- Decide what counts as a conversion before you look at any dashboard. Most disputes are definition problems.
- Track calls. In Madrid service sectors the phone is still where the money arrives, and untracked calls make organic look worthless.
- Cost per enquiry beats return on ad spend for local businesses, because revenue lands offline.
- Report the same numbers every month, even the ones that got worse.
Reporting is where most SEO and paid search relationships quietly break: not because results are bad, but because nobody agreed what would be counted. This is a measurement setup that holds up to a finance director's questions.
Step 1: define the conversions
Write down, before anything is configured, which actions count and what each is worth. For a typical Madrid service business that means: form submission, phone call over a minimum duration, WhatsApp or message thread started, direction request, and booking. Assign a rough value to each from your own close rate and average job size. Rough is fine; absent is not.
The minimum call duration matters more than it sounds. Counting every ring as a conversion inflates paid campaigns that generate misdials and wrong numbers.
Step 2: instrument the channels properly
- Analytics with server-side or consent-aware tagging, configured so organic, paid, direct and referral do not blur into each other.
- Call tracking with a number that swaps by source, or at minimum a distinct number on the paid landing pages.
- Search Console connected, for the query-level view that analytics cannot give you.
- Business Profile insights, exported monthly, since calls and direction requests from the profile never appear in site analytics at all.
- CRM or a simple enquiry log with the source recorded at intake, which is the only place the offline outcome exists.
Step 3: the metrics worth reporting
| Metric | Answers | Fails when |
|---|---|---|
| Enquiries by channel | Is anything arriving at all | Used alone: says nothing about quality |
| Cost per enquiry | Is it getting cheaper or dearer | Channels have very different lead quality |
| Qualified rate | Is the traffic the right traffic | Nobody in sales marks the CRM |
| Cost per customer | The only number the owner cares about | The cycle is longer than the reporting period |
| Assisted conversions | Whether a channel helps others convert | Treated as if it were a direct conversion |
| Impression share of the target cluster | Whether the ground is being taken | Confused with traffic. It is a leading signal |
- Enquiries by channel and district. The headline number, split so you can see where growth came from.
- Cost per enquiry by channel. Paid is straightforward; for SEO, divide the retainer plus content costs by enquiries attributed to organic. It will look bad in month two and good in month eight, which is the point.
- Assisted conversions. How often each channel appears in a path it did not close. This is what stops paid brand campaigns taking credit for organic work.
- Visibility, as a diagnostic. Impressions and average position by cluster, tracked from the district, to explain why the enquiry numbers moved.
- Close rate by source. The number that occasionally reveals that the cheapest enquiries are also the worst.
Attribution without the arguments
Every attribution model is a point of view, not a measurement. Knowing what each one is biased towards ends most of the argument:
| Model | Flatters | Use it when |
|---|---|---|
| Last click | Paid search and brand terms | Cycles are short and the budget question is simple |
| First click | SEO and social | You are asking what starts demand, not what closes it |
| Linear or position-based | Nothing in particular | Several channels are genuinely involved |
| Data-driven | Whatever the data supports | You have enough conversions for it to mean anything |
Pick one, state it in the report, and stop changing it. A channel that looks better under a different model has not actually improved.
Last-click is the default and it systematically over-credits whatever sits at the end of the path - usually branded paid search. Data-driven or position-based models spread the credit more fairly, but no model is true. The practical compromise: pick one model, use it consistently, and check it once a quarter against the enquiry log, which is the only ground truth you have.
Set the lookback window to match your sales cycle. Thirty days for reactive services, ninety for considered purchases. Businesses that leave it at the default and then argue about which channel works are usually arguing about a setting.
“If two channels are each credited with the same enquiry, the problem is not the channels. It is that nobody wrote down which model the report uses.”
What a good monthly report looks like
One page. Enquiries this month against the last three, split by channel and district. Cost per enquiry for each channel. What was delivered. What is planned. One paragraph explaining the biggest movement, up or down, with a cause rather than a shrug. Everything else belongs in an appendix nobody reads unless something looks wrong.
Frequently asked questions
Why does my analytics number never match my sales number?
Because consent banners, ad blockers, cross-device journeys and offline conversions all remove data. Expect a gap of 10-30% and reconcile against the enquiry log rather than trying to close it.
Is return on ad spend the right metric for a local business?
Rarely, because the revenue lands offline days later. Cost per qualified enquiry, plus a close rate you measure yourself, is more honest and easier to act on.
How long before SEO shows up in these numbers?
Should we pause ads once SEO is working?
Rarely, and never on the commercial terms. Ads and organic on the same query usually earn more together than either alone, and pausing removes the fastest signal you have about what converts.
Can we use paid search to test SEO ideas?
Yes, and it is one of the best uses of a small ad budget. A term that does not convert on ads will not convert on organic either, and you learn it in a fortnight instead of six months.
Local Pack effects in four to eight weeks, organic enquiry volume from month three to six. Cost per enquiry for SEO improves for as long as the work continues, which is the whole argument for it.
Get the measurement set up once, properly
Conversions defined, calls tracked, channels separated, one report you can take to a board meeting.
Talk to Madrid SEORelated: SEO services and Local SEO.