First, decide what engagement is for
Engagement is not a business outcome. It is a signal that your creative is holding attention, and a cheap way to build familiarity before you ask for anything. Treating it as the goal produces accounts that look wonderful and sell nothing.
In one plywood account we ran, eleven of twelve campaigns carried engagement and traffic objectives at fractions of a rupee per action, while a single lead campaign produced 278 Instant Form leads at ₹12.53 and 287 messaging conversations. Both were doing their job. Reporting them as one blended figure would have been misleading, which is why we never do it. The full breakdown is in the plywood case study.
Lever one: creative angle, not creative polish
The single largest cost difference we have measured came from the message, not the production quality. In a solar account in Bareilly, four ads ran on identical ₹250 daily budgets in the same week. Two used one angle and produced leads at ₹23.82 and ₹23.95. Two used a different angle and cost ₹87.59 and ₹97.04.
A four-fold difference, same city, same budget, same audience. The full test is documented on the Bareilly page.
What makes a test like that readable:
| Condition | Why it matters |
|---|---|
| Identical budgets | Different budgets change delivery, so you compare the auction rather than the message |
| Two ads per angle | One ad winning could be chance; two converging is a result |
| Same audience and placements | Change one variable, and make it the message |
| Let each ad exit the learning phase | Calling a winner on day three is calling noise |
Lever two: frequency is your real ceiling
Frequency is impressions divided by reach — how many times the average person saw your ad. It is the number most advertisers never look at, and it explains most of the mystery cost increases.
Across our accounts the pattern is consistent. In Kanpur we delivered 12.37M impressions against 3.69M reach, a frequency above three, and cost per lead started drifting up regardless of what else we adjusted. In Varanasi, frequency stayed at 2.1 and the account held at ₹10.87 per lead.
| What you see | Likely cause | What to change |
|---|---|---|
| Cost rising, frequency above 3 | Creative fatigue | New creative, hold the budget |
| Cost rising, frequency below 2 | Audience or offer mismatch | Test a different angle |
| High engagement, no conversions | Creative entertains but does not sell | Rework the offer, not the targeting |
| Almost no delivery, tiny spend | Auction is not picking the ad up | Check the creative, not the audience |
The practical rule: once frequency passes roughly three, adding budget buys repeat views of an ad people have already ignored. The fix is new creative, not more money.
Lever three: stop narrowing the audience
Narrow targeting feels efficient and usually is not. In a Haryana solar account, an open-targeted ad produced leads at ₹32.69 while a manually narrowed version of the same offer cost ₹53.85 — and could only reach a third as many people. We saw the identical pattern in a second Haryana account.
The mechanism is straightforward: a smaller pool gives the auction fewer cheap placements to find. You pay more and your ceiling is lower. Meta's own guidance has moved in the same direction for years, and the current documentation sits at Meta for Business.
Where narrowing genuinely helps: geography. A salon with a five-kilometre catchment should be tight on radius and loose on everything else.
Lever four: match the objective to the ask
An engagement objective optimises for people who like and comment. A lead objective optimises for people who submit forms. They are different audiences within the same pool, and choosing the wrong one produces exactly the wrong result cheaply.
If you want enquiries, run a lead objective even though the cost per action looks worse. An account full of ₹0.03 engagements and no enquiries is not a cheap account, it is an unproductive one.
Lever five: read the zero-result rows correctly
In a seven-ad test in Prayagraj, two ads returned nothing — and had spent under ₹5 each. That is not a failed creative, that is the auction never giving those ads delivery. Judging them as losers removes angles that were never actually tested.
Always read the spend column before the results column. It is the most common misreading of an ads dashboard we come across.
What we do not recommend
Deleting underperformers. Cap the budget instead. A paused campaign loses its learning phase if you restart it, and a capped one stays valid as a comparison point when the market shifts.
Doubling a winning budget overnight. It usually resets delivery and produces a week of expensive results that look like the change failed. Raise gradually and give it seven to ten days.
Judging a test in three days. Each ad needs to exit the learning phase. On modest budgets that is two to three weeks.
A checklist you can run this week
Open the account and check, in this order: frequency on every active ad set; whether any two ads share an angle or whether all six are variations of one idea; whether your objective matches what you actually want; whether any campaign is spending meaningfully with no results; and whether the audience is narrowed for a reason you can defend.
In our experience that five-minute pass finds the problem in most underperforming accounts before anyone touches a creative brief.
How Independent Agency works on Meta accounts
We run Meta Ads from Lucknow for businesses across India, and we publish the results rather than describing them. The portfolio carries 37 unedited account screenshots with client names blurred, including the campaigns that cost far too much. If you want the full cost range in one category, the solar marketing page shows twelve markets from ₹5.86 to ₹63.11 per lead.
Frequently asked questions
What is a good frequency for Meta Ads?
Below three for most accounts. Past that, additional budget buys repeat views rather than new people and cost per result drifts up regardless of other changes.
How many creatives should I test on Meta?
At least two genuinely different angles with two ads each, on identical budgets. Six variations of one idea is one test, not six.
Does narrowing my audience lower cost per result?
Usually not. In our Haryana accounts open targeting produced leads at ₹32.69 against ₹53.85 for a narrowed version of the same offer, and reached three times as many people.
Why do my ads get engagement but no leads?
Usually because the objective is engagement, or the creative entertains without making an offer. Change the objective and the offer before changing the targeting.
Should I delete campaigns that are not working?
Cap the budget instead. A deleted campaign takes its learning with it and is no longer available as a comparison point when the market changes.
How long should a Meta Ads test run?
Long enough for each ad to exit the learning phase — typically two to three weeks on modest budgets.
What does engagement actually cost on Meta?
In accounts we have run, post engagement cost between ₹0.02 and ₹0.62 per action. Form leads in the same account cost ₹12.53, which is why the two should never be blended.
Performance figures on this page come from selected campaigns and reporting periods in our portfolio records through September 2026. Several screenshots are filtered views of the same ad account, so totals are calculated from distinct accounts only. Results vary with market, budget, targeting, offer, competition and sales follow-up.