Semcasting unveils ‘first self-serve deterministic attribution platform’
Employing IP and street addresses, the platform makes available the process that Semcasting had been providing as a professional service.
Attribution is one of those services that seems to take place inside a black box, where all kinds of algorithms make all kinds of assumptions.
To help make the box less obscure, Andover, Massachusetts-based Semcasting is out today with The Attributor, which it describes as the first self-service deterministic attribution platform.
The platform, CEO/founder Ray Kingman told me, is transparent about its processes, and it makes clear when matches are made and when not. The company has been offering this process as a professional service for the last year, and now it’s available in a self-service platform with usage pricing.
How it works. The company specializes in audience targeting via IP address, and that’s a key element of The Attributor, which is designed to show if the viewers of your site or online ad made a purchase or took some other desired action.
Kingman suggested the following typical use case for The Attributor.
Let’s say AutoTrader.com wants to determine which car buyers went to its site, so it can show a possible causal connection to auto dealerships that advertise on AutoTrader. In addition to web site visitors, each Study could also have up to four other audiences as the “base” for a match, such as emails or profiles in a customer relationship management system.
In the case of site visitors, Semcasting converts the web traffic logs from AutoTrader.com to street addresses of the web visitors. It does this by matching static IP addresses of visitors to its graph of user profiles, which contain frequently used static IPs — such as ones for a home or office – and a street address for that person or household.
If the IP is a dynamic one, such as used by mobile towers, the Semcasting platform collects time signals to determine the latitude/longiture of the towers, and it then can match the device ID that passed by those towers at that time, to the street address.
The Attributor is integrated with a variety of sales-oriented databases, such as the Dealer Management System that compiles all car sales in realtime. Through the Dealer Management System, the dealership in question provides a list of street addresses for all its car buyers in, say, the last month, and the match is made. Semcasting says it doesn’t keep any first-party personal data, such as the buyers’ names.
Other audience segments or “tactics” could also be employed to match the IP addresses of visitors to web sites or to those who were delivered ad impressions. These can include email addresses of prospects, purchases recorded in customer relationship management systems, or similar data. Once the data is onboarded, Kingman added, the platform does the rest.
He said the 85-90 percent match rate is about 70 to 80 percent accurate. If the IP address is for, say, an apartment building, the platform makes a probabilistic determination of the likely tenant, unless there is other data.
Why this matters. Attribution helps to determine if your ad or other marketing spend has been effective for the desired outcome, such as purchases.
Probabilistic attribution makes a variety of assumptions and combines probabilities. By contrast, a deterministic attribution is more definite and can be more accurate, as it matches a persistent identifier like a street address. But deterministic attribution is usually conducted via a professional service, so a self-serve, usage-based platform can make this approach easier and less expensive.
Semcasting says other useful questions its Attributor platform can answer include:
- How many of my CRM customers saw my connected TV advertising?
- What percent of individuals who received our CRM emails eventually showed up at the website or in the store?
- There were very few clicks on our display ads but website visits increased measurably during the campaign. Were those visits the result of our online advertising efforts?
Opinions expressed in this article are those of the guest author and not necessarily MarTech Today. Staff authors are listed here.