WhatsApp API metrics that trigger real action

Most teams that wire up the WhatsApp Business API stop measuring once the first template goes out. Messages get delivered, the dashboard stays green, and nobody looks again until a template is paused or the monthly bill jumps. The damage lands on ops and growth leads at Indian SMBs running WhatsApp as a real channel — support, order updates, re-engagement — without a shared definition of what working looks like. A short list of metrics with thresholds agreed in advance fixes most of it.

Which metrics deserve a dashboard

Five numbers are enough to run a WhatsApp channel; more than that and nobody opens the dashboard. Delivery rate, read rate, reply rate, block rate, and cost per qualified conversation will catch almost every problem before it gets expensive. Track them per template and per campaign rather than as one account-wide average, because a healthy utility-template read rate will hide a failing marketing one.

• Delivery rate — messages WhatsApp accepted and pushed to the device.
• Read rate — blue-tick reads, split by template category.
• Reply rate — inbound replies within 24 hours; the only honest engagement signal.
• Block rate — recipients who blocked or reported the sender.
• Cost per qualified conversation — spend divided by conversations that clear your own definition of useful.

If you capture these yourself instead of reading a BSP dashboard, the architecture notes cover a simple webhook-to-table pattern that holds up at SMB volumes.

The thresholds that should trigger action

Three tripwires deserve a same-day look; the rest can wait for a weekly review. Page someone when block rate crosses 0.5% in any 24-hour window, when a template's quality rating drops to yellow, or when cost per qualified conversation runs at more than twice your four-week baseline. Everything else belongs in a Monday meeting:

• Read rate under 50% on a marketing template after its first 1,000 sends — rewrite the opening line and the offer, not the send time.
• Reply rate under 2% on a cold list — the list or the offer is wrong; pause the campaign.
• First response time above 10 minutes in business hours on a support number — add routing or an auto-acknowledgement.
• Opt-out rate above 1% in a month — you are over-messaging; cut frequency before you rewrite copy.

Treat these as starting points rather than law. A small, well-consented list will beat every one of them comfortably.

Cost per qualified conversation is the number that survives a budget review

Cost per message is a vanity number; cost per qualified conversation is the one a finance lead will argue about. Take a B2B distributor sending 12,000 marketing template messages a month at roughly ₹0.80 each — about ₹9,600 in conversation charges before tooling, at recent India rates. A 58% read rate gives 6,960 reads, or ₹1.38 per read. If 3% of those reads turn into a reply or an order-form click, cost per qualified conversation is ₹46.

Now send 6,000 instead: only contacts who engaged in the last 90 days. At the same rate that is ₹4,800, a 70% read rate gives 4,200 reads, and a 5% qualify rate gives 210 conversations — ₹23 each. Same budget, half the waste, and a smaller list whose consent you can defend if it is ever questioned under the DPDP Act.

Template quality rating moves before your block rate does

Meta's green, yellow and red quality rating is the cheapest early warning you get, because it reacts to recipient feedback before your block rate visibly moves. The rating is per template, and a red rating can pause that template outright — which kills a campaign mid-week with no same-day appeal. When a template goes yellow, pause it yourself, exclude anyone who has not engaged in 60-90 days, and check the variables; a placeholder that resolves to a stale first name is a common cause of blocks. Shorten the message, make the offer concrete, and retest on your most engaged segment before sending to the full list. Swapping an emoji is not a fix.

Inbound flows need a different scorecard

On inbound conversations, speed beats engagement — a read rate tells you nothing about whether the customer got help. Track first-response time, containment rate (the share resolved without a human), and handoff failure rate. For B2B support numbers, a first response under 5 minutes between 9am and 7pm IST is a fair target; outside those hours an explicit auto-reply with a real callback time beats silence.

Containment of 40-65% is realistic for order status, delivery rescheduling and basic account queries. Below 30% after four weeks usually means the menu has too many branches or the bot cannot see order data. Above 85% is worth a CSAT check, because it can mean you are deflecting people who needed a person. A small classification layer is often what moves containment from the thirties into the fifties, and every handoff failure — the bot says it is connecting you and nothing happens — should be logged and treated as a bug, not a statistic.

Where to start

Pick three metrics from the list above — read rate, block rate and cost per qualified conversation is a reasonable first set — put them on one screen, and name the person who gets pinged when a threshold trips. Instrument the webhook before you build the dashboard, because a week of untagged message data is worse than none at all. If you would rather not build that plumbing in-house, this is the kind of WhatsApp, CRM and reporting work we do at Maysan Labs.