What is numeric coverage in FMCG distribution?

The count of outlets that took the goods on an invoice in a period. A visit alone is not coverage.

What does weighted coverage tell an AI agent that numeric coverage does not?

How much of network volume or value those outlets represent. A few large accounts can hold weighted coverage up while the network thins.

Where must the agent refuse to build coverage from?

Beat targets, salesperson reports, or an industry average. Only invoices with an outlet identity.

How an AI agent reads numeric and weighted store coverage in FMCG distribution

Store coverage says how many outlets had the goods, not how much they sold. Numeric coverage counts outlets. Weighted coverage weights those outlets by volume or value. In FMCG distribution you need both, and each one lies on its own. Didrah’s AI agent builds coverage from outlet invoices. If there is no invoice, coverage must not be guessed from a salesperson’s call plan.

How numeric and weighted coverage lead to different distribution decisions

High numeric coverage and low weighted coverage means the SKU sits in many small shops and is missing from high-volume points. High weighted coverage and low numeric coverage means a few wholesalers or hypermarkets carry the book, and the network is brittle. If a sales manager watches only one, they either push the team for invoice count or stay hooked on a few large accounts. The agent should put both in one claim and say which channel built the gap.

Which supply-chain evidence an AI agent uses for coverage

Coverage evidence is an invoice with an outlet identity in a time window. A call without an invoice is not coverage. A beat target is not coverage. If the customer is still flagged active but has no invoice, that point is lost coverage, even if a visit was logged. Agentic AI must keep active customer and coverage apart: one means still on the book, the other means they took goods in the period.

How the agent ties a coverage drop to a distribution action

If numeric coverage fell and volume held, small outlets dropped while large ones still buy. The action is usually the beat, not brand budget. If both fell, the next question is regional warehouse stock or price. Human-in-the-loop means the agent proposes a fallback route; a manager signs cutting a channel. If confidence is medium because outlet identity is messy, fix the data first.

What coverage question to ask a distribution AI agent

Do not treat a call-plan target as coverage; the invoice is the evidence unit.
Always ask for numeric and weighted together, same region, same week.
If outlet identity is incomplete, the agent should return a data gap, not an invented percentage.
Put a coverage drop on the table before you raise trade marketing spend.

Check coverage every week with the distribution AI agent

Once a week, ask for numeric and weighted on the same brand and region. If numeric fell and weighted held, open the small-outlet beat. If weighted fell and numeric held, look at large accounts first, not the high-call sellers. That split is the agent’s job because it repeats on invoices, not once in a seasonal pack. If outlet identity was incomplete that week, do not put a percentage in the pack. A made-up coverage figure is worse than an honest data gap. It sends trade marketing money to a shop that does not exist in the data.