Having implemented the basic use cases for a DMP, most organisations are still falling short of their full potential. For most organisations the biggest value of a DMP is from these basic use cases; excluding known customers from expensive acquisition campaigns, and using 2nd and 3rd party data to extend your reach and increase your relevance. However, there are other strategies which will drive additional value to your business where your DMP will provide a crucial role. This article explores a number of strategies which organisations can implement to drive more value from their DMP.
First - a quick refresher, what is your DMP doing today? Regardless of which DMP you use today, if you've implemented the basic use cases described above, it will be doing these three things:Phase 1) Device Identification and Data ingestion
When a new visitor hits your website or app, the DMP will first try to identify the device. To do this, DMPs look for an existing cookie or device ID to identify who they are. If there is no cookie present, then a new ID will be generated and set in the cookie. If the cookie is present then this is a return visitor, so we can match the behaviour within this current session, to their past activity.
Following this device identification process, the ID will be categories based on predefined rules configured within the DMP. For example, if the visitor browsers to a product category of "jeans", they may be classified with an interest in "jeans". Whilst the visitor continues to browse the site, more data will be collected about the behaviour of the device. At the end of their session, the DMP will have categorised this device ID based on the the visitors activity:
Now that the DMP has identified the device and captured data from its behaviour, you can now create new audiences based on this data. For example, you could create an audience of visitors who have viewed the Jeans category but have not purchased, or visitors who have read a white paper and watched a video but haven't signed up to your next webinar.
So far we have only dealt with 1st party data - data which is collect from your own website. One major use case for a DMP is the ability to use 2nd and 3rd party data - this is data available from your partners or from external data providers who are selling data. Behind the scenes, your DMP will be automatically swapping IDs with 2nd and 3rd party data providers. This gives you access to additional data on your visitors, such as their age, gender, interests, intent to purchase or financial status. This data can be used to enrich your first party data - meaning you re-targeting campaigns can be more granular and targeted. Alternatively this 2nd and 3rd third party data can fuel your acquisition campaigns by targeting devices which have shown an interest or intent to purchase on other websites.
Audiences can also be built to be excluded from campaigns. This is another major use case for a DMP. You can create an audience from devices which have recently purchased your product and service, and then exclude this device from your expensive acquisition campaigns. The savings in ad-spend from this use case alone are often enough to pay for the DMP - several times over!Phase 3) Syndicate audiences to third parties
Finally the DMP will share your audience to third party DSPs to buy ad inventory when a member of your audience appears on a website. The audience will have been created and enriched in the DMP, before being sent to the DSP to buy inventory.
The above process is roughly the same for all DMPs which I have worked with. However, this is often where the similarities end. Below I discuss some advance DMP strategies which will drive more value from your existing DMP investment.
In the scenarios I've outlined above, I always refer to the "device" being targeted, rather than the "person".
Many organisations have implemented their DMP to collect and classify device attributes. In 2017, the average person has access to 3.64 devices - this includes access to laptops, desktops, tablets and mobiles. By only collecting data at the device level, we (as marketers) are not seeing the full picture of the customer. Here's an example; a customer browsers your website on their laptop and abandons the basket. This device is now in an "Abandoned Basket" audience, and you start targeting display ads at the device to complete there purchase. Next, the person goes onto their tablet (a different device) and completes the purchase. Despite making a purchase on the tablet, the laptop device is still in the "Abandoned Basket" audience, so you'll continue to pay for ads that re-target this customer. Cross device identification solves this problem.
Cross device identification means that the DMP recognises that two or more devices are used by the same person. If we re-imagine the above scenario with cross device identification in place, then we can stop paying for display ads on the desktop, when a purchase is made on their tablet. This saves you money on display spend, as well as improving the customer experience.
Cross-device identification can work in two ways: Deterministic or Probabilistic. I won't cover the details of the device matching process now (maybe in a future post). But in summary, the DMP is able to link multiple devices to the same person either when the visitor logs in with the same userId across multiple devices (deterministic), or guessing that two devices belong to the same person because of the location, timestamp and behaviour of two devices (probabilistic).
If you aren't identifying people across devices today, then there is certainly value in doing so.
DMPs were initially created to help marketers to improve their return-on-ad-spend. However since their creation, they have matured into a fully cross-channel solution. Today DMPs can manage audiences across display ads, search, onsite optimisation, mobile apps and even email.
Channels such as display and search (PPC) are both a major part of most organisations marketing budgets. Both of these can be optimised by only bidding on you audiences you define in your DMP. For example, if your product is not suitable for a particular age group (such as a high value product which an 18 year old couldn't afford), then you can exclude this audience from PPC campaigns within Google's search results. Equally if a known customer of yours searches Google for your keywords, you may want to bid more to ensure you're listed in the results.
So what happens once they land on your site? Ideally your website will be optimised to help the visitor find relevant content as part of their journey. However, many companies today only optimise the site experience based on the referral source (where the visitor came from) or data collected within their current browsing session. Whilst this can be valuable data, the DMP contains a wealth of data which can be used to optimise the on-site experience. By utilising past behaviour, plus other 1st, 2nd and 3rd party data, you can ensure the most relevant content and experience is delivered to each visitor to your site. With the correct strategy, this will drive increased conversions on your site.
So how does this integration work? Typically when a visitor lands on your website, your DMP can feedback all known data about the individual (which may include data collected from other devices - see above!) to your website optimisation tool. Predefined rules within your website optimisation tool will then decide which content, recommendations, calls-to-action, layouts etc..is most relevant for this visitor. This DMP use case will accelerate your sales cycle and increase conversions by creating the most engaging user journeys for visitors as they transition from offsite to onsite browsing.
Testing should be built into everything you do. Email campaigns, website content and display ad units - everything should be continually tested to ensure you make incremental improvements towards your KPIs.
Today you can A/B or MVT test in most email providers, through your website optimisation solution, in Facebook or through your DSP. The problem is that each test is executed in a silo, and does not consider the entire customer journey. For example, you can create an A/B test in your email provider to test different variations on a call-to-action. Your email provider will then execute the test and measure the results. It will monitor conversions against each test variant, and tell you which call-to-action within the email lead to the highest conversion rate. However, the email provider has no control over the other touch points your organisation has with the visitor. There will be display ads, social ads, mobile ads, and on-site content that the visitor is exposed to which drive the conversion.
How valid are the results of a test when you evaluate the results based on a single channel?
The root of this problem, is that the audience was spilt into test groups within the email provider. This means there is no way to control the other touch points (display ads or on-site experience) made with the customer.
The solution is to create your test groups in your DMP. This way you can control the entire customer journey across all channels, ensuring they get consistent messages whist you choose which elements of the journey to test.
Today's leading DMPs will allow you to setup test groups and control groups, and will subsequently monitor a conversion rate to identify the winning group. This can be used to test everything from an ad's creative, to multiple DSP executing the buy, to onsite and mobile experience.
Whilst nearly all marketers today agree that testing should be a number one priority, most agree that they are not testing enough. By having a "test everything" philosophy grounded into your digital strategy, your DMP should be the solution to segment and measure audiences. This will give you more confidence in the validity of your results, and will increase the scope of what you can test.
A large proportion of marketing budgets today are spent on display advertising, and with this, more pressure is placed on marketing teams to improve the return on ad spend. So if acquisition is one of your KPIs, you should be using "look a like" models as a way to find the best new customers.
The name is pretty self-explanatory. Look a like models take a seed audience as input - such as your best customers. They then mine a third party dataset to find other devices which have shown similar attributes as your best customers. In other words, it helps you to find people who look like your best customers.
Here's how it works. You start by creating a seed audience from your first party data. For example, customers how have purchased recently, spent more an £100 and have visited the site frequently. You can then build a model based on this audience. Whilst each DMP may have its own algorithm to do this, fundamentally the DMP will analyse every data point associated with each member of the audience to calculate which data points occur more frequently in your seed audience than the general population. For example, my seed audience may be 5 times more likely to be female, and 7 times more likely to be aged 18-35 than the average member of the population. The DMP will actually analyse hundreds of thousands of data points, to identify which ones occur most frequently in your seed audience. Once the model has been built, the DMP will know which data attributes are most common in your seed audience and how these are weighted i.e. age is a more significant factor than age. The DMP can now mine other 3rd party data sets to find other devices which have the same combination of attributes. There is always a trade-off to be made between reach and accuracy. You can have a high reach and accept that some of those you target may not be a good match, or you can have a more accurate match but to a smaller audience.
The end result of this process is a new audience that shares the same attributes as your seed audience. You can now target this audience from your DSP as part of your acquisition strategy.
This should be a no-brainer if you're already using 3rd party data in acquisition campaigns, as it will immediately increase the relevance of your campaigns.
Your DMP will become more valuable as it ingests more data. The more data you have on your visitors, the more targeted and relevant your audiences will be. So far in this article I've focused on 1st party data from web browsing, as well as 2nd and 3rd party data.
However, most DMPs will ingest your 1st party data from many other sources such as offline data (such as CRM), ad server logs, email and mobile apps.
Lets take a look at the value each of these data sources brings.Offline/CRM Data
If you have customer data stored in an external system, this may be a valuable source of data which you can activate to target your paid media campaigns (display, PPC, social). For example, if you have a loyalty scheme where customers earn points, or achieve reward tiers (Gold, Silver, Bronze), then this data can be uploaded to the DMP. Audiences can be created to separate Gold tier customers from Bronze tier customers, so each loyalty tier experiences more relevant content and offers. Or if a customer has raised a complaint with the call centre, the visitor can be placed in an exclusion audience to remove them targeted ads until the issue is resolved.
This method is reliant on your customers authenticating on your site, for the DMP to match your onsite data to your offsite data. If your site visitors authenticate infrequently (for example, visitors only sign in annually), you can also onboard your offline data through a third party partner.
The result of this process will be additional data attributes that you can use for enriching data, granular targeting as well as excluding from targeting.Ad Impression Data
Ingesting ad impression data can be particularly valuable to cap the number of ad impressions to a single person, as well as helping to move a person down a predefined sales funnel.
A common way to track an ad impression is generate a pixel tag from your DMP and insert that pixel into the creative of the display ad. In theory where ever that ad is displayed, the pixel will fire back to the DMP. The DMP can now track that the ad has been display and to whom. The DMP can now count the number of times a display ad has been seen by a person, and prevent further impressions once a pre-defined threshold has been reach. However - there is a slight glitch here - many ad networks will block the pixel - so using this method will not provide 100% coverage.
Some DMPs (such as Adobe Audience Manager) can capture Google DCM log files so that impressions, clicks and conversions can be captured without the need for a pixel to be inserted in the ad creative. This provides a much greater coverage (for users of Google DCM), to be able to cap ad impressions at the person level.Email
Knowing which of your emails a person opens or not could be a valuable data attribute in moving a customer down a sales funnel. Your DMP should be able to generate a pixel to place inside your email creative to track when it is opened.Mobile App
Finally it's worth adding that everything that can be tracked from your website, can also be tracked within your mobile app. All behaviour within a mobile app (product views, categories view, conversions or features used) can be captured with an SDK. An SDK is is a piece of code embedded into your mobile app that communicates with the DMP to track the customers activity. Rather than a cookie, a device ID is generated to identify the device. Other than that, the process of creating and managing audiences, and syndicating audiences to a DSP are the same.
If you have a mobile, embedding your DMP's SDK should be a quick-win to capture additional data on your customers. Recent studies have shown that users behaviour differs between mobile apps and website. Capturing data from mobile devices will provide you with a more complete view of your customers attributes, and help deliver more targeted messages across all channels.