IT

Google Ads for startups and Meta Ads, one test at a time

The board

How do you test ads without burning the budget?

By writing every campaign as an experiment: a hypothesis, an audience, a metric and a stopping rule, all set before spending. Here's the board we start from with a startup; filter it by channel and goal. If the idea still needs validating, start with the "Validate" cards.

Experiments showing: 8 of 8

  • Google · SearchValidate

    Do people searching for this actually want it?

    Hypothesis
    People searching the category click an ad built on our main benefit and leave their email.
    Audience
    High-intent searches, exact and phrase match.
    Metric
    Waitlist sign-ups, not clicks.
    Stop
    Test budget spent with no sign-ups: change the promise, don't raise the spend.

  • Meta · Feed and ReelsValidate

    Which promise stops the scroll?

    Hypothesis
    Of three angles (time, mistakes, control), one brings sign-ups at a clearly lower cost.
    Audience
    One broad audience, equal spend per angle.
    Metric
    Cost per sign-up for each angle, at equal spend.
    Stop
    One angle clearly ahead, or budget gone.

  • Google · SearchLeads and demos

    Does the problem sell better than the category?

    Hypothesis
    People searching the problem, not the product, book more demos: there's less competition there.
    Audience
    Keywords phrased as problems or questions, one landing page per group.
    Metric
    Demos booked, and how many become customers in the CRM.
    Stop
    Planned clicks with no demos: rewrite the landing page.

  • Meta · Instant formLeads and demos

    In-platform form or landing page?

    Hypothesis
    The instant form brings more leads; the landing page brings better ones.
    Audience
    The same audience split in two.
    Metric
    Qualified leads, not total leads.
    Stop
    Enough leads to compare their quality in the CRM.

  • Google · App campaignUsers

    Do installs turn into users?

    Hypothesis
    Optimising for completed onboarding brings users who stay, not just downloads.
    Audience
    Chosen by the platform from the event.
    Metric
    Cost per completed onboarding.
    Stop
    Event too rare for the algorithm: switch to a more frequent step.

  • Meta · Lookalike audienceUsers

    Do your first customers point to the next ones?

    Hypothesis
    A lookalike of the first customers activates better than an interest-based audience.
    Audience
    Lookalike of the customer list, against interests.
    Metric
    Cost per activated user.
    Stop
    Customer list too small: postpone the test.

  • Google · Display and YouTubeWin back

    Do pricing-page visitors come back?

    Hypothesis
    People who viewed pricing without converting return when they meet an answer to their objection.
    Audience
    Pricing-page visitors, customers excluded.
    Metric
    Conversions from returning visitors.
    Stop
    High frequency, no conversions.

  • Meta · StoriesWin back

    Do quick answers close the sign-up?

    Hypothesis
    People who left sign-up halfway finish it after short answers to their doubts.
    Audience
    Sign-ups started but not finished, last few days.
    Metric
    Completed sign-ups from this audience.
    Stop
    Frequency rising with no sign-ups.

Two channels

Google Ads or Meta Ads for a startup?

Google Ads catches people already searching for a solution; Meta Ads puts the product in front of people who don't yet know they need it. Many B2B startups begin with Google, and new or highly visual products with Meta; most end up needing both, at different moments.

Google Ads and Meta Ads for a startup: intent, formats, audience, when each works
Google AdsMeta Ads
IntentCaptures demand that already exists: the person is searchingCreates demand: the person is scrolling, not searching
FormatsSearch ads, Shopping, YouTube, Display, app campaignsImages, video and carousels in the feed, Stories and Reels on Facebook and Instagram, instant forms
AudienceKeywords and search intent, plus audience signalsInterests, behaviours, lookalikes of your customers
When it worksPeople already search for the solution, or the problem, in clear termsThe product is new or explains itself in seconds, and the audience is easy to define
What you need firstA landing page per group of searches, conversions trackedPlenty of creative to rotate, Pixel and Conversions API set up
First pitfallPaying for clicks on searches that are too broadOptimising for clicks and likes instead of customers

The cycle

How does an ad testing cycle work?

Five steps on repeat: hypothesis, tracking, small launch, reading, decision. Every round should teach you something, even when the test fails: it's the method behind our growth marketing.

  1. Hypothesis

    One sentence: what changes, for whom, and which number we expect to move. If it can't be written that way, it isn't a test yet.

  2. Tracking

    Before spending, we check the conversion really gets recorded, from landing page to platform to CRM.

  3. Small launch

    Budget and duration agreed upfront, one variable at a time, everything else held still.

  4. Reading

    What counts is the chosen metric and the leads in the CRM, not clicks and likes.

  5. Decision

    Scale, change or stop, and write down what we learned: the next round starts from there.

Tracking

How do you track conversions for a startup?

Decide first which action counts (a demo booked, a sign-up, a purchase), record it with the Google tag and with Meta's Pixel and Conversions API, and connect the CRM so you know which leads become customers.

  1. Ad

    The click carries its campaign and message.

  2. Landing page

    One page per promise, fast, with a single call to action.

  3. Event

    The action that matters gets recorded, respecting cookie consent.

  4. Platform

    Google and Meta receive the conversion and learn who to look for.

  5. CRM

    The lead becomes a customer, or not: that's what sets the budget.

Google Ads campaigns and tracking are run by My Web Lab, the group's web agency, the same team that builds the website: landing pages and measurement are built together.

If tracking is wrong, everything is wrong: platforms optimise for the signal they receive. An event counted twice or a form that doesn't fire sends spend in the wrong direction.

Budget

How much budget does a startup need for Google Ads?

There's no single right figure. A test budget depends on what a click costs in your market, how many clicks it takes to get a conversion and how many conversions you need to make a call. We work it out with you, starting from what you can afford to spend on learning.

Test budget≈cost per click÷conversion rate×conversions to decide

Cost per click
Set by the auction: crowded markets and very commercial keywords cost more. Estimate it before you start, with Google's Keyword Planner.
Conversion rate
How many landing-page visitors take the action. It depends on the promise, the page and how clear the offer is.
Conversions to decide
How many you need to tell a result from noise. With too few, a good ad can look bad, and the other way round.
A glass balance scale, level: a few frosted cubes on one side, a single lilac crystal on the other

Also in the mix

  • How many experiments run at once
  • The length of your sales cycle
  • What a customer is worth over time, and the margin

Questions about Google Ads and Meta Ads

How long does it take to know whether a campaign works?

It depends on how many conversions come in, not on the number of days: with few leads a day, it takes longer to read a result. That's why duration, budget and the stopping rule are set before launch, and nothing changes mid-test.

Do we need a full website or is a landing page enough?

For a test, one landing page per promise is enough: fast, with a single call to action. As campaigns grow you need a site that handles the traffic and builds trust; more on our startup website page.

Should we start with ads or with SEO?

Ads give you answers in days; SEO builds, over months, traffic you don't pay for per click. Many startups begin with small paid tests and use what they learn to choose their content; see SEO for startups.

How much does campaign management cost?

It depends on the channels, the number of campaigns and languages, how much has to be built (landing pages, tracking, creative) and how often you test. We don't publish a price list: the factors are on our pricing and timelines page.

Can you guarantee a cost per customer?

No: it depends on the auction, the market and the product, and nobody fully controls those. What we can promise is the method: clear hypotheses, verified tracking, decisions made on the numbers.

Further reading

Your first test fits in one line. Let's write it.

We reply within one business day.

Let's talk