Building a Universal Event Taxonomy
Nancy Dong, CEO of Roster, on the rise of ops-centric tooling
The hard part is not collecting clicks, it is deciding what each action actually means in the context of a sales process. Roster is trying to turn messy tool level exhaust from Salesforce, Gong, and Salesloft into one comparable timeline of work, then connect that timeline to outcomes like ramp time and quota attainment. That requires translating many different low level actions into a shared model of steps like prep, follow up, coaching, and deal progression.
-
Each source system records activity in its own shape. Salesforce stores tasks and events as CRM activity objects tied to accounts, leads, contacts, and opportunities, while Gong exposes call level fields like participants, topics, snippets, account name, and opportunity data. A universal log has to normalize all of that into one sequence before any analysis works.
-
The real modeling step is categorization. Listening to a rep call in Gong, editing opportunity notes in Salesforce, and sending a recap in Salesloft are different raw events, but Roster needs to infer that they are all parts of a broader workflow such as deal inspection or post call follow up. That is what makes the data apples to apples instead of just a pile of timestamps.
-
This is why the product sits closer to an operational system of record than a BI dashboard. Traditional BI can show that a quarter is off plan, but it usually does not encode the business context needed to say which behaviors changed, which process broke, and what managers should fix next. The value is in turning behavior data into recommended actions.
If this model matures, ops software starts to look more like a behavioral ledger for internal work. The companies that win here will not just ingest more data, they will define the standard event taxonomy for how revenue teams actually operate, and use that shared model to power recommendations, automation, and eventually cross functional expansion beyond sales.