Impressions needed calculator

Conversions wanted is the outcome the period has to produce: orders, qualified leads, signups. One kind of conversion at a time, because the rates below differ for each.

Conversion rate is the share of clicks that become that conversion, from your own reporting. Landing page, offer, audience and device mix all move it, so take it from the closest comparable period you have rather than from an article.

Click-through rate is the share of impressions that become clicks, again from your own account and for the same placement. A search rate and a feed rate are not interchangeable and using one for the other will be wrong by a wide multiple.

Cost per thousand impressions is optional, and it is what turns a volume plan into a budget. Take it from a recent report for the same placement, because inventory prices vary far more between placements than between advertisers.

The arithmetic is three steps and they are worth walking through once, because seeing them stated makes the sensitivity obvious.

150 conversions at a 2.5 per cent conversion rate needs 6,000 clicks. 6,000 clicks at a 1.2 per cent click-through rate needs 500,000 impressions. 500,000 impressions at a cost of 9 per thousand costs 4,500.

Every figure in that chain is a multiplication of the ones before it, which is the single most important property of this calculation. The two rates compound: impressions needed is the target divided by the product of the two, so an error in either moves the answer proportionally and an error in both moves it a great deal more. A click-through rate a third lower than assumed raises the impressions needed by half, and the budget with it.

A forward plan answers a question nobody asked. The business rarely wants to know what 5,000 of media produces; it wants to know what delivering the quarter costs.

Planning backwards also exposes infeasibility early, which is the most valuable thing a plan can do. A target that requires more impressions than the addressable audience can absorb is not a budget problem, and no amount of money fixes it. Neither is a target that requires a conversion rate the site has never achieved. Both of those look perfectly reasonable as a budget request and are visible immediately as a volume requirement.

The third thing it exposes is where the leverage is. The two rates sit in the denominator, so improving either reduces the impressions and the budget more than proportionally. Lifting the conversion rate from 2.5 to 3 per cent cuts the impressions needed by a sixth, at no media cost at all. That comparison — a landing page change against a budget increase — is one this calculation makes concrete and most plans never state.

The auction's response to your own spend. Buying more impressions in the same audience usually costs more per thousand than the first ones did, so a plan built on a current cost per thousand understates the budget at materially higher volumes.

Audience exhaustion. The plan is in impressions, not people. If the audience is small relative to the impressions required, frequency climbs and the click-through rate falls, which invalidates the rate the plan was built on precisely when the plan is being executed.

Time. 500,000 impressions is a different plan over a week than over a quarter, in what it costs and in what it does to frequency. The period belongs in the plan even though it is not in the arithmetic.

The learning cost of automated bidding. A campaign needs a certain volume of conversions before automated bidding stabilises. A plan that is technically feasible but delivers the conversions too slowly to escape that state will underperform its own rates.

Variance. A plan for 150 conversions will not deliver 150. Small numbers move a great deal, and a plan for fewer than a hundred conversions should be read as a range rather than a figure.

Everything that is not this channel. The chain assumes the conversions come from the impressions it plans. In a business where people see an ad, search the name a week later and buy from somewhere else entirely, the plan will be wrong in both directions at once: the channel produces conversions it is never credited with, and it is credited with conversions that would have happened anyway.

Plan a range rather than a point. Run the calculation with the rates you expect, then again with rates a quarter worse, and treat the second as the budget the business should be prepared to spend to hit the target. The gap between the two is the honest measure of how much the plan depends on assumptions.

Phase it. Commit enough budget to measure the real rates, replace the assumptions with measurements, and recompute before committing the rest. A plan that is recalculated twice during a quarter is worth more than one that was precise on the day it was written.

And check the plan against the two things it cannot compute: whether the audience is large enough to absorb the impressions, and whether the period is long enough to serve them without the frequency doing the work instead of the reach.