What ₹500 a Day on Meta Ads Can Actually Teach You

Most small businesses in Delhi start Meta advertising — Facebook and Instagram ads, run through the same system — with a number in mind rather than a question. Five hundred rupees a day. Maybe a thousand. Twenty thousand for the month, see how it goes.

That is a perfectly reasonable place to start. What is not reasonable is the expectation that usually comes attached: that after a month you will know something. Which audience works. Which creative wins. Whether Instagram beats Facebook for your business. A modest budget buys you a small number of outcomes, and a small number of outcomes cannot tell you much with confidence. This article is about which questions your money can actually answer, and which ones you will have to answer some other way.

The fifty-event problem

Meta's own advertiser documentation describes something called the learning phase. When a new ad set starts running, the delivery system is working out who to show your ads to. During that period, performance is unstable and cost per result swings around. Meta states that an ad set generally exits the learning phase after roughly 50 optimisation events in a seven-day window. An optimisation event is whatever you told the campaign to optimise for — a lead form submission, a purchase, a landing page view, a message.

That threshold is the single most useful number in small-budget advertising, because it tells you what your spend needs to produce before the system is even settled, let alone before you can judge it.

Work backwards. Suppose you are optimising for leads and your cost per lead lands at ₹300. Fifty leads a week means ₹15,000 a week, or roughly ₹2,100 a day, on one ad set. If you are running three ad sets to test three audiences, that is ₹6,300 a day to get all three out of learning.

Most small advertisers are spending a fraction of that. Which means the honest description of a ₹500-a-day campaign is: it is permanently in the learning phase. Meta will still deliver it. You will still get results. But the system never reaches the stability its own documentation treats as the starting line, and every edit you make restarts the clock.

The arithmetic of small numbers

Put the learning phase aside for a moment and look at plain counting. At ₹500 a day and ₹250 a lead, you get two leads a day. Sixty a month. Split across two ad sets, that is thirty each.

Now one ad set produces 34 leads and the other produces 26. Is the first audience better? Almost certainly not in any way you can rely on. That gap is well within the range you would expect from pure chance with numbers that small. Flip a coin sixty times and you will rarely get thirty-thirty. Nobody concludes the coin is biased.

This is the quiet tragedy of small-budget testing. The reports look precise. The Ads Manager shows cost per result to two decimal places. The precision is real; the reliability is not. You will make a confident decision to kill the losing audience, and half the time you will have killed the better one.

It gets worse when the thing you are measuring is rarer than a lead. If you sell a ₹60,000 service and close one client a month from ads, a month of data contains exactly one data point. You cannot test anything with one data point. You can only observe.

What a small budget genuinely tells you

This is not an argument against advertising with ₹500 a day. It is an argument for asking it the right questions. There are several things a small spend answers well, because they do not depend on small differences between similar options.

Whether anything happens at all

Big effects show up in small samples. If one creative gets 40 leads and another gets 3, you do not need statistics. You need to make more of the first one. Small budgets are good at finding disasters and finding runaway winners. They are bad at everything in between.

Your rough cost of attention

Cost per thousand impressions and cost per link click stabilise much faster than cost per lead, because there are vastly more of them. A ₹500 day might buy two leads but several hundred clicks' worth of data over a month. That tells you what it costs to get a Delhi audience to look at you — useful for planning, and reasonably trustworthy.

Whether your offer makes sense to strangers

This is the most valuable thing and the least discussed. If people click and then nothing happens, the problem is rarely targeting. It is that the page they land on does not answer the question the ad raised. You find that out fast and cheaply, and fixing it is free. If your landing pages are not built to convert, that is a web development problem, not an advertising one, and no amount of budget will paper over it.

What people actually ask

The comments, the WhatsApp messages, the objections that come back — these are qualitative and they arrive in quantity even at low spend. Read every one. They are worth more than the dashboard.

What it cannot tell you

Which of two similar audiences is better. The difference is usually small, and small differences need volume to detect.
Whether Instagram or Facebook placement is better for you. Same problem, plus Meta's automatic placements move spend around during the test.
Whether a headline change helped. Copy A/B tests need far more conversions than a small budget produces.
What your true cost per acquisition is over time. One good month and one bad month at this scale is normal noise, not a trend.
Whether last month's dip was the algorithm. It might have been. It might have been a festival, a competitor's sale, or nothing.

How to spend a small budget well

Run one ad set, not five. Concentrate every rupee behind a single audience so it gets as close to the fifty-event threshold as your budget allows. Splitting a small budget guarantees that nothing exits learning.
Use a broad audience and let Meta find people. Narrow interest stacking made sense when the system needed help. With limited spend, hand-built micro-audiences mostly just raise your costs.
Optimise for an event that actually happens. If purchases are rare, optimise for a step further up — a form fill, a message, an add to cart. More events means more signal, even if each one is worth less.
Change big things, not small things. Test a video against a static image, not one headline against another. Test a free consultation against a price-led offer. Only large differences are visible at this scale.
Stop editing. Every meaningful change to an ad set restarts the learning phase. Leave it alone for at least a week unless something is obviously broken.
Judge on a quarter, not a week. Sixty leads is a thin month. One hundred and eighty is a number you can start to reason about.
Track what happens after the click. Which enquiries became quotes, which quotes became jobs. Meta cannot see this. If those enquiries are landing in a shared inbox and getting lost, a simple CRM system will tell you more about ad quality than the dashboard will.

What not to buy

A few things get sold to small advertisers that we would not pay for at this budget level.

Elaborate audience research decks. Someone charging you to build fourteen interest-based audiences for a ₹15,000 monthly spend is selling complexity you cannot afford to run. You will never get any of them out of learning.

Monthly reports full of percentage changes. "Cost per lead down 18% month on month" is meaningless on sixty leads. If a report leads with percentages instead of absolute numbers, it is decoration. Ask for raw counts.

Retargeting campaigns before you have traffic. Retargeting works by showing ads to people who already visited. If your site gets a few hundred visitors a month, your retargeting pool is too thin to deliver properly and you are just splitting an already-small budget.

A separate campaign for every service. Tempting, structurally tidy, and fatal at low spend. Pick the service with the best margin and advertise that.

Follower growth as a paid objective. If you want an audience on Instagram, earn it with consistent organic posting. Paying for followers who did not choose you is the most expensive way to make a number go up.

The honest summary

A modest Meta budget is a way of buying distribution and a small stream of enquiries. It is a poor laboratory. Treat it as the first and you will be pleased. Treat it as the second and you will spend a year making confident decisions on evidence that cannot support them.

There is also plenty here that nobody outside Meta can state with certainty. How the delivery system weighs signals, how much a pause really costs you, whether a given account behaves differently from another — these are things experienced advertisers have opinions about, not facts. Be suspicious of anyone who explains Meta's system to you without ever saying "probably".

What to do next

Open Ads Manager and look at one thing: the total number of optimisation events your account produced last month, not the cost per event. If it is under about 200, you do not have a testing programme — you have a single campaign, and you should structure it that way. Collapse your ad sets into one, pick your strongest offer, and leave it running untouched for a fortnight.

Then make sure the traffic has somewhere good to land. Most small-budget campaigns underperform because of the page, not the ad. If you want a second pair of eyes on either side of that — the campaign structure or the page it points at — get in touch and tell us what you are spending and what you are selling.

Questions we get asked

What is the minimum daily budget for Meta ads in India?

There is no technical minimum that matters much — Meta will run ads on a few hundred rupees a day. The more useful question is what your cost per result is. Meta's documentation says an ad set generally needs about 50 optimisation events in seven days to exit the learning phase, so multiply your expected cost per lead by 50 and divide by seven to see what a fully stable ad set would cost per day. Most small businesses will be below that, and that is fine as long as you know it.

Why does my cost per lead keep changing so much?

Partly because small numbers are volatile — two leads on Tuesday and none on Wednesday looks dramatic as a percentage but means almost nothing. Partly because if your ad set never exits the learning phase, Meta is still exploring who to show your ads to. Judge performance over a month or a quarter, using absolute counts rather than percentage swings.

Should I run separate campaigns for Facebook and Instagram?

At a small budget, usually not. Splitting spend across placements means neither gets enough volume to settle, and you still will not be able to tell which performed better with any confidence. Use automatic placements and let the system allocate, then revisit the question if your spend grows substantially.

How long should I leave an ad running before deciding it does not work?

At least a week without edits, and preferably two. Every meaningful change to an ad set restarts the learning phase, so frequent tinkering keeps you permanently at the unstable stage. The exception is something obviously broken — a dead link, a form that does not submit, spend burning with zero clicks.

Is boosting a post cheaper than running a proper campaign?

Boosting is not cheaper per result in any reliable sense; it is simply simpler, with fewer objectives and controls available. At very small budgets the practical difference narrows, because you are not doing sophisticated structuring either way. The bigger factor in your cost is usually the offer and the landing page, not which interface you used.