Pricing and ad questions

A PM interview pricing question asks you to defend a price with no sales data, and an ad question asks what one click is worth to the business buying it. Cracking the PM Interview files ad questions under pricing, which fits: both turn on what something is worth to whoever pays.

How would you price a paid plan for a hiking-map app?

Give the prompt edges before any numbers. The app is generic and these are my assumptions: 500,000 people a month use a free version with trail maps, and the paid plan adds maps you can download for use without signal, plus route planning. My goal is the most profit from the plan over the next year. If the interviewer says the goal is growth, the answer moves lower, so ask.

I look for three reference points: what the plan is worth, what people would use instead and what each user costs.

What the plan is worth to a hiker: the ceiling

Start with one person and ask what they'd gain in a year, in money. Here's a regular who hikes 20 days a year:

• Paper maps they'd stop buying: 3 new areas a year × $15 a map = $45.

• Planning time saved: 20 hikes × 15 minutes = 5 hours. Hobby time is worth less than work time, so I'll value it at $10 an hour: 5 × $10 = $50.

• Total: $45 + $50 = $95 a year.

Now someone who hikes 2 days on holiday. One $15 map, plus 2 × 15 minutes = 30 minutes at $10 an hour, which is $5. Total: $20.

Those are ceilings. Nobody pays full value, because then the purchase gains them nothing. The gap between $95 and $20 also tells you the plan is for regulars. Not getting lost may matter more than either line, but I can't defend a number for it, so I'd name it and leave it out.

Try this step alone on the case below: work out only what the product is worth to one buyer, in dollars, then stop.

What a hiker would use instead: the anchor

An anchor is the price people hold yours against. A hiker can't feel $95 of worth, but they can see another app's price tag. Their alternatives:

• Free: their phone's built-in maps, or this app's free version.

• Paper: $15 an area, from above.

• Rival hiking apps: assume the best-known ones charge $36 a year.

If you don't know a rival's price, say your guess:

"I think rival apps sit around $36 a year. Tell me if that's off, and I'll adjust."

So their real ceiling is close to $36 unless this app is clearly better, whatever the $95 says. The paper map helps you sell it: $36 ÷ $15 = 2.4 maps, and they buy 3 a year.

Should you undercut? A much lower price invites a rival to match it, and then you've both lost money. I'd start at the anchor and let a test move me.

For the case below, say only what the buyer would compare this product to and what that costs, then stop.

What each user costs you: the floor

Below the floor, each sale loses money. My assumptions: map data and servers cost $2 per paying user per year, support costs $1, and the app store keeps 15% of the price. So each buyer leaves the price × 85%, minus $3. That reaches zero at $3 ÷ 85%, about $3.50.

The floor rarely sets a software price, but it lets you check whether the business works at all. A team of three at $150,000 each costs $450,000 a year. At $36, each buyer leaves $36 × 85% minus $3 = $27.60, and $450,000 ÷ $27.60 is about 16,300 buyers. 16,300 ÷ 500,000 is about 3.3% of monthly users, possible but far from certain.

That gives a floor of about $3.50, an anchor of $36 and a ceiling of $95 for a regular. Reasoning won't narrow that any further.

How to test a price without a demand curve

A demand curve is a chart of how many people buy at each price. Nobody will hand you one, and you only need a few points from it. Pick three prices around the anchor, show each to a random group of new users, count the buyers, and work out what you keep.

I made these results up to show the method. Here $36 wins, even though $24 sells more plans and $48 keeps more per sale. Given my goal, the last column decides.

I'd tell the interviewer where this test falls short:

• Small counts. The gaps in the last column could be chance. Telling the prices apart with a standard test (a 5% risk of a false alarm, an 80% chance of catching a real gap) takes roughly 4,000 to 10,000 people per price.

• First sale only. A price that sells well and doesn't renew loses later, so I'd read renewals at month 12.

• People compare notes. Someone paying $48 who meets a friend paying $24 will complain. I'd test on new users over a short window and give the lowest price to anyone who asks.

Read the case below as a result someone just handed you, and write only what you'd check before trusting it.

Choosing a pricing model for this app

Compare only the models that could fit this app.

What the pick gives up is the holiday hiker, whose $20 ceiling sits well below $36. They stay on the free version. Our longer article on pricing strategy questions compares models in general.

"I'd charge about $36 a year and keep a free version, with offline maps behind the plan. A regular hiker gets about $95 a year of value, but rivals sit near $36, so that's where I'd start. I'd test $24 and $48 against it on new users, keep whichever leaves the most money per 1,000 people who see the offer, and check renewals a year later."

For the cases below, name two models that could fit and one reason against each, then stop.

How search ads are priced: four ratios

Now an online bike shop buying search ads, the kind that appear when someone searches for "gravel bike." Assume one week looks like this: the ad appeared 50,000 times, got 2,000 clicks, cost $2,400 and led to 30 orders.

• Click-through rate (CTR): the share of showings that get a click. 2,000 ÷ 50,000 = 4%.

• Cost per click (CPC): what each click cost. $2,400 ÷ 2,000 = $1.20.

• Cost per thousand impressions (CPM): an impression is one showing of the ad, and the M comes from mille, Latin for thousand. 50,000 showings is 50 thousands, so $2,400 ÷ 50 = $48.

• Conversion rate: the share of clicks that become orders. 30 ÷ 2,000 = 1.5%.

They connect: 1,000 showings × 4% = 40 clicks, and 40 × $1.20 = $48, the same CPM. The shop pays per click, so $48 is only what its clicks came to per thousand showings in my made-up week. It isn't a display-ad price, so don't quote it as one.

In the case below, say which of these ratios each price you quote belongs to before you quote it.

The most the shop should pay for a click

Profit per order, on my assumptions: a $600 average order × 30% margin = $180, less $60 for shipping, payment fees and returns, leaves $120.

1,000 clicks × 1.5% = 15 orders. 15 × $120 = $1,800 of profit. $1,800 ÷ 1,000 clicks = $1.80 a click.

$1.80 is break-even, so it's the most the shop should pay, and it should want to pay less. At today's $1.20 it keeps $0.60 a click: 2,000 × $0.60 = $1,200 a week.

Per thousand impressions, 40 clicks × $1.80 = $72. So the shop could pay a CPM of up to $72, but only while its click-through rate holds. If the ad slips down the page and CTR falls to 3%, that's 30 clicks × $1.80 = $54.

Two things move the limit, and I'd raise them unprompted:

• Repeat buyers raise it. If each buyer later returns for tires and chains worth $40 of profit, 15 × ($120 + $40) = $2,400, or $2.40 a click.

• Buyers who'd have come anyway lower it. If the shop also shows up in the unpaid results for "gravel bike," some people who click the ad would have clicked that listing instead and bought. If a third of these orders would have happened regardless, only 10 count: 10 × $120 = $1,200, or $1.20 a click. That's what the shop pays today, so the ads would be earning nothing.

What auctions and quality scores change

The $1.80 caps what the shop should pay. What it does pay is generally set by an auction, run each time someone searches, so it depends on who else is bidding. Platforms also tend to weigh how likely each ad is to be clicked and how useful the ad and its landing page look, often summed up as a quality score.

Here's why they'd care. Suppose a rival pays $3 a click, but its ad has a 1.5% CTR. Per thousand showings that brings the platform 15 clicks × $3 = $45. The shop at $1.20 and 4% brings 40 × $1.20 = $48, so the lower price per click can be worth more to the platform.

A better ad can therefore pay less for the same spot, a poor one more, and the price paid can sit below the limit. Exact rules differ by platform and change over time, so in an interview I'd describe the idea and avoid stating a formula as fact.

Display ads sold through ad exchanges run separate auctions with their own rules. Google Ad Manager, for one, said in 2019 that it was moving to a first-price auction, where the winner pays what it bid.

"With a 1.5% conversion rate and $120 of profit an order, a click is worth $1.80, so that's the shop's limit, or $72 per thousand at a 4% click-through rate. I'd bid below that to leave some profit, and lower it further if many of these buyers would have found the shop anyway. What we'd actually pay depends on the auction, so I'd watch the real cost per click against that limit each week."

For the case below, describe a single auction from start to finish in your own words, then stop.