Skip to content

Digital Marketing Foundations · AI-assisted marketing and your 90-day plan · lesson 15 of 15 · 16 min

Your 90-day marketing plan: from diagnosis to decisions

Why 90 days

A year is too long to plan in detail – platforms, costs and your own data will change. A week is too short to learn anything. Ninety days is long enough to run several proper tests and short enough to stay focused. This lesson pulls every module of the course into one practical plan.

The seven parts of the plan

  1. Diagnosis – where you are now: funnel numbers, biggest leak, unit economics, channel performance, tracking health.
  2. Goal – one business outcome with a number and a date ("120 new customers a month by 31 December at nCAC ≤ £35").
  3. Customer and positioning – your one-card persona and positioning statement.
  4. Three bets – the few big moves most likely to reach the goal (for example "fix checkout completion", "launch creator-led TikTok", "build a welcome email flow").
  5. Experiments backlog – specific tests under each bet, prioritised.
  6. Budget and calendar – money and weeks allocated, including learning budgets.
  7. Measurement and decision rules – dashboard, UTM plan, review rhythm, stop and scale rules.

Step 1: diagnose with numbers you already have

Pull the last 90 days:

  • Funnel: landing views → product views → add to cart → checkout → purchase (or enquiries → calls → clients).
  • Unit economics: CAC or nCAC, contribution margin, break-even ROAS, payback.
  • Channels: spend, revenue (from the store), CPA by channel.
  • Tracking health: reconciliation ratios, share of Unassigned traffic, consent audit result.

Write three sentences: the biggest leak, the most efficient channel and the biggest measurement gap.

Step 2: prioritise experiments with ICE

For each test idea, score 1–10 for Impact (if it works, how much will it move the goal?), Confidence (how much evidence do we have?) and Ease (time and cost). In Google Sheets, with scores in B, C and D:

ICE score:  =ROUND((B2*C2*D2)^(1/3), 1)
Rank:       =RANK(E2, $E$2:$E$30)

(The cube root keeps the score on the same 1–10 scale.) ICE is a conversation tool, not truth – its value is making assumptions explicit.

| Test | Bet | Impact | Confidence | Ease | ICE | |---|---|---|---|---|---| | Show delivery fee on product page | Fix checkout | 7 | 8 | 9 | 8.0 | | Guest checkout | Fix checkout | 8 | 7 | 6 | 7.0 | | 3 creator videos, problem-first hook | TikTok | 8 | 5 | 6 | 6.2 | | Welcome flow (5 emails) | Email | 6 | 7 | 7 | 6.6 | | New brand video ad | Awareness | 5 | 3 | 3 | 3.6 |

Step 3: write each test as a hypothesis

Because [evidence], we believe [change] will cause [metric] to move from [x] to [y]
for [audience]. We will run it for [time/spend] and call it if [decision rule].

Example: "Because 40% of checkouts complete and session recordings show exits on the fee screen, we believe showing delivery costs on the product page will lift checkout completion from 40% to 50% for mobile visitors. We will run it for two weeks and keep it if completion rises by at least five points with no drop in average order value."

Step 4: the 13-week calendar

| Weeks | Focus | |---|---| | 1–2 | Fix tracking (UTM plan, consent audit, test purchase), baseline dashboard, quick wins | | 3–6 | Run top-ranked tests for bets 1 and 2; weekly review | | 7–8 | Scale winners (20–30% budget steps), stop losers, launch next tests | | 9–12 | Bet 3 and second-round tests; build owned assets (email flows, WhatsApp opt-in) | | 13 | Review against goal; write the next 90-day plan |

Step 5: decision rules and rhythm

  • Weekly (30 minutes): dashboard, what/why/next summary, test status.
  • Monthly: unit economics, reconciliation, budget reallocation.
  • Rules: stop-loss per test, scale thresholds, "no changes to a test before its minimum run".

Using AI to pressure-test your plan

After drafting, ask an assistant to be your critic:

Act as a sceptical head of marketing. Here is my 90-day plan (below) and my
dashboard numbers. 1) List the five weakest assumptions and what evidence would
test each. 2) Check every calculation and flag errors. 3) Identify any compliance
risks (consent, disclosure, claims). 4) Suggest one test I have missed.
Do not rewrite the plan. [paste plan with no personal data]

Worked example: a Karachi meal-prep service (illustrative)

  • Diagnosis: 2,000 trial sign-ups a month but only 18% convert to weekly plans; Instagram drives most sign-ups; WhatsApp replies average two hours; no welcome flow; UTMs inconsistent.
  • Goal: 500 active weekly subscribers by the end of the quarter at CAC ≤ PKR 3,000.
  • Bets: (1) trial-to-plan conversion – WhatsApp quick replies and a 5-message onboarding sequence for opted-in trials; (2) creator-led Reels with the "lunch sorted on 12-hour shifts" angle; (3) referral offer for existing subscribers.
  • Top tests: reply within 10 minutes with quick replies (ICE 8.3); day-3 "how was your first meal?" message (7.4); two nurse and banker creators (6.5).
  • Rules: scale a creator when their sign-up CPA is 20% under target for two weeks; stop any test after 2× target CPA with no conversions.

By week 8, trial-to-plan conversion has improved, creator content outperforms brand-made ads, and the referral test is scheduled for weeks 9–12.

Common mistakes

  • Ten bets instead of three.
  • Tests with no hypothesis, metric or stop rule.
  • Skipping tracking fixes in weeks 1–2, so no result can be trusted.
  • Never writing the week-13 review.

How to measure success

At day 90 you should be able to say: did we hit the goal; which bets worked and why; what we learned from failed tests; and what the next 90 days will focus on. Where to next: Paid Social Advertising and Email Marketing and Automation go deeper on the most common bets, and AI Performance Marketing and Privacy-First Measurement take you to advanced practice.

Video lecture: Your 90-day plan: turning the course into action

Lecture coming soon · 11 chapters · about 8 minutes. Read the full transcript below.

  1. Your 90-day plan
  2. Why 90 days
  3. Seven parts
  4. Step 1: diagnose
  5. Step 2: prioritise with ICE
  6. Step 3: hypothesis template
  7. Step 4: 13-week calendar
  8. Example: Karachi meal-prep (illustrative)
  9. Tests, rules, results
  10. Pressure-test and avoid
  11. Course recap and try this now

Lecture transcript

Your 90-day plan

You've now got all the pieces. How people buy, research, positioning, channels, budgets, metrics, dashboards, tracking, consent and AI. But pieces don't grow a business. A plan does. And most marketing plans fail in one of two ways. They're either a vague wish, like grow the brand this year, or a list of fifty ideas with no priorities. In this final lecture, you'll learn to build a ninety-day plan with seven parts, prioritise experiments with a simple scoring method, write every test as a hypothesis with a decision rule, and lay it out on a thirteen-week calendar. By the end, you'll have a plan you can start on Monday.

Why 90 days

Why ninety days? A year is too long to plan in detail. Platforms, costs and your own data will change long before December. A week is too short to learn anything reliable. Ninety days is the sweet spot. Long enough to run several proper tests, short enough to stay focused. Think of it like a football season split into thirds. You set a target, play the matches, review the tape, and adjust your tactics before the next third. You don't redesign the team after every game, and you don't wait until the end of the season to notice you're losing.

Seven parts

Here are the seven parts. One, diagnosis: where you are now, in numbers. Two, a goal: one outcome, with a number and a date. Three, your customer and positioning: the one-card persona and your positioning statement. Four, three bets: the few big moves most likely to hit the goal. Not ten. Three. Five, an experiments backlog: specific tests under each bet, ranked. Six, budget and calendar: money and weeks, including learning budgets. And seven, measurement and decision rules: your dashboard, UTM plan, review rhythm, and when to stop or scale. Every one of those comes straight from a lesson you've already done.

Step 1: diagnose

Start with diagnosis, using numbers you already have. Pull the last ninety days of your funnel, from landing views to purchases, or from enquiries to clients. Pull your unit economics: new-customer CAC, contribution margin, break-even ROAS and payback. Pull channel results using store revenue. And check your tracking health: reconciliation ratios, how much traffic lands in Unassigned, and your consent audit. Then write just three sentences. The biggest leak in the funnel. The most efficient channel. And the biggest measurement gap. Those three sentences will shape everything else, and they stop you from starting with whatever idea is most exciting.

Step 2: prioritise with ICE

Next, prioritise. You'll have more ideas than time, so score each test idea from one to ten on three things. Impact: if it works, how much will it move the goal? Confidence: how much evidence do we have that it'll work? And Ease: how cheap and quick is it? That's called ICE. In Google Sheets, one formula combines the three, and the formula is in the lesson text. Here's the important bit. ICE isn't the truth. It's a conversation tool. Its real value is making assumptions visible. When someone scores confidence as eight, you get to ask: what's the evidence?

Step 3: hypothesis template

Then write each test as a hypothesis. Because of this evidence, we believe this change will move this metric from x to y for this audience. We'll run it for this long, and call it if this rule is met. Let's do a simple worked example from the candle brand we met in the first lesson. Because only forty percent of checkouts complete, and recordings show people leaving on the fee screen, we believe showing delivery costs on the product page will lift checkout completion from forty to fifty percent for mobile visitors. We'll run it for two weeks and keep it if completion rises at least five points with no drop in order value.

Step 4: 13-week calendar

Now lay it out on a thirteen-week calendar. Weeks one and two: fix tracking. Your UTM plan, consent audit and a test purchase, plus your baseline dashboard and quick wins. Weeks three to six: run your top-ranked tests for bets one and two, with a weekly review. Weeks seven and eight: scale winners in steps of twenty to thirty percent, stop losers, and launch the next tests. Weeks nine to twelve: bet three and second-round tests, plus building owned assets like email flows and WhatsApp opt-ins. And week thirteen: review against the goal and write the next ninety-day plan. Why fix tracking first? Because otherwise, no result you see can be trusted.

Example: Karachi meal-prep (illustrative)

Now a realistic business scenario, with illustrative numbers. A meal-prep service in Karachi gets two thousand trial sign-ups a month, but only eighteen percent convert to weekly plans. Instagram drives most sign-ups. WhatsApp replies take two hours on average. There's no welcome flow, and UTMs are a mess. Goal: five hundred active weekly subscribers by the end of the quarter, at a CAC of three thousand rupees or less. Three bets. One: trial-to-plan conversion, with WhatsApp quick replies and an onboarding sequence for opted-in trials. Two: creator-led Reels using the lunch sorted on twelve-hour shifts angle. Three: a referral offer for existing subscribers.

Tests, rules, results

Their top tests by ICE: replying within ten minutes using quick replies, a day-three how was your first meal message, and two creators, a nurse and a banker, who match their best customers. Decision rules: scale a creator when their sign-up CPA is twenty percent under target for two weeks, and stop any test after twice the target CPA with no conversions. By week eight, trial-to-plan conversion has improved, creator content is beating brand-made ads, and the referral test is scheduled for weeks nine to twelve. Notice what made it work: three bets, ranked tests, clear rules, and tracking fixed first.

Pressure-test and avoid

Before you commit, pressure-test the plan with AI. Paste your plan and dashboard numbers, with no personal data, and ask an assistant to act as a sceptical head of marketing. Ask it to list the five weakest assumptions and what evidence would test each, to check every calculation, to flag compliance risks like consent, disclosure and claims, and to suggest one test you missed. Tell it not to rewrite the plan. You want a critic, not a ghostwriter. Then watch the classic mistakes. Ten bets instead of three. Tests with no hypothesis, metric or stop rule. Skipping the tracking fixes. And never writing the week-thirteen review.

Course recap and try this now

Let's recap the whole course in one breath. Understand how people buy and what they're trying to achieve. Position your offer clearly. Pick channels by job. Budget backwards from a goal and check it against your margins. Measure with the right formulas, report what, why and next, track with a clean UTM plan, respect consent, and use AI honestly. Your ninety-day plan turns all of that into three bets, ranked tests and clear rules. Here's your try this now, and it's the big one. Write your own ninety-day plan using the seven parts and the templates in the lesson text. Then continue to Paid Social Advertising or Email Marketing to go deeper on your biggest bet.

Key takeaways

  • A 90-day plan has seven parts: diagnosis, goal, customer and positioning, three bets, experiments, budget and calendar, measurement and rules.
  • Diagnose with real funnel, unit-economics, channel and tracking data before choosing bets.
  • Prioritise tests with ICE, write each as a hypothesis with a decision rule, and fix tracking first.
  • Review weekly, reallocate monthly and write a week-13 review that feeds the next plan.

Try it

Write your own 90-day plan using the seven parts: diagnosis sentences, one goal, three bets, an ICE-ranked backlog of at least six tests with hypotheses, a 13-week calendar and your decision rules.