Restaurant ads audience drift after first week?

ranktracker

I've seen this exact pattern with half a dozen restaurant clients. The first week crushes it - right demos, right local area, decent engagement. Then the algo decides it knows better and starts chasing cheap clicks from 60+ users who aren't ever walking through the door.

The culprit (as someone in the thread mentioned) is almost always the engagement signal. Older users share the trendy promo with their grandkids, the algo sees a ton of shares, and decides to optimise for that behaviour instead of actual inquiries or conversions.

Here's what's worked for me:

Switch the optimisation goal to 'Conversions' and make sure you're tracking a meaningful action (e.g. 'Order Now' button click or phone call) - not just link clicks.
Set a strict frequency cap of 2 per person per week. Stops the same retiree sharing it five times.
Add a negative audience of '65+' and exclude existing customers who might be resharing.
Use a dynamic geo-fence with a radius that matches delivery zones.

Yes, narrowing reach feels counterintuitive, but better to spend £100 on 10 genuine leads than £500 on a thousand non-converting eyeballs from a different generation.

Anyone else find a workaround that sticks?

pixelperfect

i've seen this exact pattern with a few restaurant clients I've worked with, and it's frustrating. What happens is the platform initially tests your ads on a narrow audience slice to gauge reactions. once that early data runs out, the algorithm starts optimising for the cheapest engagement signals rather than your actual intent. older users tend to click, watch longer, and interact more predictably, so delivery gradually drifts-even when your targeting hasn't changed.

the real kicker is that location and age targeting don't override optimisation goals the way you'd expect. If your objective is engagement, traffic, or messages, the system will favour whoever is most likely to perform that action at low cost. Funnily enough, student promos and trendy food visuals often perform well with older users because they scroll and engage more, even if they never step foot in the restaurant. it's a common blind spot.

socialbutterfly

Oh, I've seen this one before 😅 My best guess? older family members are sharing it with the grandkids - that cross-generational engagement just throws the algorithm completely off. Meta's ad delivery gets all confused when the same content is getting reactions from totally different age brackets. Drives me nuts every time

bouncekiller

Oh honey, you're still trying to limit reach? That's like putting a tiny fence around a paddock that's already full of the wrong horses. Might as well just set your budget on fire and call it a day. If Meta's already decided your audience is a bunch of random cat-lovers and late-night kebab dodgers, limiting reach just means showing the same useless ads to fewer people. Genius move.

trafficcop

I've seen this exact mess play out with local restaurant accounts, especially in tight geo-targets. First week? Meta gobbles up those high-intent users like they're fresh bread. But once frequency spikes and the local pool runs dry, the algorithm gets lazy and chases cheap engagement - usually from an older crowd who just loves clicking anything. It's not always your targeting that's broken, it's that the signal-to-noise ratio tanks because your audience is tiny. Best trick? Swap creatives more often and optimise for messages or bookings instead of engagement. That usually stops the weird demographic drift.

ctrjunkie

This is such a classic Meta audience drift problem - the platform just chases the cheapest engagement signal it can find. That's why a promo aimed at students ends up served to 60+ audiences who clicked something once. Tightening demographics won't fix it long-term, it's baked into the ad delivery logic.

What campaign objective are you using?

communityfirst

It's always frustrating when a campaign that starts so strong suddenly veers off course. I've seen exactly this pattern with local restaurant accounts - the algorithm isn't being malicious, but it does prioritise the easiest engagement after the learning phase ends. that means it'll happily show your ad to people who tap or scroll by without ever booking a table, especially older audiences who tend to click more. The result? impressions waste and a dead campaign.

from my experience, the real fix is tightening your conversion tracking and switching your campaign objective away from traffic or engagement. Use conversions as your goal, and make sure your pixel is properly tracking reservations or phone clicks - that's the signal the algorithm needs to stay on target. also, building lookalike audiences from your actual customer base or website visitors works far better than broad location targeting. The first week's performance is real - you just need to lock in that audience before the system goes rogue on you

analyticsjunkie

yeah, I've seen this happen more times than I've seen a restaurant fail to geo-fence their own damn building. we call it the "grandma takeover" in my agency - once Meta decides your budget is better spent on Boomer clicks that cost pennies and never convert, the algorithm just drifts. and yeah, that broadening once the local pocket saturates? Classic Meta move. suddenly your ad's serving 50-year-olds three towns over who "like" the photo of the burger but will never drive 20 minutes for it.

What's your optimisation goal? if it's engagement, you're basically feeding the beast exactly what it wants. Might be time to switch to something that actually punishes cheap clicks - like calls or messages

liamcopy

This exact thing happened with a casual dining chain I worked with in Toronto.

First 5-7 days looked solid. Tight targeting, decent CTR. Then around day 8, delivery silently shifted into 55+ audiences even though we'd set 21-40. Engagement stayed flat. But dine-in reservations? They tanked. Went from around 40 bookings a week down to less than 20. CTR was still hovering near 4%. Makes you think Meta's working fine. It's not.

What's happening is "audience expansion" creep. Meta respects your filters early, then once it collects proxy signals (clicks, messages, directions requests) it starts chasing cheaper "likely converters". For restaurants those proxies are garbage. Someone clicks a menu link because they're bored, not because they're booking a table. Meta optimises for that engagement and drifts into older, less valuable audiences.

We rebuilt the campaign around reservation confirmations only. Fed actual booking data back through CAPI. Audience drift dropped massively. Within a month we were back to 50+ bookings a week and cost per reservation fell from $19 to $11. Meta finally started chasing verified diners instead of click-happy randoms.

When you check your campaigns, are those older audiences actually generating real reservations? Or just engagement with no footfall?

dailygrind

I've been noticing the exact same issue with local restaurant clients. Meta's algorithm loves to chase the cheapest CPMs and easiest engagement signals, and that almost always means skewing toward older demographics who scroll longer and click more passively. Once the system gets a whiff of that lower-friction audience, it starts dumping impressions on them even if they haven't been near your restaurant in years. The platform doesn't care about intent-it just optimises for whatever action is cheapest to buy.

A few things I've seen make this drift worse:

Engagement-based creative - If your ads are designed to get comments, shares, or reactions, Meta will funnel them toward people who engage for entertainment, not dining intent. Retirees love a good "looks delicious" comment but rarely convert.
Automated audience expansion - That default "broader targeting" option is a killer for local restaurants. It expands way beyond your radius and demographic signals once the campaign runs past a week.
Optimisation objective mismatch - If you're optimising for video views or landing page views instead of actual conversions (website purchases, calls, or messages), you're basically telling Meta to find the cheapest eyeballs, not the most likely customers.

I'd honestly check which optimisation objectives your best-performing campaigns are using. Switch to conversion-based (if you have the pixel data) or even a manual bid floor to cap the CPM spikes. Also, look at the delivery breakdown in Ads Manager-see if the age distribution is creeping up after day 5. That's usually the first red flag.

Would be curious to know what objectives you're running now. That'll tell you if the algorithm is working for you or just working the system.