How to read your own analytics in fifteen minutes

There is no value in reading reports built on broken measurement, and broken measurement is common enough that this check comes first. It takes four minutes and it determines whether anything below is worth doing.

1. 1.Submit a test enquiry through your own form and watch it register as a conversion. If it does not, stop here — everything downstream is wrong.
2. 2.Check whether one enquiry produces one conversion or two. Double-counting is the most common tracking fault and it makes every cost figure look half what it is.
3. 3.Compare last month’s conversions against your inbox or CRM. They will not match exactly and they should be close. A large gap is a tracking problem, not a marketing one.
4. 4.Check whether phone calls are counted at all. For many businesses they are the majority of enquiries and they appear nowhere.

Everything a business owner actually needs is in these four. Set each to compare against the same period last year rather than last month, because month-on-month comparison mostly measures how many working days a month had.

The fourth is in Google Ads rather than analytics, and it is the one most business owners have never opened. It is also the most immediately useful, because it is the only report that shows waste in plain language rather than as a number.

Each of these appears prominently, moves visibly, and tells you very little. Watching them produces confident decisions in arbitrary directions.

• Sessions. Traffic is an input, not an outcome, and it is trivially increased by attracting people who will never buy. A month with fewer sessions and more enquiries is a better month, and a report leading with traffic is usually leading with it for a reason.
• Bounce rate. It has always been a poor signal and in GA4 it is derived from engagement rather than measured directly. A page answering a question completely, so the visitor leaves satisfied, records the same thing as a page nobody could use.
• Average position. It averages across every query, device and location you appear for, which produces a number that moves for reasons unrelated to anything you did. Position for the specific phrases you care about is meaningful; the average is not.

If the four reports above look like nonsense, it is usually not because analytics is hard. It is because four settings are at their defaults, and each one distorts the picture in a way that is invisible unless you know to look for it.

• Internal traffic is not excluded. Your own team, and your agency, browsing the site daily. On a low-traffic site this can be a substantial share of everything recorded, and it inflates exactly the pages your team looks at rather than the ones customers do.
• Payment and booking domains appear as referrals. A visitor who goes to a payment provider and returns is recorded as a new visit from that provider, which quietly credits your checkout with acquiring the customer it just took payment from.
• Conversions are not defined against real business events. If the only conversion is a thank-you pageview, anybody who reloads that page counts twice, and any enquiry arriving by phone counts not at all.
• Data retention is at the default. GA4 keeps event data for two months unless changed, extending to fourteen, and the change is not retrospective. It costs nothing to fix and the data lost before you fix it is gone permanently.

All four take an afternoon between them, and none requires an agency. They are worth doing before drawing any conclusion from a report, because each one produces confident numbers that happen to be wrong.

A sequence rather than a dashboard. Doing it in this order means each step provides context for the next, and the whole thing fits comfortably into a quarter of an hour.

1. 1.Three minutes: conversions this month against the same month last year. One number, one comparison. This is the headline and nothing else is.
2. 2.Three minutes: conversions by channel, same comparison. Which channels moved, and did anything disappear entirely?
3. 3.Four minutes: the search terms report, sorted by cost. Read the top thirty as text. Anything obviously irrelevant is money that left the building.
4. 4.Three minutes: landing pages by entrances, and check the conversion rate of the top five. A high-traffic page converting poorly is the most valuable thing on this list.
5. 5.Two minutes: write down one question this raised. Take it to whoever manages the marketing. That question is worth more than the fourteen minutes that produced it.

Analytics has real limits and being clear about them prevents decisions that the data does not actually support.

It cannot see people who declined cookies, which is a meaningful and variable proportion of visitors. It cannot see the podcast, the recommendation or the conversation that made somebody search your name, so branded search takes credit for demand created elsewhere. It cannot follow somebody who researched on their phone and bought on a laptop a week later. And it attributes to the last click, which systematically over-credits remarketing and branded search while under-crediting whatever created the demand.

None of that makes analytics useless. It makes it one input among several, best read alongside the self-reported field on your enquiry form and what your sales team actually hears on calls. A decision supported by all three is safe; one supported only by an attribution model usually is not.