Support Deflection Honesty Grader

The honesty of a support deflection (or containment) rate. For each cohort you enter the claimed rate plus how the deflected bucket actually split — genuinely self-served & resolved, abandoned, repeat contact, and reached-a-human-anyway (the four shares total 100%). It returns the genuinely-resolved share (honesty), the real defensible rate, the hidden-demand share, and a verdict: REAL / SOFT / FALSE DEFLECTION. It grades the metric claim from your own data — never people, and it's not a guaranteed or benchmarked rate.

Because "deflected" usually means "didn't reach an agent," which silently includes the people who just gave up. A bot can hit 60% deflection by burying the path to a human — but that demand didn't vanish, it moved to abandonment, repeat contacts, and churn. The headline rate looks great right up until someone asks how many of those actually got resolved. This splits the bucket so you know that before they ask.

Real deflection rate = claimed rate × the genuinely-resolved share. An 80% claim where 88% of the bucket truly self-served is a real rate of 70% — the number that survives a follow-up question. The kit computes it per cohort so you report the defensible figure instead of the inflated headline. In the worked example, a Billing bot's 62% claim is really 34%.

Because hidden demand is a veto. The hollow-deflection gate is dispositive: if a third or more of the deflected bucket (35%+) is abandonment plus repeat contact, the cohort is FALSE DEFLECTION no matter how the resolved share looks. The Billing bot has 55% resolved — above the FALSE floor on its own — yet gates to FALSE because 40% of its bucket is people abandoning or coming back. A deflection that mostly produces give-ups isn't resolution; it's hidden demand wearing a good metric's clothes.

Different jobs, and they pair. The AI Support Deflection Kit ($89) is the toolkit for building honest deflection — escalation rules, a playbook, KB templates, a CSAT tracker, a prompt pack. This grader is the measurement that audits whether a deflection rate you're already reporting is honest. Build it with the Kit, prove it's real with this. It's the same "do it / grade the result" split as the marketing line's measurement kit.

One .xlsx with three tabs — Start Here, Deflection Honesty (a row per cohort: claimed rate + the four-way split → honesty, real rate, hidden-demand share, verdict), and a Dashboard roll-up — that opens in Excel, Google Sheets, or Numbers. It's deterministic and offline, nothing uploaded. It applies no industry benchmark and isn't a guaranteed rate or financial advice; it just makes the number you report defensible.