The short version
- If a report leads with impressions, reach or rankings, it is describing activity rather than results.
- Three numbers matter to most businesses: cost per qualified lead, cost per customer, and return on spend. Everything else is diagnostic.
- A useful report says what changed, what was learned, and what happens next. Most say what happened.
- Ask for the numbers that would show the work was not going well. If they are absent, that is the finding.
Marketing reports have a reputation problem, and it is deserved. Too many are assembled to demonstrate effort rather than to inform a decision. They are long, they are green, and after reading one you still could not say whether to spend more next month.
Here is how to tell the difference, and what to ask for instead.
The metrics that flatter
None of these are useless. All of them are misleading when they lead the report.
Impressions and reach
How many times something could have been seen. Easy to grow by widening targeting, which usually makes performance worse. A rising impression count alongside flat enquiries is a sign of dilution, not growth.
Clicks and sessions
Better, because somebody did something. Still one step removed from anything commercial, and easily inflated by attracting the wrong people more cheaply.
Keyword rankings
Useful as a diagnostic, poor as a headline. Rankings vary by location, device and personalisation, and a position on a term nobody searches with intent to buy is worth nothing. Ranking well for high-volume, low-intent terms is a common and expensive trap.
Engagement rate and time on page
Genuinely ambiguous. A long time on your pricing page might mean careful consideration or total confusion. Never report these without a hypothesis attached.
Social followers
Almost never connected to revenue for a B2B business. Worth tracking if community is part of your strategy. Worth ignoring otherwise.
For every number in the report, ask: if this went up thirty per cent next month, would we do anything differently? If not, it does not belong on page one.
The metrics that matter
Cost per qualified lead
Not cost per enquiry. Qualified means somebody has judged the enquiry to be a genuine prospect. This requires a definition of qualified that sales and marketing both accept, and it is worth the argument it takes to agree one.
Cost per customer
The number the business actually cares about. It requires closing the loop between the marketing platforms and whatever system records sales, which is a piece of work in itself but changes every conversation once it exists.
Return on ad spend, or better, on margin
Revenue is flattering. If your margins vary between products or services, a campaign selling the low-margin one can look excellent and contribute little. Use margin where you can get it.
Pipeline contribution
For long sales cycles, the sale may be a year away. Reporting on qualified pipeline created gives you a leading indicator instead of waiting twelve months to find out.
Blended cost per acquisition
Total marketing cost, including fees, divided by total new customers. Crude, and it ignores attribution entirely, which is exactly why it is useful as a sense check. If the sophisticated numbers look wonderful and this one does not move, something is wrong with the sophisticated numbers.
What a report should actually contain
Four sections, in this order.
- The headline. Three or four commercial numbers against target and against last period. One page. Somebody should be able to read only this and know how things are going.
- What we did and what we learned. Changes made, in plain language, with what each one produced. Including the ones that did not work, because a report with no failures in it is not a report, it is a brochure.
- Diagnostics. The supporting detail. Channel breakdowns, search terms, page performance. This is where impressions and rankings legitimately live.
- What happens next. Specific actions, with owners and dates, including anything needed from you. This is the section that turns a report into a working document.
If your current report has section three and nothing else, you are receiving a data export rather than an analysis.
Not sure whether your reporting is telling you the truth? Send us a recent report. We will tell you what it is not showing you and what you should be asking for, with no obligation attached.
Questions to ask your agency
Reasonable questions, and the answers are informative regardless of what they are.
- What is our cost per customer by channel, and how confident are you in that figure?
- Which campaign is performing worst, and what are you doing about it?
- What did we try last quarter that did not work?
- What would you do differently if we increased the budget by half? What if we halved it?
- What is the single biggest constraint on our results right now, and is it something you control?
The last one is the most revealing. A good agency will often tell you the constraint is on your side, such as response times, the quality of the offer, or a broken CRM. That is uncomfortable and usually true.
On reporting honestly
A report that is green every month is not a sign of a well-run account. It is a sign that somebody is choosing the metrics after seeing the data.
Real marketing has bad months. Seasonality moves, competitors change their bids, a landing page test fails, a platform changes something without warning. An agency that reports those plainly, with what they did about it, is more valuable than one that never has a difficult conversation. The second kind is not doing less badly. They are just not telling you.
This is how we report on paid media and SEO, and it is built on the attribution work described in our guide to tracking which marketing generates revenue.
Common questions
How often should we get a report?
Monthly for most businesses, with a quarterly session that looks at strategy rather than numbers. Weekly reporting usually produces reaction to noise rather than decisions.
Our agency says attribution is impossible now. Is that true?
Perfect attribution is not achievable and never really was. Useful attribution is very achievable. If the conclusion drawn from privacy changes is that nothing can be measured, ask what they have tried.
Should we build our own dashboard?
If you have the data plumbing in place, a simple dashboard from your own sources is excellent, because it removes the question of who chose the numbers. It is not a substitute for someone explaining what the numbers mean.
What if we do not have enough volume for any of this to be significant?
Then report the raw counts honestly, avoid percentage changes on small numbers, and lengthen your reporting period. Low volume is a reason to be careful with statistics, not a reason to fall back on impressions.
