Evaluate · Account intelligence
Finding the accounts most likely to move.
I was working with a target universe of more than 1,500 prospective advertisers, which created a practical question: which accounts should Sales prioritise?
Company size, industry and geography could tell us whether an organisation matched the target profile, but I wanted to know whether we could also identify behaviour that suggested an advertiser might be more open to experimenting with a new advertising platform.
I developed an account qualification approach that combined those behavioural signals with account fit and first party engagement, then measured whether the resulting scores were actually associated with stronger Sales progression.
The challenge
How do you decide which accounts deserve Sales attention when more than 1,500 companies fit your target profile?
We had a large universe of prospective advertisers that met our basic qualification criteria, but Sales couldn't pursue every account with the same level of attention.
I needed a way to distinguish companies that simply looked like good prospects on paper from those showing behaviour that suggested they might actually be more open to trying a new advertising platform.
That meant finding signals that could tell us something useful about buying propensity before Sales invested significant time in the account.
The hypothesis
Could an advertiser's existing media behaviour tell us something about its appetite for experimentation?
The technology already present on an advertiser's website provided a way to observe part of its media behaviour.
Meta and Google were widely used across sophisticated digital advertisers, which meant their presence alone offered relatively little differentiation. Platforms such as Snapchat, Reddit and other less ubiquitous channels were more interesting because their presence suggested that an advertiser had already allocated budget beyond the most established parts of the media mix.
I wanted to test whether that willingness to experiment elsewhere could become a useful signal when deciding which accounts might be more receptive to another platform.
The build
Testing the hypothesis across more than 1,500 companies required a way to conduct the research at scale.
Manually inspecting every advertiser website would have made the idea impractical, so I used AI as a development partner to build a scanner that could conduct the technical research automatically.
The scanner crawled advertiser websites and inspected page source, script tags, tag manager configurations and known pixel domains for evidence of advertising technology. A structured technology library then mapped those detections to the platforms being used by each advertiser.
Where the technical evidence was clear, classification remained deterministic. The system relied on observable evidence such as scripts, pixels and domains rather than asking an AI model to infer whether an advertiser used a particular platform.
AI helped me accelerate the development of the crawler, structure the pattern matching logic and work through the exceptions involved in turning the original hypothesis into a usable system.
The resulting workflow reduced the amount of manual research required by approximately 70%.
The signals
I combined external media behaviour with what we were already seeing inside our own environment.
The website analysis gave us one view of an advertiser's behaviour. An organisation using several less ubiquitous advertising platforms provided stronger evidence of experimentation than one whose detected media technology was concentrated around Meta and Google.
I then combined that external behaviour with first party engagement.
Industry specific assets were already attracting people from companies in our target universe, so repeated engagement from relevant employees within the same organisation provided another piece of evidence about potential interest.
Neither signal determined prioritisation on its own. Together with account fit, they gave us a richer picture of which organisations appeared to deserve attention.
Prioritisation
The account score brought the different pieces of evidence together.
The resulting model combined account fit, experimental media behaviour and first party engagement so that accounts could accumulate evidence across several dimensions rather than being prioritised because of one isolated action.
That gave Sales a way to distinguish organisations showing several indicators of potential readiness from organisations that simply satisfied the basic target profile.
The purpose of the score was practical: create a better ordering of the account universe so that limited Sales attention could be concentrated where the evidence was strongest.
What happened
We measured the strength of the signals by looking at what happened after accounts reached Sales.
High scoring accounts converted from MQL to SQL at 34%, compared with 22% across the broader qualified account population, representing a 55% relative improvement in conversion.
That downstream performance gave us a way to evaluate the quality of the underlying signals rather than assuming that every observed behaviour represented intent.
Signals associated with stronger commercial progression could receive greater weight over time, while signals that appeared interesting but did not translate into progression could be reduced.
The account score could therefore improve through the outcomes it produced, creating a feedback loop between the original signal hypothesis and actual Sales conversion.
Takeaway
Better account scoring helps put attention where the evidence is strongest.
The value of the system was not the number of accounts it could analyse. It was the ability to turn a large prospect universe into a more informed decision about where Sales attention should go.