How we identify red flags
  • 1 minute read
  • Contributors

How we identify red flags

Article summary

Our red flag warning alerts you when a prospect sends an email that indicates there is potential risk to the deal, whether it signifies cause for alarm, a delay, or rejection.

How does Gong determine that an email contains a red flag?

We've developed a proprietary machine learning-based model for determining whether an email contains a red flag.

As a baseline, our model has an advanced, contextual understanding of human language drawn from extensive pre-training that includes reading millions of pages from all over the internet.

We then trained our model specifically on what a red flag is by feeding it a few hundred real examples from sales emails drawn from multiple customers in multiple industries. For example, sentences such as the following:

  • Although the team feels strongly about your products, they are unsure about the funding aspect of it.

  • I have talked to the decision-makers and they haven’t acted like they are interested at this time.

  • We are cutting our budgets across the whole business and are unable to commit to any additional costs at the moment.

  • This project has had to drop down the priority list.

We trained the model to generalize language patterns and understand similar intent, even if phrased differently.

We scan emails sent by your prospects, and apply the model to identify red flags.

Does Gong rely on certain keywords?

We don't. Instead of using keywords, our machine learning-based models take into account the full context of the communication to assess whether or not the email presents a red flag.

For example, the words “unsure” or “priority” in the examples given above are not strong enough indicators that the statements are red flags, and the full sentences put them in context. We can understand the intent of what’s being said, regardless of whether or not we've seen the exact words before.

How accurately can Gong identify red flags?

Our red flag detector can detect red flags with a precision of 90%. In other words, based on our test data, up to 10% of the red flags flagged by the system may be judged by experts to not truly be red flags.

We continuously adapt and improve our models to stay current with language, and we add examples of red flags as they’re identified.

Was this article helpful?


Eddy AI, a genAI helper, will scrub our help center to give you an answer that summarizes our content. Ask a question in plain language and let me do the rest.