When discussing marketing and personalization, we all know it is table stakes to know your clients well. After all, if you don't know them, how will you serve them the best way possible?
Let's say you work for a financial services company. How will you know what services the customer is interested in? Is it worth focusing on publicizing investment products if they are interested in a no-annual-fee credit card?
Thousands of platforms promise to help brands to fully understand their clients — what they buy, how often they access the website, what their interests are, etc. However, the technical data world can be pretty imprecise at an individual level. There is plenty of fake information on individual customers.
Understanding each customer as unique is essential so that brands know which users their media campaigns draw into their websites. That is impossible if you can't recognize the same person interacting with various devices.
Take a minute and ask yourself: how many devices do you or your family use daily? In 2021, a research study by Deloitte collected data stating that the average US household has 25 connected devices – more than double the 11 that the average household had in 2019.
As digital campaigns proliferate on countless devices and channels, the challenge of a marketing professional becomes increasingly more outstanding. Understanding the characteristics of those who use only one device is no longer a competitive advantage: now, we need to understand the person who interacts with at least two devices.
Identity resolution is the solution for recognizing your customers. This awareness enables us to design a coherent image of people as they navigate among channels and devices and integrate through different “identifiers”. It can include, for example, emails, phone numbers, cookies, and device IDs.
How your technology partners approach identity resolution significantly impacts all your marketing decisions. For that reason, we need to better understand the concepts linked to it.
An individual's identity includes offline identifiers — such as name, address, and phone numbers — and online ones — such as email addresses, cookies, device IDs, etc. Identity resolution may look complex, but failure to understand how it works can result in decisions based on false information.
An isolated identifier is not enough to identify a person in most cases. For example, let's look at two identifiers: address and name. The address may not be enough when used alone, as it can relate to more than one person. However, when linked to a name, there are increased chances of recognizing a person.
Identity resolution makes it possible to connect the dots between these identifiers more precisely. This is scalable and compatible with specific privacy policies for producing a stable and consistent view of your customer. Knowing the right questions regarding the approach a technological partner uses for identity resolution is essential to ensure your methods will increase your ROI.
Next, we'll discuss how identity resolution works and the different market approaches.
There are two approaches to identity resolution: deterministic and probabilistic. These terms tend to imply value judgments, yet no characterization is precise.
Deterministic data can be an email address related to a cookie or a device ID. Probabilistic data helps make inferences based on your customer's behavioral signs and interactions.
The deterministic approach evaluates an explicit link among identifiers, while the probabilistic method shows how implicitly strong that connection is. It is essential to assess three factors when using identity resolution:
- Precision: the importance of using only precise links to ensure that the resolution always connects identifiers that belong to the same person
- Wealth: the complexity of understanding the customers' profiles, across digital and offline channels, devices, demographics, etc
- Scale: the amount of data and the possibility of capturing more links on the customer and creating the largest possible target public.
Both deterministic and probabilistic methods have advantages and disadvantages when seeking these objectives for actual overall identity resolution.
The deterministic method helps create a single customer profile from two directly linked identifiers. That is possible, for example, when a user logs in to a site using an email address on a computer and later purchases on the same website from their smartphone. Then, the two identities — one from the computer and the other from the smartphone — will be linked through email, and the user profile will contain information from both devices.
Although the deterministic approach is genuinely omnichannel, as it connects identifiers from the digital and offline worlds, it can be problematic, as users usually share certain app accounts among relatives and friends. When a user shares a Netflix password or their computer with a friend for them to check their emails, we can have incorrect links. Besides that, due to its nature, deterministic data are limited in quantity — as there is a small and finite number of explicit connections — thus interfering with the model's scalability.
On the other hand, the probabilistic approach analyzes various signs to find out which identifiers can be linked with high reliability. As an example, suppose the same device ID and desktop cookie visits a website from a home IP address, several nights a week. In that case, the probabilistic methods would conclude that all these signs belong to the same family.
Additionally, suppose we have a scenario where two devices — a computer and a smartphone — share the same data frequently. In that case, if both are connected simultaneously to various IP addresses — at home, at work, at an airport, etc — we can assume that these two devices belong to the same person.
The probabilistic approach is handy for ruling out false information. It analyzes a large amount of data instead of only binary correspondences, as in the deterministic method. On the other hand, as this approach is limited to online data, it is ineffective when discussing identity resolution involving real-world interactions.
Some technologies use deterministic and probabilistic methods to create a precise and scalable approach. Theoretically, this seems like the best of both worlds. Unfortunately, it is slightly different when put into practice due to low-quality data, inaccurate links, and contaminated data clusters.
For the hybrid method to succeed, marketing professionals need to apply deterministic approach results as a source for probabilistic approaches, so they can evaluate which connections are relevant. In other words, you can't just match deterministic data with probabilistic data after creating both by different methods. It would be like adding yeast to a piece of baked bread and expecting it to keep rising.
Identity data has become vital in acquiring and increasing customer loyalty.
This means that you can accurately segment an individual or a family within the graph, so you can connect contact point data with purchasing data or other results. To do this correctly, you need a stable end-to-end identity resolution system, which has the ability to evaluate and make decisions on an individual level, tracking each contact point in a granular way.
The importance of a robust identity resolution system for any user-centric brands has been indisputable for quite some time now. Today, marketers are more demanding when it comes to identity and seek an in-depth understanding of the methodologies used by their technological partners. Understanding the data science behind keywords is essential, going beyond what the word "determinism" suggests and the risk that "probabilistic" implies.
You need an in-depth understanding of identity resolution that will allow you to confirm if the marketing platform and measurement approach can operate at the highest levels of a granular, consistent, and stable identity. If you cannot deliver the best experience online, offline, and across multiple channels to the right consumer, your results are unlikely to be satisfactory, no matter how good your ideas, campaigns, messages, or algorithms are.
At Croct we offer identity resolution as one of our features. After all, there's nothing better than understanding your users in-depth through their interaction across multiple channels. We believe this is the best way to create a unique and personalized experience for each of them.
We use the hybrid method in the right way, so you can get the best out of your data, regardless of how many devices each user accesses or how many different domains your company has.
In addition, we only use data provided by your users, which ensures privacy and security at all times on the user's journey. If you want to understand better how our engine works or how we can help you create better browsing experiences, please get in touch with us! We will be happy to chat with you :)