Meta
Medium
Should Meta Local Ads Optimise for Reach or Conversion?
A local business (salon, restaurant, clinic) doesn't want impressions or even clicks. They want footfall/bookings/calls within a small radius. Their real success metric is incremental customers per rupee spent.
A salon owner in XYZ locality doesn't want reach. It wants more paying customers than it would have got without the ad. That's the only truth.
By this logic alone, conversion wins. Simple. So why is this even a debate?
But there is a real tradeoff.
Impressions happen thousands of times a day → the model learns reach easily.
Conversions for a salon happen maybe 10 times a week → the model has almost nothing to learn from. So conversion optimisation for a local advertiser means asking the model to predict something rare, using data it mostly can't observe.
That's the whole tension: the goal you want has no data; the goal with data isn't what you want.
What can be done?
Optimise for middle signals. "Get Directions" taps, call-button taps, WhatsApp messages. These happen 20–50x more often than purchases, and someone tapping "Get Directions" is very likely to visit. Enough data + close to the goal. Best of both.
Borrow data. One salon has 10 conversions a week. But Meta has thousands of salons. Pool their learning — what predicts a salon visit in one area predicts it elsewhere too. Now conversion optimisation works even for small advertisers.
We should also measure: did the advertiser come back and spend again next month? A local business owner re-spends only if the shop got busier. Retention of ad spend is the honest metric.
Community Answers (2)
For me reach and conversion aren’t equal objectives. Reach measures distribution, conversion measures outcome, first principal thinking. For any advertising product, your ultimate goal is driving outcome; that is right north star metric to measure. But real question is whether conversion is measurable at the scale needed for the system to optimize effectivity. If signal is too sparse, then system cannot optimize effectively and platform looks broken even when isn’t. So, the answer isn’t reach vs conversion, it is about finding the right optimization signal at each stage. Early in a campaign, optimize for frequent engagement signals that correlate with the outcome you actually want. As real conversion data builds up, shift the optimization towards conversion directly. Role of advertisers shouldn’t have to manage this shift themselves, the platform should detect when enough signal exists and move optimization automatically. In the end, for me conversion is the goal. But getting there means the system has to earn its way from easy signals to hard ones.
Local ads have limited reach, hence limited data. Reach is an easy goal, because you have to just distribute the impressions without worrying about if the impressions are reaching to right people also or not. Conversion seems a lucrative metrics, but there could be a pitfall, i.e. not having enough conversion instances might keep the model guessing. Hence as a result the cost per conversion will come out to be very high (since the model is still learning and not purely efficient) Hence, focus shall be on reach if the ad account or ad account X campaign is new. Once there are enough signals, the optimisation can shift to conversion.