Your company has a Marketing Strategy, right? It’s that set of 102 slides presented by the CMO at the offsite last quarter, immediately after lunch on the second day, the session you may have nodded off in (it’s ok, nobody noticed. Probably). It was the one that talked about customer personas and brand positioning and social buzz, and had that video towards the end that made everybody laugh (and made you wake up with a start).
Your company may also have a data strategy. At the offsite, it was relegated to the end of the third day, after the diversity session and that presentation about patent law. Unfortunately several people had to leave early to catch their flights, so quite a few people missed it. The guy talked about using big data to drive product innovation through continuous improvement, and he may (at the very end, when your bladder was distracting you) have mentioned using data for marketing. But that was something of an afterthought, and was delivered with almost a sneer of disdain, as if using your company’s precious data for the slightly grubby purpose of marketing somehow cheapened it.
Which is a shame, because Marketing is one of the most noble and enlightened ways to use data, delivering a direct kick to the company’s bottom line that is hard to achieve by other means. So when it comes to data, your marketing shouldn’t just grab whatever table scraps it can and be grateful; it should actually drive the data that you produce in the first place. This is why you don’t just need a marketing strategy, or a data strategy: You need a Marketing Data Strategy.
A Marketing Data What?
What even is a Marketing Data Strategy? Is it even a thing? It certainly doesn’t get many hits on Bing, and those hits it does get tend to be about building a data-driven Marketing Strategy (i.e. a marketing strategy that focuses on data-driven activities). But that’s not what a Marketing Data Strategy is, or at least, that’s not my definition, which is:
A Marketing Data Strategy is a strategy for acquiring, managing, enriching and using data for marketing.
The four boldface words are the key here. If you want to make the best use of data for your marketing, you need to be thinking about how you can get hold of the data you need, how you can make it as useful as possible, and how you can use your marketing efforts themselves to generate even more useful data – creating a positive feedback loop and even contributing to the pool of Big Data that your Big Data guy is so excited about turning into an asset for the company.
Building your Marketing Data Strategy
So know that you know why it’s important to have a Marketing Data Strategy, how do you put one together? Everyone loves a list, so here are six steps you can take to build and then start executing on your Marketing Data Strategy.
Step 1: Be clear on your marketing goals and approach
This seems obvious, but it’s a frequently missed step. Having a clear understanding of what you’re trying to achieve with your digital marketing will help you to determine what data you need, and what you need to do with/to it to make it work for you. Ideally, you already have a marketing strategy that captures a lot of this, though the connection between the lofty goals of a marketing strategy (sorry, Marketing MBA people) and the practical data needs to execute the strategy are not always clear.
Here are a few questions you should be asking:
Get new customers, or nurture existing ones? If your primary goal is to attract new customers, you’ll need to think differently about data (for example relying on third-party sources) than if you are looking to deepen your relationship with your existing customers (about whom you presumably have some data already).
What are your goals & success criteria? If you are aiming to drive sales, are you more interested in revenue, or margin? If you’re looking to drive engagement or loyalty, are you interested in active users/customers, or engagement depth (such as frequency of usage)?
Which communications strategies & channels? The environments in which you want to engage your audience make a big difference to your data needs – for example, you may have more data at your disposal to target people using your website compared to social or mobile channels.
Who’s your target audience? What attributes identify the people you’d most like to reach with your marketing? Are they primarily demographic (e.g. gender, age, locale) or behavioral (e.g. frequent users, new users)?
What is your conversion funnel? Can you convert customers entirely online, or do you need to hand over to humans (e.g. in store) at some point? If the latter, you’ll need a way to integrate offline transaction data with your online data.
These questions will not only help you identify the data you’ll need, but also some of the data that you can expect to generate with your marketing.
Step 2: Identify the most important data for your marketing efforts
Once you’re clear on your goals and success criteria, you need to consider what data is going to be needed to help you achieve them, and to measure your success.
The best way to break this down is to consider which events (or activities)you need to capture and then which attributes (or dimensions) you need on those events. But how to pick the events and attributes you need?
Let’s start with the events. If your marketing goals include driving revenue, you will need revenue (sales) events in your data, such as actual purchase amounts. If you are looking to drive adoption, then you might need product activation events. If engagement is your goal, then you will need engagement events – this might be usage of your product, or engagement with your company website or via social channels.
Next up are the attributes. Which data points about your customers do you think would be most useful for targeted marketing? Does your product particularly appeal to men, or women, or people within a certain geography or demographic group?
For example, say you’re an online gambling business. You will have identified that geo/location information is very important (because online gambling is banned in some countries, such as the US). Therefore, good quality location information will be an important attribute of your data sources.
At this step in the process, try not to trip yourself up by second-guessing how easy or difficult it will be to capture a particular event or attribute. That’s what the next step (the data audit) is for.
Step 3: Audit your data sources
Now to the exciting part – a data audit! I’m sure the very term sends shivers of anticipation down your spine. But if you skip this step, you’ll be flying blind, or worse, making costly investments in acquiring data that you already have.
The principle of the data audit is relatively simple – for every dataset you have which describes your audience/customers and their interaction with you, write down whether (and at what kind of quality) they contain the data you need, as identified in the previous step:
- Events (e.g. purchases, engagement)
- Attributes (aka dimensions, e.g. geography, demographics)
- IDs (e.g. cookies, email addresses, customer IDs)
The key to keeping this process from consuming a ton of time and energy is to make sure you’re focusing on the events, attributes and IDs which are going to be useful for your marketing efforts. Documenting datasets in a structured way is notoriously challenging (some of the datasets we have here at Microsoft have hundreds or even thousands of attributes), so keep it simple, especially the first time around – you can always go back and add to your audit knowledge base later on.
The one type of data you probably do want to be fairly inclusive with is ID data. Unless you already have a good idea which ID (or IDs) you are going to use to stitch together your data, you should capture details of any ID data in your datasets. This will be important for the next step.
To get you started on this process, I’ve created a very simple data audit template which you can download here. You’re welcome.
Step 4: Decide on a common ID (or IDs)
This is a crucial step. In order for you to build a rich profile of your users/customers that will enable you to target them effectively with marketing, you need to be able to stitch the various sources of data about them together, and for this you need a common ID.
Unless you’re spectacularly lucky, you won’t be issuing (or logging) a single ID consistently across all touchpoints with your users, especially if you have things like retail stores, where IDing your customers reliably is pretty difficult (well, for the time being, at least). So you’ll need to pick an ID and use this as the basis for a strategy to stitch together data.
When deciding which ID or IDs to use, take into consideration the following attributes:
- The persistence of the ID. You might have a cookie that you set when people come visit your website, but cookie churn ensures that that ID (if it isn’t linked to a login) will change fairly regularly for many of your users, and once it’s gone, it won’t come back.
- The coverage of the ID. You might have a great ID that you capture when people make a purchase, or sign up for online support, but if it only covers a small fraction of your users, it will be of limited use as a foundation for targeted marketing unless you can extend its reach.
- Where the ID shows up. If your ID is present in the channels that you want to use for marketing (such as your own website), you’re in good shape. More likely, you’ll have an ID which has good representation in some channels, but you want to find those users in another channel, where the ID is not present.
- Privacy implications. User email addresses can be a good ID, but if you start transmitting large numbers of email addresses around your organization, you could end up in hot water from a privacy perspective. Likewise other sensitive data like Social Security Numbers or credit card numbers – do not use these as IDs.
- Uniqueness to your organization. If you issue your own ID (e.g. a customer number) that can have benefits in terms of separating your users from lists or extended audiences coming from other providers; though on the other hand, if you use a common ID (like a Facebook login), that can make joining data externally easier later.
Whichever ID you pick, you will need to figure out how you can extend its reach into the datasets where you don’t currently see it. There are a couple of broad strategies for achieving this:
- Look for technical strategies to extend the ID’s reach, such as cookie-matching with a third-party provider like a DMP. This can work well if you’re using multiple digital touchpoints like web and mobile (though mobile is still a challenge across multiple platforms).
- Look for strategies to increase the number of signed-in or persistently identified users across your touchpoints. This requires you to have a good reason to get people to sign up (or sign in with a third-party service like Facebook) in the first place, which is more of a business challenge than a technical one.
As you work through this, make sure you focus on the touchpoints/channels where you most want to be able to deliver targeted messaging – for example, you might decide that you really want to be able to send targeted emails and complement this with messaging on your website. In that case, finding a way to join ID data between those two specific environments should be your first priority.
Step 5: Find out what gaps you really need to fill
Your data audit and decisions around IDs will hopefully have given you some fairly good indications of where you’re weak in your data. For example, you may know that you want to target your marketing according to geography, but have very little geographic data for your users. But before you run off to put a bunch of effort into getting hold of this data, you should try to verify whether a particular event or attribute will actually help you deliver more effective marketing.
The best way to do this is to run some test marketing with a subset of your audience who has a particular attribute or behavior, and compare the results with similar messaging to a group which does not have this attribute (but are as similar in other regards as you can make them). I could write another whole post on this topic of A/B testing, because there is a myriad of ways that you can mess up a test like this and invalidate your results, or I could just recommend you read the work of my illustrious Microsoft colleague, Ronny Kohavi.
If you are able to run a reasonably unbiased bit of test marketing, you will discover whether the datapoint(s) you were interested in actually make a difference to marketing outcomes, and are therefore worth pursuing more of. You can end up in a bit of a chicken-and-egg situation in this regard, because of course you need data in the first place to test its impact, and even if you do have some data, you need to test over a sufficiently large population to be able to draw reliable conclusions. To address this, you could try working with a third-party data provider over a limited portion of your user base, or over a population the provider provides.
Step 6: Fix what you can, patch what you can’t, keep feeding the beast
Once you’ve figured out which data you actually need and the gaps you need to fill, the last part of your Marketing Data Strategy is about tactics to actually get this data. Of course the tactics then represent an ongoing (and never-ending) process to get better and better data about your audience. Here are four approaches you can use to get the data you need:
Measure it. Adding instrumentation to your website, your product, your mobile apps, or other digital touchpoints is (in principal) a straightforward way of getting behavioral events and attributes about your users. In practice, of course, a host of challenges exist, such as actually getting the instrumentation done, getting the signals back to your datacenter, and striking a balance between well-intentioned monitoring of your users and appearing to snoop on them (we know a little bit about the challenges of striking this balance).
Gather it. If you are after explicit user attributes such as age or gender, the best way to get this data is to ask your users for it. But of course, people aren’t just going to give you this information for no reason, and an over-nosy registration or checkout form is a sure-fire way to increase drop-out from your site, which can cost you money (just ask Bryan Eisenberg). So you will need to find clever ways of gathering this data which are linked to concrete benefits for your audience.
Model it. A third way to fill in data gaps is to use data modeling to extrapolate attributes that you have on some of your audience to another part of your audience. You can use predictive or affinity modeling to model an existing attribute (e.g. gender) by using the behavioral attributes of existing users whose gender you know to predict the gender of users you don’t know; or you can use similar techniques to model more abstract attributes, such as affinity for a particular product (based on signals you already have for some of your users who have recently purchased that product). In both cases you need some data to base your models on and a large enough group to make your predictions reasonably accurate. I’ll explore these modeling techniques in another post.
Buy it. If you have money to spend, you can often (not always) buy the data you need. The simplest (and crudest) version of this is old-fashioned list-buying – you buy a standalone list of emails (possibly with some other attributes) and get spamming. The advantage of this method is that you don’t need any data of your own to go down this path; the disadvantages are that it’s a horrible way to do marketing, will deliver very poor response rates, and could even damage your brand if you’re seen as spamming people. The (much) better approach is to look for data brokers that can provide data that you can join to your existing user/customer data (e.g. they have a record for user firstname.lastname@example.org and so do you, so you can join the data together using the email address as a key).
Once you’ve determined which data makes the most difference to your marketing, and have hit upon a strategy (or strategies) to get more of this data, you need to keep feeding the beast. You won’t get all the data you need – whether you’re measuring it, asking for it, or modeling it – right away, so you’ll need to keep going, adjusting your approach as you go and learn about the quality of the data you’re collecting. Hopefully you can reduce your dependency on bought data as you go.
Finally, don’t forget – all this marketing you’re doing (or plan to do) is itself a very valuable source of data about your users. You should make sure you have a means to capture data about the marketing you’re exposing your users to, and how they’re responding to it, because this data is useful not just for refining your marketing as you go along, but can actually be useful other areas of your business such as product development or support. Perhaps you’ll even get your company’s Big Data people to have a bit more begrudging respect for marketing.
Read the original version of this post on my blog here.