You met the customer funnel in Module 3, with a goal for each stage: get noticed, get considered, get chosen, get chosen again. Now you turn those goals into numbers.

Objectives come first. A KPI (key performance indicator) is one of the few measures that show whether an objective is being met. So start with the objective, not the data. A useful objective is SMART: specific, measurable, achievable, relevant and time-bound. "Grow the business" is a wish. "Grow repeat bookings from 30% to 35% of all bookings by 31 March next year" is an objective you can measure.

The KPI tree. Draw it from the top down:

1. The business goal, such as more revenue, more members or lower costs.
2. The marketing objectives that drive it, such as more first-time customers or fewer customers leaving.
3. The KPIs that show each objective moving.

Every KPI should link upwards to an objective. If it doesn't, it may be interesting, but it isn't a KPI for this goal.

One KPI per funnel stage. Pick one measure for each stage you care about. Awareness might be visits from new users. Consideration might be enquiries or product page views. Purchase might be the conversion rate. Loyalty might be the repeat rate or the churn rate.

Leading and lagging. A lagging indicator tells you what has already happened, such as revenue or customers lost. A leading indicator moves first and gives you warning, such as sign-ups started this week. You need both: lagging to judge results, leading to act in time.

Vanity metrics. A vanity metric looks impressive but doesn't show whether an objective is being met: followers, likes, impressions, total page views. Ask: "If this number doubled, what would we do differently?" If the answer is "nothing", leave it off.

UK example: the GOV.UK Service Manual. Its page "Data you must publish" (last updated 6 April 2021) tells government service teams they must publish data on four mandatory KPIs: cost per transaction, user satisfaction, completion rate and digital take-up. Four numbers cover what a service costs, whether users are happy, whether they finish the task and whether they choose the online route. That's the lesson for you: a small, balanced set, each tied to what the service is for.

You've chosen your KPIs. Now find out where each number comes from, and what could make it wrong.

How web analytics works. A small piece of code, called a tag, sits on each page of a website. When something happens, such as a page loading, a click, a sign-up or a purchase, the tag sends a record to the analytics tool. Each record is an event. Events from one visit are grouped into a session, which ends after a period of inactivity. A conversion is an event you've marked as one that matters. Google Analytics 4 (GA4) is one widely used example. It is built around events, and it now calls the events you mark as important "key events"; Google Ads still calls them conversions. Menus and names change, so learn the ideas, not the buttons.

Web analytics isn't the only source. Sales usually come from the shop or booking system, subscribers from the subscription system, opens from the email platform, and reach from each social platform's insights. For money, the order system is usually more reliable than analytics.

Campaign tags. UTM parameters are labels added to the end of a link you share: utm_source (who sent it, such as newsletter), utm_medium (the kind of channel, such as email) and utm_campaign (which campaign). Google's help page says the values are case sensitive, so "Email" and "email" show as two lines. Agree one naming list. Untagged links often show up as direct visits, and the campaign loses the credit.

Consent. You met cookie consent in Module 5; now see what it does to your numbers. The ICO's guidance on storage and access technologies, updated after the Data (Use and Access) Act 2025, describes an exception for analytics used only to collect statistics to improve your own service: visitors must get clear information and an easy way to object. If the same data is used for advertising, consent is needed. Either way, visits from people who decline or object are usually missing or only estimated, so analytics usually undercounts. This isn't legal advice: check the ICO's current guidance.

Data quality problems to look for: pages with missing tags; bots and your own staff's visits; consent gaps; test or duplicate orders; and a definition that changed partway through the year.

AI activity. Give an AI assistant a simple chart or table with no personal data, such as twelve months of website visits, and ask it why a number changed. Then check each reason it gives against what else you know: campaigns you ran, the calendar, website changes, tracking changes. Mark each reason confirmed, possible or guessed. The AI can't see your calendar, so it often fills the gap with a confident guess.

A baseline is the value of a measure before you change anything. It's the starting point for every target and the yardstick for every result. Read the worked example later in this topic before you start: it sets Larkfield's baselines and shows the "from the outside" method.

Start with last year. The best baseline is your organisation's own figures for the last full year, or the same period you'll later judge results on. Three rules keep it honest:

• Same period: compare like with like, a year with a year or a quarter with the same quarter.
• Same definition: if "active subscribers" means the average over the year, keep it that way; don't switch to the year-end figure.
• Check it adds up: orders times average order value should equal revenue, and revenue should match the accounts.

Seasonality. Many organisations have busy and quiet months. A garden centre peaks in spring; Larkfield's gift buying is concentrated before Christmas. So compare a campaign month with the same month last year, not the month before. Otherwise the season gets the credit, or the blame, for your marketing.

When there's no data. Many learners can't see internal figures, and that's fine. Estimate from public information instead: a free traffic-estimate tool, the number of public reviews, follower counts on each platform, or accounts filed at Companies House (larger companies show turnover; many small ones file shorter accounts without it). Write "estimate" beside the figure, with the source, the date and why you trust it. An honest estimate is useful. An unlabelled guess is dangerous, because the next person will treat it as fact.

How reliable is it? Use the same scale as your topic 1 source log:

• High: your organisation's own records, checked against another source.
• Medium: your own analytics with known gaps, or a well-documented public source.
• Low: outside tool estimates and figures worked out from clues.

A flat baseline matters too. If a number has been steady, that tells you what would probably happen if you did nothing different. You'll need that when you judge whether a campaign really worked, later in the course.

A dashboard is not a report of everything you can count. It's a single page that helps someone decide what to do next. Most dashboards fail because they show too much and nobody knows what any number is for.

Keep it short. Four to eight KPIs on one page, taken from your KPI tree. If someone asks for a number to be added, ask which decision it will inform.

Give each KPI five things:

1. An owner: the person or role who explains why it moved and acts on it. If you work alone, the owner can be "me" or a role such as "marketing lead".
2. A baseline: where it started, from Lesson 6.3.
3. A target: where the objective says it should be, and by when.
4. A rhythm: how often it's checked. Leading indicators might be weekly; lagging ones monthly or quarterly.
5. A decision: write it as "If this falls below X, we will...". If you can't finish the sentence, question whether the KPI belongs on the page.

Show change, not just numbers. "2,400 members" means little on its own. "2,400, up 150 on the baseline of 2,250, target 2,500" tells you where you are. Where there's a season, show the same period last year beside it.

Colours need rules. Red, amber and green are quick to read, but only if you've decided in advance what each colour means. Otherwise people argue about colours instead of acting.

Sketch it first. A pencil sketch or a simple table is enough for this task. Later you might build it in a spreadsheet or a free dashboard tool such as Looker Studio, but the thinking is the same. Test the sketch with one question: could someone pick it up on a Monday morning and know what to do?

You've now built every piece of your Foundation stage work. A strategic audit puts it together: a structured review of where your organisation stands now, ending with the main issues to act on. It's Milestone 1, and the Practitioner stage starts from it. Read the worked example later in this topic first: it shows Maya Collins's audit for Larkfield.

The sections, each drawn from earlier work:

Summarise, don't paste. Each section needs two or three sentences and the evidence behind them. Keep it in the present tense: this is where the organisation is, not what you plan. Then name your weakest evidence. Every audit has some, and saying so is a strength.

Choosing three issues. An issue is a problem or an opportunity that matters to the business goal. Test each candidate:

• Does it link to a KPI, so you'll know if it improves?
• Is there evidence for it, or only a hunch?
• Can marketing act on it?

Leadership: make the call. This is your first leadership moment in the course. You'll rarely have perfect evidence, and a list of fifteen problems isn't a decision. Choose three. For each, say why it matters, what the evidence is and what would change your mind. If someone more senior disagrees, your evidence is what you bring to the conversation.

Your three issues become the starting point for your campaign plan's objectives at Milestone 2, and your baselines are what you'll judge results against at the end of the course.