Experiment library

Growth experiments for the AI era

The old growth-hacking playbooks are full of tactics for platforms that no longer exist. These 30 experiments are built for how marketing works now, grouped by the four systems that turn attention into revenue.

Create demandGet in front of the right people before they need you.

ConvertTurn a visit into an enquiry, and a quote into a job.

AnswerRespond to every enquiry in minutes, at any hour.

Wake the databaseEarn again from people who already know you.

How to use the library

Pick one experiment from the system where you lose the most money (the cost per booked job calculator will tell you which). Write down the measure and the stop condition before you start, run it for the time shown, then keep it or kill it. Costs are rough ranges for a small Australian service business.

Create demand

Founder-voice clips at volume

Short clips of the owner explaining real jobs will bring enquiries more cheaply than polished brand video.

The old way: One produced brand video a quarter, edited by an agency.
The AI-era way: Film one hour with the owner. An AI editor cuts 20 clips, writes captions and five alternative hooks for each. Post daily and boost the two best each week.
Typical cost: $300 to $1,500 a month
Time to signal: 4 to 6 weeks
Measure: Cost per enquiry from video, against your other channels
Stop if: No clip beats your median view-through rate after 20 posts

Create demand

Answer-engine seeding

Publishing one original statistic about your trade will get you cited when people ask ChatGPT or Perplexity about it.

The old way: Blog posts written for keywords, hoping to reach page one.
The AI-era way: Run a small survey or analyse your own job data, publish the numbers on one clear page, then check every week whether ChatGPT, Perplexity and Gemini cite it.
Typical cost: Staff time, plus $0 to $500 for a survey
Time to signal: 6 to 10 weeks
Measure: Citations in AI answers, links earned and branded search growth
Stop if: No citations and no links after 10 weeks

Create demand

Suburb pages built from real jobs

Pages that describe genuine jobs in each suburb will rank and convert better than templated location pages.

The old way: One template copied across 50 suburb names.
The AI-era way: An AI drafts a page per suburb from completed job notes and photos (problem, fix, price range, time on site). A person checks every page before it goes live.
Typical cost: Staff time, about 20 minutes per page
Time to signal: 6 to 12 weeks
Measure: Impressions, calls and form fills per page
Stop if: Pages still not indexed after 8 weeks; merge them into fewer, stronger pages

Create demand

Thirty ad variants from one brief

Giving the ad platform many genuinely different creatives will lower cost per lead more than testing one headline at a time.

The old way: A/B test two headlines for a month.
The AI-era way: Generate 30 variants (angle, image, hook, offer) from one brief, launch them together and let delivery find winners. Retire losers weekly and generate replacements from the winners.
Typical cost: Your normal ad budget, plus tool fees under $100 a month
Time to signal: 2 to 4 weeks
Measure: Cost per lead against your current best ad
Stop if: Cost per lead isn't at least 15% better after spending twice your target cost per job

Create demand

Ads written in your customers' words

Ad copy built from phrases in your reviews will get more clicks than copy written by the business.

The old way: Brainstorm benefits in a meeting.
The AI-era way: Export your Google reviews, have an LLM group them into the top five reasons people chose you, and write ads that use customers' exact phrases.
Typical cost: Under $100
Time to signal: 2 to 3 weeks
Measure: Click-through rate and cost per lead
Stop if: No lift in click-through rate after 2,000 impressions per variant

Create demand

Find the themes competitors aren't advertising

There are services or problems your competitors don't advertise, where clicks are cheaper.

The old way: Guess from a keyword tool.
The AI-era way: Review competitors' live ads in Google's Ads Transparency Center, have an LLM summarise their themes, and launch small campaigns on the services nobody covers.
Typical cost: $500 to $1,000 test budget
Time to signal: 2 to 4 weeks
Measure: Cost per lead and impression share on the new themes
Stop if: Cost per lead on the new themes is worse than your account average

Create demand

Weekly founder point-of-view posts (B2B)

A clear weekly opinion from the founder will bring inbound conversations from buyers who already trust them.

The old way: A content calendar of company news.
The AI-era way: Record a five-minute voice note each week. An AI turns it into one post in the founder's voice, which they edit and publish themselves.
Typical cost: About 30 minutes a week
Time to signal: 8 to 12 weeks
Measure: Inbound messages and calls booked that mention the posts
Stop if: No buyer conversations after 12 posts

Create demand

Referral loop with adjacent trades

Businesses that serve the same customers just before or after you will refer work if asked properly.

The old way: Hand out business cards at networking events.
The AI-era way: List 30 adjacent businesses in your area (for an electrician: builders, solar installers, property managers). An AI drafts a personalised note to each; you follow up by phone.
Typical cost: Staff time only
Time to signal: 4 to 8 weeks
Measure: Referred jobs per month
Stop if: No referrals from 30 partners contacted twice

Convert

Instant price range on the quote form

Showing a likely price range before someone submits will bring more, and better qualified, enquiries.

The old way: "Call us for a quote."
The AI-era way: An estimator built from your past jobs returns a range as the visitor fills in the form, with what changes the price.
Typical cost: $1,000 to $5,000 to build
Time to signal: 3 to 6 weeks
Measure: Form completion rate and quote acceptance rate
Stop if: Completion rises but quote acceptance falls (you're attracting price shoppers)

Convert

A landing page headline for every search

A headline that matches what someone searched will convert better than one generic headline.

The old way: One landing page for every ad.
The AI-era way: An LLM writes a headline and first paragraph for each ad group's theme. The page swaps them in from the URL, so "emergency plumber" and "hot water repair" see different openings.
Typical cost: A day of development
Time to signal: 2 to 4 weeks
Measure: Landing page conversion rate by ad group
Stop if: No ad group improves after 200 visits each

Convert

Three-field form, questions after

Asking only for name, mobile and job up front will lift enquiries, and qualifying afterwards keeps quality.

The old way: A ten-field form that asks everything at once.
The AI-era way: Collect three fields, then an AI asks the remaining questions by SMS or chat straight after submission.
Typical cost: Under $500 to set up
Time to signal: 2 to 3 weeks
Measure: Form completion rate and the share of leads that get quoted
Stop if: Quoted leads fall by more than completions rise

Convert

The right review next to the button

Showing a review about the same service and suburb beside the call to action will lift conversion.

The old way: A star rating badge in the footer.
The AI-era way: Tag your reviews by service and suburb with an LLM, then show the closest match next to each page's main button.
Typical cost: A day of development
Time to signal: 3 to 4 weeks
Measure: Conversion rate on pages with matched reviews
Stop if: No lift after 500 visits

Convert

Website chat that books the job

A chat assistant that knows your services and prices, and ends by booking a slot, will convert visitors who won't call.

The old way: A chat widget that collects an email and replies the next day.
The AI-era way: An AI assistant trained on your services, service area and price ranges, connected to your calendar, so a chat ends in a booking or a call-back.
Typical cost: $100 to $500 a month
Time to signal: 3 to 4 weeks
Measure: Bookings from chat as a share of visitors
Stop if: Fewer than one booking per 500 chat sessions

Convert

Personalised quote follow-up

Following up every quote with messages that mention the actual job will raise the win rate.

The old way: Send the quote and wait.
The AI-era way: Day 2, 5 and 10 messages drafted by an AI from the quote and site notes ("the switchboard upgrade we discussed"), sent from a named person.
Typical cost: Under $200 a month in CRM automation
Time to signal: 4 to 6 weeks
Measure: Quote win rate
Stop if: Win rate unchanged after 50 quotes

Convert

Sixty-second video quote walkthrough

A short video from the technician explaining the quote will win more large jobs than a PDF alone.

The old way: Email a PDF quote.
The AI-era way: The technician records 60 seconds on their phone at the site. AI adds captions and a written summary, and the link goes out with the quote.
Typical cost: Staff time only
Time to signal: 4 to 8 weeks
Measure: Win rate on quotes above your average job value
Stop if: No lift after 30 video quotes

Convert

Weekly message-match check

Pages drift out of step with what people search; catching that weekly will keep conversion rates up.

The old way: A website review once a year.
The AI-era way: Each week, an LLM compares your top search terms with the page each one lands on and flags mismatches to fix.
Typical cost: Under $50 a month
Time to signal: Ongoing; first fixes within 2 weeks
Measure: Conversion rate on the pages you fixed
Stop if: Fixed pages show no change after a month

Answer

AI answers after-hours calls

Answering every after-hours call will turn up jobs that currently go to whoever answers first.

The old way: Voicemail, returned the next morning.
The AI-era way: A voice agent answers after 5pm and on weekends, books routine work into the calendar and sends urgent jobs to the on-call phone.
Typical cost: $150 to $600 a month
Time to signal: 4 weeks
Measure: After-hours jobs booked per month
Stop if: Fewer after-hours jobs than the agent costs, after two months

Answer

Five-minute follow-up on every form

Contacting web enquiries within five minutes will reach far more of them than replying within the hour.

The old way: Someone checks the inbox between jobs.
The AI-era way: An SMS from a named person within a minute of the form, then an AI-qualified call-back, with the lead passed to a human once qualified.
Typical cost: Under $200 a month
Time to signal: 2 to 3 weeks
Measure: Contact rate and median time to first contact
Stop if: Contact rate unchanged after 100 leads

Answer

Sixty-second callback button

Visitors who won't call will take a call from you if it happens straight away.

The old way: "We'll call you back within one business day."
The AI-era way: The visitor enters their number, and an AI voice agent calls within a minute to qualify the job and book a time.
Typical cost: $100 to $400 a month
Time to signal: 3 to 4 weeks
Measure: Callback requests and the share that book
Stop if: Fewer than 1% of visitors use it after a month

Answer

Missed-call text-back

Texting every missed caller within 30 seconds will save conversations that would otherwise go to a competitor.

The old way: Call back when you get a chance.
The AI-era way: Every missed call triggers an SMS with a booking link and an AI that can answer questions by text.
Typical cost: Under $100 a month
Time to signal: 2 weeks
Measure: Missed callers who reply or book
Stop if: Fewer than 10% of missed callers reply

Answer

AI qualifies, the owner closes

Letting an AI ask the first three questions will save the owner's time for the jobs worth winning.

The old way: The owner answers every call, including tyre-kickers.
The AI-era way: A voice or SMS agent asks about job type, timing and location, then sends hot leads straight to the owner's phone and books the rest.
Typical cost: $150 to $600 a month
Time to signal: 4 weeks
Measure: Owner hours per job won, and close rate on hot leads
Stop if: Close rate on hot leads falls

Answer

Mystery-shop your own business

You don't know how fast you really respond until you test it.

The old way: Assume the team is on it.
The AI-era way: Each month, send five test enquiries (form, call, after-hours call, email, social message) and log the response time to each.
Typical cost: One hour a month
Time to signal: Immediate
Measure: Median response time by channel
Stop if: Never. Keep running it

Answer

Weekend coverage trial

Weekend enquiries are a large, ignored share of demand.

The old way: Closed until Monday.
The AI-era way: Run an AI agent on Saturdays and Sundays only for four weeks and compare bookings with the four weeks before.
Typical cost: $100 to $300 for the trial
Time to signal: 4 weeks
Measure: Weekend bookings
Stop if: Weekend bookings don't cover the agent's cost

Wake the database

Revive 12-month-old quotes

Many people who didn't go ahead a year ago still have the problem.

The old way: Old quotes sit in a folder.
The AI-era way: An AI drafts a personal SMS for each quote from the original job notes. Send in batches of 50 so the team can handle replies.
Typical cost: Under $200 plus staff time
Time to signal: 1 to 2 weeks
Measure: Reply rate and revived revenue
Stop if: Reply rate under 2% across two batches

Wake the database

Review request when the job closes

Asking at the moment the job is marked complete will roughly double the number of reviews you get.

The old way: Ask for reviews when you remember.
The AI-era way: The CRM sends a request as soon as the job status changes to complete, with a reply flow that sends unhappy customers to the owner first.
Typical cost: Under $100 a month
Time to signal: 4 weeks
Measure: Reviews per completed job
Stop if: Fewer than one review per 10 jobs

Wake the database

Reminders based on asset age

Past customers with ageing equipment (hot water units, switchboards, air conditioners) are your warmest new jobs.

The old way: Wait for it to break and hope they call you.
The AI-era way: An AI reads past job records, estimates what was installed when, and drafts a service or replacement reminder for each customer as it ages.
Typical cost: Staff time plus messaging costs
Time to signal: 2 to 4 weeks
Measure: Jobs booked from reminders
Stop if: Under 1% of reminders book

Wake the database

Lapsed customer check-in

Customers who haven't used you in 18 months will come back if the message is about them, not a discount.

The old way: A newsletter with a 10% off code.
The AI-era way: An AI writes each lapsed customer a short note referencing their last job and the season, with one useful next step.
Typical cost: Under $200
Time to signal: 2 weeks
Measure: Reply rate and booked jobs
Stop if: No bookings from 200 messages

Wake the database

Ask for a referral right after a five-star review

People who have just praised you publicly are the most likely to refer someone.

The old way: A generic two-sided referral scheme on the website.
The AI-era way: When a five-star review lands, send a personal thank-you that quotes their review and asks who else they know who might need you.
Typical cost: Under $50 a month
Time to signal: 4 to 6 weeks
Measure: Referred enquiries per month
Stop if: No referrals from 30 asks

Wake the database

Seasonal and weather triggers

A message timed to the weather or season (storm season, the first hot week) gets far more replies than one sent at a random time.

The old way: The same email blast each quarter.
The AI-era way: Segment past customers by job type, then send a message an AI has drafted for each segment when the forecast or season makes it relevant.
Typical cost: Under $200 per campaign
Time to signal: 1 week after each trigger
Measure: Reply rate against your usual campaigns
Stop if: No better than your normal campaigns after two triggers

Wake the database

Clean the CRM before you campaign

A cleaned, enriched list will lift every reactivation campaign that follows.

The old way: Send to everything and accept the bounces.
The AI-era way: An AI removes duplicates, fills in missing suburbs and job types from old notes, and flags each contact's past spend before any campaign goes out.
Typical cost: Staff time, a few hours
Time to signal: Immediate
Measure: Share of contacts you can reach, and response rates on the next campaign
Stop if: Not applicable; do it once, then keep it clean

Retired tactics

Staples of the old growth playbooks that platform changes or AI have made obsolete, and what replaced them.

Retired	What replaced it
Gender-focused ad copy	Broad targeting with many creative variants. The platform finds the audience.
Manual A/B tests of one headline	Dozens of generated variants, with delivery picking the winners.
Gated PDF guides	Open, citable pages. AI answers summarise what's behind the gate anyway.
Periscope and Twitter Lead Cards	Short-form founder video and conversational lead capture.
"Call us during business hours"	An AI that answers every call, at any hour.
Exit-intent discount popups	A chat that answers the question that was about to make them leave.

Start with a diagnostic.

The 90-Day Plan: a diagnostic of your marketing and a week-by-week plan, built in two weeks. Yours to keep, whether or not we run it.

Get your 90-day plan Talk to us