 3 standard deviations above the prior year average this month’ deflects unhappiness towards the math more than ‘looks like you had some promotion problems last month’ – the statistics are pointing out an issue, not you!
2. ‘The New Normal’
Consumer affairs has deep ties to the quality and manufacturing functions. Leveraging consumer feedback to continually improve products and processes is fundamental to the value of our function!
As such many of us have adopted ‘complaints per million’ as a key measure and one of the most important tools we use to frame consumer comments. Built into this measure is the assumption that there is a relationship between the number of units consumed (or manufactured, or shipped, or purchased – they all point the same way!) and the number of consumers who will contact a brand on a given issue. If we sell/make/ship more or fewer units we should get a corresponding increase/decrease in contacts – normalizing those contacts by the product volume gives us a stable base to isolate ‘real’ issues by changes in this normalized measure.
There are however a few challenges with this approach, including:
- Matching shipment/production/sales data to contacts can require a lot of effort (and sometimes the data has errors)
- The ‘lag’ between production/shipment/sales data may not reflect reality (what is the real lag time from plant to use? is it the same for all of your brands?)
- Many topics simply do not fit this type of analysis (Promotion complaints per million units produced?)
We have found that normalizing by proportion is an effective way to identify trends that are not simply due to an increase or decrease in overall contact volume. This approach looks at the mix of contacts in a period (a day, a week, a month) and compares it to the mix over the prior year. You can then identify changes in what consumers are saying in a ‘normalized’ way that does not depend on any other data – such as “over the past year advertising complaints have been 2.4% of our total contact volume, this month they were 0.5%.” Clients that use this approach have found that they identify the same issues when normalizing by proportion that show up when looking at parts per million, with less effort.
And when you combine proportion and z-score you get a powerful way to quickly identify the short list of meaningful changes that are worth sharing or investigating!
3. Lessons from Quality
We have talked a bit about the quality / manufacturing heritage that has shaped many consumer relations departments. In our work with clients and their quality partners we have some terrific methodologies that consumer care teams can consider adopting in the process of identifying consumer insights.
Specifically, the discipline of statistical process control (which is often adopted as a version of Six Sigma) has a lot of great perspective on looking at data. One of the core tenants of the approach is to identify and eliminate meaningful variability, so there are proven ideas as to what constitutes a ‘meaningful’ change that should be investigated.
Our clients have pointed us in the direction of control charts and ‘special cause tests’ as tools to identify changes in consumer comments. In many ways this is the most complex of the three approaches but you likely have allies in your quality and manufacturing teams that can help you adopt these practices.
Bringing it to life.
All of these methods can be leveraged in Excel at only the cost of the effort to set them up and manipulate the data (some of the above links can get you started). We have also built the capability into our INCITE analytics tool (Yes, I think that is a better approach - if you would like to see it please give me a shout! email@example.com). No matter what route you take, we think that each of these three tools is worth considering for consumer care teams who are looking to grow their consumer insight identification in 2018 and beyond!