Three titles, 14 days, one page

A daily publishing report: who is playing, whether they come back, whether the players you bought have paid for themselves yet — and whether today's numbers can be trusted.

61,773players a day

$12,183revenue in the window

$1.44cost per install

1/3titles profitable per install

4checks wanting attention

Every number here is invented. This is a working rebuild of a report that ran daily for a games publisher — same structure, same questions, generated data. None of the original company's numbers, titles or systems are in it.

Use the buttons above to walk through it. Each view answers one question, and says which decision it is meant to support.

Who is playing

Daily and monthly players per title, and how many of the month's players show up on an average day. That last number is the one to watch: it moves before revenue does.

61,773players a day, all titles

82,774new players in the window

Title	Players/day	Players/week	Players/month	Stickiness	New players
Emberfall Saga	21,770	45,131	81,525	27%	27,828
Tidebreak Arena	24,272	50,149	83,889	29%	33,506
Lanternwood	15,731	33,910	72,417	22%	21,440

Stickiness is daily players divided by monthly players. Around 20% means a typical player shows up about six days a month; above 30% means they are in most days. It falls before revenue falls, which is why it is on the first screen rather than buried in a monthly deck.

The first five minutes

Where the people you just paid for go before they ever play. Every step is a percentage of the step above it, so the compounding is visible — four steps at 85% each is only 52% end to end, and it is the 85% that gets quoted in meetings.

Emberfall Saga

27,828 installs became 11,953 finished tutorials — 43% of what you paid for.

Install

27,828 · 100%

The store reports it. Nothing of yours has run yet.

First open

21,314 · 77%

The biggest drop in most games — and the one nobody instruments, because your SDK starts here.

Account created

18,111 · 65%

Every field you add to this form costs you players.

Tutorial started

16,837 · 60%

They are in. This step is mostly load time.

Tutorial finished

11,953 · 43%

Where a long tutorial shows up as lost players.

Tidebreak Arena

33,506 installs became 13,853 finished tutorials — 41% of what you paid for.

Install

33,506 · 100%

The store reports it. Nothing of yours has run yet.

First open

25,433 · 76%

The biggest drop in most games — and the one nobody instruments, because your SDK starts here.

Account created

22,530 · 67%

Every field you add to this form costs you players.

Tutorial started

19,905 · 59%

They are in. This step is mostly load time.

Tutorial finished

13,853 · 41%

Where a long tutorial shows up as lost players.

Lanternwood

21,440 installs became 10,655 finished tutorials — 50% of what you paid for.

Install

21,440 · 100%

The store reports it. Nothing of yours has run yet.

First open

17,323 · 81%

The biggest drop in most games — and the one nobody instruments, because your SDK starts here.

Account created

15,506 · 72%

Every field you add to this form costs you players.

Tutorial started

14,377 · 67%

They are in. This step is mostly load time.

Tutorial finished

10,655 · 50%

Where a long tutorial shows up as lost players.

The install-to-open drop is usually the largest and is almost never instrumented, because the analytics tool only starts reporting once the app opens. It is measured here by comparing what the store billed against what the game first saw.

Do they come back

Of the players who installed on a given day, how many are still opening the game 1, 7, 30 and 90 days later. This is the number that decides what an install is worth, and therefore what you can afford to pay for one.

Emberfall Saga

40%D1

28%D2

21%D3

17%D7

14%D14

12%D30

11%D60

10%D90

57% gone within 30 days · 9% of those came back on their own, without being paid for a second time.

Tidebreak Arena

46%D1

29%D2

21%D3

16%D7

13%D14

11%D30

9%D60

8%D90

60% gone within 30 days · 10% of those came back on their own, without being paid for a second time.

Lanternwood

41%D1

24%D2

16%D3

12%D7

9%D14

8%D30

6%D60

5%D90

55% gone within 30 days · 12% of those came back on their own, without being paid for a second time.

Revival — players who left and returned unprompted — is tracked separately because paying to re-acquire someone who was coming back anyway is one of the easiest ways to waste a marketing budget.

Does an install pay for itself

What you paid for a player, against what that player has actually spent by day 7, 30 and 180. The gap between the last two columns is the entire argument for patience.

Title	Cost per install	Back by D7	Back by D30	Back by D180	D30 return	Breaks even	Profit per install
Emberfall Saga	$1.44	$0.98	$2.24	$5.93	155%	day 30	+$4.48
Tidebreak Arena	$1.47	$0.21	$0.52	$1.01	35%	never	−$0.46
Lanternwood	$1.42	$0.05	$0.14	$0.25	10%	never	−$1.17

Revenue per player, as it arrives

Most of a player's value shows up after day 30. A campaign judged on its seven-day return looks like a failure while it is still on its way to profit — which is how profitable campaigns get switched off.

Title	D0	D1	D7	D14	D30	D60	D90	D180
Emberfall Saga	$0.40	$0.29	$0.98	$1.28	$2.24	$3.38	$4.03	$5.93
Tidebreak Arena	$0.09	$0.06	$0.21	$0.34	$0.52	$0.66	$0.77	$1.01
Lanternwood	$0.02	$0.02	$0.05	$0.08	$0.14	$0.17	$0.20	$0.25

Where the marketing money went

Spend from the ad networks, installs from the attribution provider — never both from the same place, for the reason in the last column.

	Spend	Installs	Network claimed	Over-claim	Cost per install	Return
facebook	$993,630	786,697	1,166,314	1.48×	$1.263	111%
google	$343,962	170,956	211,794	1.24×	$2.012	67%
tiktok	$57,669	10,286	16,398	1.59×	$5.606	26%

Averaging each channel's cost per install instead of weighting it by spend gives $2.96 — $1.52 high. That is the single most common reason a marketing report disagrees with finance.

Over-claim is what a network says it delivered divided by what was actually attributed to it. It is above 1.0 everywhere, always, because each network counts installs it believes it influenced and those windows overlap. Add the three together and you get more installs than the game had.

By title

	Spend	Installs	Network claimed	Over-claim	Cost per install	Return
emberfall	$557,528	386,984	559,391	1.45×	$1.441	94%
lanternwood	$418,166	295,484	418,529	1.42×	$1.415	103%
tidebreak	$419,567	285,471	416,586	1.46×	$1.470	93%

Can you trust today's report

The checks the report runs on itself before anyone reads it. A daily report that is quietly wrong is worse than one that is late, because the decisions get made anyway.

2passed

3worth a look

1needs fixing

Spend matched to installswarn$62,628 of $1,395,260 unmatched (4.5%)

Money we were charged that no install can be attributed to. Usually a campaign renamed in the ad tool and not in the mapping — it reads as an expensive channel until someone checks.

Network claims vs attributedwarntiktok claims 1.59× what was attributed

Every network counts installs it believes it influenced, and the windows overlap. Adding the networks' own numbers together would overstate installs for the whole title — which is why cost comes from the networks and installs come from the provider, never both from one place.

Every source reported every dayfailfacebook missing 1d, google missing 1d, tiktok missing 1d

A source that delivered nothing looks exactly like a quiet day in every chart downstream. This is the check that tells the two apart.

Revenue agrees across systemsokstore vs in-game ledger differ by 0.11%

Refunds, currency rounding and store fees land at different times in each system. A gap under about 1% is normal; a gap that grows week over week is not.

Yesterday's numbers heldwarn1 metric(s) restated after late data

Attribution keeps arriving for days. If yesterday's install count moves after someone has already reported it, they need to hear it from the report rather than from the person they reported it to.

No duplicated eventsok20 duplicate rows dropped

A retried webhook delivers the same purchase twice. Nothing in a revenue total looks wrong afterwards — it is just larger.

Deliberately not a wall of green ticks. A check that always passes is one people stop reading, so each of these has failed at least once in the system it came from.

How this is put together

One page, built once a day, that needs no login and no dashboard seat. The short version of what happens before you open it.

Collect

3 networks

Each ad network in its own format, its own currency and its own calendar. One of them reports in a different timezone, which is worth a day of spend if nobody notices.

Agree on a number

1 source of truth

Cost from the networks, installs and revenue from the attribution provider. Never the same number from two places.

Measure

1,140 rows

Cost per install weighted by what was actually spent — not the average of each channel's rate, which flatters the smallest and most expensive channel.

Check

6 checks

Missing sources, duplicate events, restated numbers, spend that matches nothing.

Publish

1 link

A page, sent to whoever needs it. No seat licence, no tool to learn.

2026-07-01, 14 days · 1,140 campaign-country-days. Rebuilt from scratch for this demo — the production original ran against a games publisher's warehouse and none of it is here. The architecture is the transferable part: which system is authoritative for which number, and what happens where they meet.