The bike shop spending €40 a day on the wrong search terms
A two-person bicycle repair shop runs Google Search ads at €40 a day on the single broad-match keyword 'bike'. Last week the campaign produced 55 clicks, cost €268 and zero enquiries. The search terms report shows 'bike racing games', 'bike wallpaper', 'cheap bike parts online' and 'bicycle repair near me'. All ads point to the shop's home page, which opens with a photo of the owners and a paragraph about their founding in 2011; the phone number is in small text in the footer. In parallel, they spent €100 boosting an Instagram post to a lookalike of their Page followers, reaching 31,000 people and generating 12 profile visits and no enquiries.
1. Where is the €368 going, and what would you change first? Prioritise the changes.
Read the worked answer
Where it goes: 55 clicks for €268 is about €4.87 a click - expensive for a repair shop, because broad match is bidding against gaming and wallpaper content. Three of the four visible search terms have no possibility of buying a repair.
Priority order - cheapest and fastest first:
- Conversion tracking. There is no way to know what an enquiry is worth or whether anything works. Calls and form submissions must be recorded before another euro is spent. Without this, every later decision is a guess.
- Negative keywords and match type. Add games, wallpaper, free, download, cheap parts, online as negatives; move from broad to phrase/exact on 'bicycle repair near me', 'bike service', 'bike repair'. This stops the bleed today at zero cost and is reversible.
- Location targeting. Restrict to a realistic catchment radius - a repair shop cannot serve someone 200 km away.
- Landing page. Send ads to a page with opening hours, prices, a map and a 'book a repair' button. A founder story with a footer phone number asks the visitor to hunt for the action.
- Ad copy mirroring repair intent rather than generic brand messaging.
Reasoning: the first two cost nothing and recover most of the waste; the landing page is where the remaining clicks convert or don't.
2. Both channels produced zero enquiries. Is that evidence they do not work for this shop? Explain.
Read the worked answer
No. Neither test was valid, so neither result tells you anything about the channel.
- Sample size: 55 clicks at a realistic 2–4% conversion rate predicts 1–2 enquiries. Zero is within normal noise. Concluding 'search does not work' from 55 clicks is like judging a coin after five tosses.
- The clicks were the wrong people: broad 'bike' served gamers and wallpaper browsers. The channel was not tested; the keyword was.
- The landing page was wrong: even a motivated visitor arrived at a founding story with the phone number in the footer.
- Instagram: a lookalike built from a tiny two-person shop's Page followers is a weak seed audience; 31,000 reach with no call to action and no tracked outcome is a branding impression, not a sales test.
The useful conclusion is not 'these channels fail' but 'these tests were not fair tests'. A fair test needs roughly 150–200 clicks per variant, correct intent targeting, a landing page with one obvious action, and conversion tracking. Only then does a zero mean something.
3. Rewrite the objective and targeting for the €100 Instagram spend, and state what you would measure.
Read the worked answer
Old version: objective = reach; audience = lookalike of Page followers; outcome measured = people reached. Reach is a delivery metric, not a business result, and the lookalike seed was too small to be meaningful.
New version
- Objective: enquiries - messages or calls. Select the messaging/engagement objective, not reach or awareness.
- Audience: a radius around the shop (say 8 km), since a repair is a local errand, optionally layered with cycling interests; exclude existing customers so the budget buys new demand. Drop the lookalike until the customer list is large enough to build a reliable one.
- Creative: one clear offer and one action - 'Punctures fixed while you wait, message us for a slot' - with the price and the location in the first line.
- Landing: the message thread itself, answered within an hour during opening hours; slow replies waste the spend.
Measure: cost per enquiry, plus enquiry-to-booking rate. Not reach, not profile visits. With €100, expect a small number of enquiries - enough to judge the cost per enquiry against the profit on a repair, not enough to judge the channel's ceiling.