Set your 'clean enough to trust' thresholds

Perfect data is impossible and chasing it is a waste of a founder's scarcest resource. The bar is not perfect, the bar is clean enough to trust, and the discipline is to define that bar concretely enough that you can measure it and an agent can enforce it. Vague aspirations like 'keep the CRM tidy' are unenforceable. Numbers are not.

Pick three to five thresholds, not forty

Resist the urge to write a forty-field perfection specification. You want a small set of thresholds, three to five, that together stand in for trustworthiness, and that you can actually monitor. Here is the set I would start with for a lean B2B operation:

• Every open deal has a valid stage, a named owner, and a defined next step. This is the single most important bar, because a pipeline of zombie deals with no next step is exactly what drives teams to the side spreadsheet. If every live deal has those three fields, the pipeline view is trustworthy enough to run a forecast on.
• Duplicate rate under about 2 per cent. Disciplined teams hold the 1 to 2 per cent benchmark; let it climb and trust erodes because nobody knows which of the three records is the real one.
• Hard-bounce rate under 2 per cent. More on why this is a hygiene metric below.

Three bars. Measurable, monitorable, and between them they cover the failure modes that actually break trust.

Why bounce rate is a hygiene metric

Founders treat bounce rate as an email problem and a hygiene metric as a database problem, and miss that they are the same problem. Stale contacts do not sit harmlessly in a corner, they actively degrade the deliverability of your outreach to the good contacts. Email lists with hard-bounce rates above 2 per cent damage your sender reputation, and above 5 per cent you risk outright blacklisting. So a list full of people who changed jobs is not just wasting a few sends, it is throttling every send, poisoning the inbox placement of the messages that would have landed. That is why bounce rate belongs on your hygiene dashboard, it is the clearest signal that contact decay is actively costing you reach.

A threshold you can measure is a threshold an agent can enforce

The deeper reason to set concrete numbers is automation. 'Keep it clean' is a wish an AI agent cannot act on. 'Flag every open deal missing a next step, alert me when duplicate rate crosses 2 per cent, suppress contacts the moment they hard-bounce' are instructions an agent can run all day without you. The act of writing your thresholds down as numbers is the act of making them machine-enforceable. Pick the three to five that matter, define them as values, and you have handed your future AI process its job description. Anything beyond that bar is effort spent polishing data nobody was going to distrust anyway.