Marketing analytics and client reporting
The average client dashboard is opened once, during the call where it was introduced. The problem is not the data. It is that a dashboard answers questions nobody asked.
Why marketing dashboards and client reports go unread
A dashboard is a general-purpose instrument. It shows many numbers so that any question might be answerable. This is exactly wrong for a client, who has one question — is this working? — and no appetite for deriving the answer themselves.
Worse, dashboards present numbers without interpretation, which quietly transfers the analytical work to the person least equipped to do it. A client looking at a 12% drop in one metric and a 4% rise in another does not know whether to be worried. They will either ignore it or ask you — and if they have to ask, the dashboard cost you time rather than saving it.
This does not mean dashboards are useless. It means they are a reference layer, not a reporting layer. The report is the story. The dashboard is where someone goes to check the story.
Decision metrics vs vanity metrics in marketing analytics
The distinction is not about which numbers are impressive. It is about whether the number would change what anyone does.
A decision metric changes behaviour when it moves. Cost per qualified lead rises 40% — you investigate and reallocate. Conversion rate on the pricing page drops — you look at what changed.
A vanity metric moves and nothing happens. Impressions, follower counts, raw sessions, "reach". Not fraudulent, and occasionally useful as context, but nobody has ever changed a plan because impressions went up.
A practical filter before adding anything to a report: if this number doubled, or halved, what would we do differently? If you cannot answer, it belongs in the appendix.
Worth noticing: reports thick with vanity metrics are usually defensive — they demonstrate activity when results are uncertain. Clients read this correctly, if not always consciously. A short report showing one disappointing number with a clear plan attached builds far more trust than a long one showing everything except the number that matters.
Marketing attribution, honestly explained
Attribution is where marketing reporting most often stops being truthful, usually without anyone intending it.
Last-click gives full credit to the final touch. It is simple, it is what most tools default to, and it systematically overvalues bottom-of-funnel channels — particularly branded search, which frequently just intercepts demand created elsewhere.
Multi-touch models distribute credit across touchpoints. More sophisticated, but the distribution rules are assumptions, and presenting modelled numbers with two decimal places gives them a precision they have not earned.
Incrementality testing — turning a channel off in some regions and measuring the difference — is the only approach that answers the actual question, which is what would have happened anyway. It is also disruptive, slow, and requires a client willing to deliberately stop spending money for a period. Very few are, which is worth acknowledging rather than pretending otherwise.
Add to this a measurement environment that keeps degrading — cookie restrictions, iOS privacy changes, consent modes — and the honest position is that your numbers are directional. Say so. A client who learns from you that attribution is approximate trusts you more than one who discovers it independently and wonders what else was overstated.
A client reporting structure that works
Four sections, in this order, one page:
- What changed. The two or three numbers the client cares about, versus last period and versus target. No preamble.
- Why. Your interpretation. This is the section that justifies your fee and the one that cannot be automated — a number a client can get themselves; the reason behind it is what they are buying.
- What we are doing about it. Specific, owned, dated. Including when the answer is "nothing, this is noise" — a defensible reason to not act is a real contribution.
- What we need from you. Blockers, named plainly. Clients routinely under-appreciate that they are the constraint, and the record matters at renewal.
Anything else — channel breakdowns, creative performance, full funnel — goes in an appendix or the dashboard for the minority who want it. Reporting length is not correlated with reporting value; if anything the relationship runs the other way.
What to automate in marketing reporting
Report production splits into two very different activities that are usually conflated — a distinction that also governs agency client management more broadly.
Assembly — pulling numbers from six platforms, normalising date ranges, reconciling definitions, building charts — is repetitive, error-prone, and consumes an embarrassing share of skilled people's time. Automate all of it. This is the clearest return available in marketing operations.
Interpretation — deciding which movements matter, why they happened, and what to do — is the actual product. Automated commentary reliably produces the tell-tale sentence: "sessions decreased by 12.4% compared to the previous period", which restates the chart directly above it and tells the reader nothing.
Where automation does help interpretation is in flagging: anomaly detection that says "this is outside its normal range, look here". Direction of attention is a genuine contribution. Conclusions are not, yet.
Client reporting cadence: how often should you report?
Weekly written updates and a monthly or quarterly conversation is the pattern that works for most retained relationships, whether you are an agency or a professional services firm. The weekly note is short and often boring — that is the feature. Reliability is what stops clients wondering, and a client who has stopped wondering does not send "just checking in" emails.
The monthly call should then be about decisions rather than a recital of numbers the client already received. If your monthly call is spent reading the report aloud, the report is not doing its job and the call is not either.
One rule worth holding: bad news travels immediately, not on the reporting schedule. A client who learns about a problem three weeks late in a scheduled report will wonder what else is being timed for convenience. The relationship damage from delayed disclosure consistently exceeds the damage from the problem itself.
Frequently asked questions
How often should we report to clients?
A short written update weekly and a decision-focused conversation monthly suits most retained relationships. Lower-touch accounts can run monthly written and quarterly calls. Whatever cadence you choose, hold it — predictability is most of the value, and a skipped update is read as a problem being hidden.
Which marketing metrics actually matter?
The ones that change a decision when they move, which depends on the business. For most, that means cost per qualified lead or acquisition, conversion rate at the stages you can influence, and revenue or pipeline contribution. Impressions, reach and raw sessions are context at best.
Is GA4 enough on its own?
It handles web behaviour reasonably and is free, which matters. It is weaker on cross-channel attribution, offline conversions and anything involving a CRM-side outcome such as lead quality or closed revenue. Most teams end up combining it with CRM data to connect marketing activity to money.
How do we explain that attribution is imprecise without undermining ourselves?
Say it early, before there is a disagreement, and pair it with what you are confident about. 'Attribution is directional — here is what we know reliably, here is what we are inferring, here is the test that would settle it' is a position of expertise. The same explanation offered after a client challenges a number sounds like an excuse.
Should clients have live dashboard access?
Generally yes — withholding it reads badly, and self-service reduces status questions. But a dashboard is not a report. Give access, then keep sending the written interpretation, or you have simply moved the analytical work onto the client and they will notice what they are paying for.