FAQs: Account Identification

The core objective of Demandbase’s Account Identification is to identify employees of prospects and customers when they are browsing your website, before they provide their contact information to you. Our company identity graph and machine learning techniques allow us to identify many of the unknown visitors visiting your website who have not yet clicked on an email, filled out a form, or provided their corporate email address. Many marketing automation solutions, such as Pardot, Marketo, or HubSpot, identify leads via a lead form-fill or email clicks with a corporate email address. While these can be useful means of identification, they occur late in the customer's journey and represent a small percentage of your total website traffic, resulting in many missed opportunities. Because Demandbase identification leverages a probabilistic AI model on top of our identity graph made up of billions of IP and cookie identifications, there is always some risk of inaccuracies. That said, our model ultimately identifies many companies visiting your website and exhibiting high intent across the Internet much earlier in the customer's journey.

A form fill is already half-way through the customer's journey. Demandbase identifies potential customers while they are still unknown.

Demandbase uses probabilities (using AI-powered predictive models) throughout the customer journey from identification through attraction, engagement, conversion, closing, retention, upsell, and cross-sell. The following plots are for illustrative purposes only.

Traditional Volume Marketing with volume leads wastes a lot of spend. Even worse, it takes away your time and focus from real opportunities and can result in lost sales due to missed opportunities.

The opposite extreme would be to wait for hand raises from customers who really want to and can buy your product. You wouldn't have any wasted spend, but waiting passively can waste time and result in lost sales due to missed opportunities.

Consider instead, Account-Based Marketing with Artificial Intelligence, which takes advantage of probabilities. These predictive models use objective criteria, so they focus your spend on accounts with more potential and drive higher success rate. According to McKinsey and Company, "It can reduce acquisition costs by as much as 50%, lift revenues by 5-15%, and increase the efficiency of marketing spend by 10-30%.”

To identify visitors, we start by using the billions of facts in our company identity graph, such as known cookies, IP addresses, 3rd party IDs, device IDs, and public VPN and ZTNA providers. We then use machine learning (AI trained on the facts in our identity graph) to predict additional identifications based on commonality in web activity across all IP addresses and cookies observed by Demandbase (e.g. our B2B DSP observes 18+ Billion signals per day). Since the predictive model is ultimately probabilistic and use of IPs by companies and employees working remotely is so dynamic, we expect some deviations from the facts. However, continuous improvement is built into our Account Identification model by enabling it to continually learn from new information (i.e. observed facts in our IP and cookie graph). 

Our algorithm has a high degree of accuracy because it has been learning and updating for over a decade with hundreds of billions of data signals observed to date. In addition, we’re always applying new techniques and new data sources in our predictive models. For example, during COVID lockdown, with so many people working from home, we learned to adjust identification based on whether the same company uses an IP address (confirming the association) or a mix of different companies uses it (showing that they are too loosely associated).

With the integration of Engagio, InsideView, and our partnership with LiveRamp, our overall identification accuracy has increased even further. With Engagio's first-party data sources, InsideView’s third-party B2B contacts, and LiveRamp’s privacy-compliant identity resolution, we have access to even more observed facts on which to train our Account Identification predictive models.”

Please inform us of inaccuracies. The more you allow us to review them in your day-to-day work, the better we can sort out different types of misidentifications, and the faster our predictive models can learn.

In the meantime, we continue to work on the accuracy of our account identification. For any discrepancies found, please report them to support@demandbase.com, so we can investigate and take corrective action.

We have data from third-party sources as well as a crawler that we use to map domains to each other, based on activity and redirects. We intentionally do not map all subdomains to top-level domains, because subdomains (especially internationally) often map to different brands that have been acquired under the parent. However, in some cases, it could be that the domain was valid when we published it and has since become invalid. We can correct it when it’s from a valid record in our database.

ABM is still a rapidly developing technology, with significant variability in the availability and character of digital signals in different geographies like North America and Europe relative to certain countries in APAC and the Middle East. For example, unfortunately we still see a higher incidence of misidentification in India, where on average one IP address is used by many more households or businesses as compared to the average IP-to-household ratio seen in North America, Europe, or ANZ. Other ABM and identification solution providers face similar challenges in APAC and the Middle East. We try to mitigate this issue as much as possible by complimenting our IP identification with Cookie identification and continuing to expand the reach of our B2B DSP (the largest source of digital signals observed globally).

ABM is still a rapidly developing technology. Much of the data we use for our product in the US and Europe is not available in other regions. To name a few, government company records, private data providers, IP registries, ad networks, and B2B websites all vary drastically in availability, cost, and technological integration needs. While we can examine a company’s website and various public and partner data sources to create an accurate company record in our system, anonymous activity by employees of that company without digital signals such as IPs and cookies associated with verifiable traffic. Other ABM and identification solution providers face similar issues in these regions.

That being said, we are constantly examining the available data sources and trying new proprietary techniques to expand our coverage in the Middle East and APAC (about 48 countries, including India, Indonesia, Pakistan, Japan, Philippines). While it is still too early to estimate a timeline for fuller coverage, rest assured this initiative is a high, long-term priority and as data and signal availability improves, we will be able to expand coverage in these regions.

Many large, global companies have multiple different domains and brands for each market or region they operate in. In these cases, most of the ABM signal we receive will be grouped at the parent domain level, for example, traffic and signal from “.mx”, “.fr”, and “.uk” domains will all be rolled up into the “.com” domain. This is because much of the signal we receive from various third-party and proprietary sources does not map reliably to a single, specific domain. This is why some features, such as advertising, allow you to filter to specific states or countries, so that you can target the right market or region you are interested in.

We recognize this kind of geographic detail is important to our clients and we are working on making it available across the platform.

In these cases, we are not always able to distinguish between traffic from the parent company and the subsidiary company or brand. Many times the subsidiary team uses parent company IT services, meaning all of their traffic comes from the same IPs as does traffic from the parent company. That being said, our algorithms and underlying data are constantly improving, so we are able to identify this traffic separately in some cases.

On average we currently identify about 20% of B2B traffic by cookie only, but it can vary widely by client. Keep in mind this would be on average about 10% of overall traffic since much of the overall traffic we see is not B2B. This again, can vary widely by client.

Demandbase Account Identification allows us to identify many of the unknown visitors to your website before they click on an email, fill out a form, or provide their corporate email address. However, there are limitations to the visitors we can identify. For instance, if a visitor clicks an ad while using the LinkedIn app on their mobile device, we may not be able to identify them and match them to their company’s account. The same challenge exists when a visitor is directed to your website from an ad using a device without cookies enabled or from an unidentifiable IP address (e.g. shared wireless network, cloud VPN, mobile network, etc.).

The most common reason that LinkedIn ad clicks may not be identifiable by us is because 80% of all engagement with LinkedIn content is now on mobile. While some of those mobile interactions occur on office or home wireless networks (which we can typically identify), nearly all of those interactions that occur while connected to mobile/cellular networks will not be identifiable. Although Demandbase does use mobile cookies to identify traffic, they’re less prevalent particularly for iOS devices (as iOS Safari disables third-party cookies by default while Android Chrome allows cookies by default).

Additionally, users often share a link directly from a LinkedIn ad to colleagues, themselves, etc. When this occurs, the UTM parameter on the URL is also copied. The sharing and distribution of these links will often result in Demandbase identifying traffic other than the intended/targeted audience on LinkedIn.

LinkedIn, along with other user-authenticated websites or social networks (i.e. “walled gardens”), can serve targeted ads regardless of which device or network a user is on. However, even though the average time on site for LinkedIn users has grown to over 7 minutes per day, the average person spends 400 minutes or more each day on all other websites. For this remaining 98% of average daily internet usage, Demandbase’s Account Identification and B2B ad tech helps our customers to reach B2B buyers everywhere else they consume content.