Measurement for SMEs when half the demand never hits analytics

Most SME professional services firms have one marketing lead and a dashboard that looks authoritative. The dashboard cannot see GPTBot, ClaudeBot or PerplexityBot, because those fetchers never run the scripts analytics depends on. Zero-click search already hid a large share of demand before assistants arrived. Partners still ask for last-click stories. The honest operating model is smaller and sterner: count what you can still see, label what you cannot, and stop scaling spend against vanity.

The diagnostic essay is Your analytics cannot see AI assistants. This piece turns that diagnosis into a working rhythm for firms without an analytics team.

What should you count instead?

Start with qualified enquiries: forms and calls that name a service, a need you actually handle, or a question answered on the site, and that the firm accepts as in scope.

Add the share of those enquiries you can attribute at all, even roughly. Add whether the pages you care about are being fetched in server logs, including known assistant user agents where you can identify them. Add branded search interest as a coarse signal. Add a lightweight assisted-mention check: periodically ask the tools your buyers use who they recommend for your category, and record what appears.

Keep the client-side suite for human post-click behaviour. Scroll depth, form starts, landing paths still matter when a person arrives. Just stop treating that view as the whole market.

Two follow-ups people usually ask next:

Is log analysis expensive?

It need not be. Even a monthly export reviewed against a short list of priority URLs beats another vanity dashboard. Analytics and Measurement is how we install that discipline.

Can we attribute revenue to AI assistants?

Rarely with honesty. You can show assisted presence, referred visits, and enquiry quality trends. Inventing a precise AI revenue line is how marketing loses partner trust.

How to report to a partnership

Bring three numbers you will defend and one explicit blind spot. Volume of qualified enquiries. Cost or effort per qualified enquiry where paid is in play. Whether priority pages are readable and fetched. Then say plainly what the dashboard cannot see. Partners prefer an uncomfortable truth to a confident fiction. The partner questions piece sits beside this: What a partner should ask before signing off marketing spend.

If logs show assistants reading pages that cannot convert, the next job is the site (Conversion, Content and Editorial), not more media. If nothing is being read, the next job is visibility (Visibility, Search and AI Visibility). Measurement is the first move at growth (Growth). Spend comes second.

When to stop scaling paid

Stop when sessions rise and qualified enquiries do not. Stop when landing pages still cannot answer who you are for. Stop when the only proof in the room is a channel report that ignores the blind spot. Paid Media amplifies destinations that are ready. It is a poor instrument for inventing readiness.

SMEs do not need a measurement cult. They need a short board pack that survives a sceptical partner. Price bands for putting that pack in place sit on Pricing. If your dashboard looks healthy and the partnership does not believe it, start a conversation. We will say whether the gap is instrumentation, the site, or a story marketing should stop telling.

Key takeaways

• Assistants and zero-click demand sit partly outside client-side analytics. Say so.
• Count qualified enquiries, fetch evidence, branded search and assisted mentions beside the dashboard.
• Report a small set of defended numbers and one labelled blind spot.
• Do not scale paid spend against vanity sessions when enquiry quality is flat.

The board does not need a bigger dashboard. It needs a shorter one, with each number tied to a source it can check and a plain note on what the firm still cannot see.