Worked example
What $100 buys a tool directory
Link data is cheap per call, so the real question is how many calls a directory makes. Here is one invented month, line by line.
SEO Tools AI research team ·

The directory below, ToolShelf, is invented. Its numbers are made up to show the arithmetic of credits. Replace them with your own listing count.
ToolShelf, one month
| Job | Calls | Credits each | Credits | Cost |
|---|---|---|---|---|
| Score all 5,000 listings with bulk_dr | 3 | 1 | 3 | $0.0003 |
| Referring domains for the top 500 listings | 1 | 2 | 2 | $0.0002 |
| Anchors for 40 featured listings, one call each | 40 | 2 | 80 | $0.008 |
| Rival overlap for 12 new tools with intersect | 12 | 2 | 24 | $0.0024 |
| Month total | 109 | $0.0109 |
Read the total twice
About a cent. That is the whole month.
The $100 free credit is 1,000,000 credits, and this month used 109 of them. Even if ToolShelf checked every listing's anchors, two credits per call times 5,000 listings is 10,000 credits, or $1.
That is why the backlink checker tool free question is slightly the wrong one for a directory. Free web forms are fine for one lookup, and in bulk the paid per-call price is so low that engineer time costs more than data.

Turn your own listing count into a budget
Your own numbers will differ from ToolShelf's. The method stays the same, and it takes less time than it reads.
Start from the number of listings and the jobs you run on them. The rest is multiplication and one division by 10,000.
Count calls, not rows
Credits follow calls. A bulk_dr call takes up to 2,000 domains, so the row count only matters for how many calls you make. Divide your listings by 2,000 and round up.
Say a directory holds 4,100 listings. That is three scoring calls, not 4,100, and the three cost three credits in total.
Keep the row count in the same sheet as the call count, so the arithmetic stays visible. A reader who sees 4,100 rows and three calls can check the sum in seconds.
Price the monthly refresh
A monthly refresh runs twelve times a year. Multiply the calls in one run by twelve, then add the detail calls you run for featured rows.
The result is a yearly credit figure you can hold against the free $100, which is 1,000,000 credits. Most directories find the number far smaller than they feared.
Leave room for a repeat
A script that runs twice, or a sheet that resends every row, doubles the bill for that month. Plan with a margin, and read the _credits value after the first real run.
If the margin is not used, the balance simply stays for next month. A surprise bill is the only kind of surprise worth designing against.
A repeat made on purpose, such as a re-run after a fix to the mapping, should be logged with its reason. The extra credit then has an explanation the next editor can read.
Build your own month in five steps
List your jobs
Write one line for each kind of call your directory makes, including the free lookups you run by hand. A missing line is the usual reason a budget comes out low.
Count the calls
For each line, count the calls in a month. Keep rows and calls apart, so a 5,000-row sheet still counts as three bulk_dr calls.
Multiply by the credit cost
Use the cost per call from the field reference: 1 credit for bulk_dr and compare, 2 for the detail actions, and 3 for search and keyword.
Convert credits to dollars
One credit is $0.0001, so 10,000 credits is one dollar. Divide the monthly total by 10,000 to see what the month costs.
Set it against the free credit
Compare the total with the $100 credit, which covers 1,000,000 credits. A month of a few hundred credits leaves nearly all of it for later months.
What we would do
With $100, with $1,000, with nothing
The spending order is almost always the same, and none of it needs a subscription.
- With nothing: use the free actions and a monthly manual pass for under 50 tools.
- With the free $100: automate the whole listing column and keep two years of credit in reserve.
- With $1,000 to spend: put the money into editorial work. Better reviews move readers more than a fifth decimal place of link data.
Where the money really goes
Data is the smallest line of a directory's budget. Hosting, review time, screenshots and outreach to vendors all cost more.
Be careful with one trap: listing economics often reward whoever pays for placement, and a link score can then be used to launder that placement as merit. Keep paid slots visibly labelled, and sort the organic list by a metric you can explain.
Mind the quality bar. Google's helpful content guidance, folded into core ranking with the March 2024 core update, favours pages that show first-hand expertise. A directory that only repeats a number adds little. Add your own testing notes beside it.
A failure is not a disaster. If a call fails with HTTP 402, the credit is gone. Top up with USDT (TRC-20) from the account page, and the sync recipe in sheet and n8n sync shows how to stop cleanly.
Three ways a small budget goes wrong
Small budgets rarely fail on price. They fail on habits that quietly multiply the calls, and each habit has a simple fix.
None of these traps is expensive on its own. Together they can make a cheap month look like a surprise.
Re-scoring what did not change
A sync that rewrites every row each week spends credits on numbers the monthly index has not moved. The monthly schedule in the sync recipe avoids that trap.
If you need an extra run, scope it to the rows that changed. A list of ten new tools does not need the whole directory scored again.
A domain that is not in the index
A domain missing from the index still rides inside its call, and the call is billed the same either way. Retrying that domain on every run buys nothing.
Keep a blank cell and a short note, then move on. The note tells the next editor why the row is empty, which a retry never can.
A hundred listings in one afternoon
Say you add a hundred tools in one sitting. The credit cost of scoring them is a fraction of a cent, so the real cost lies in review time, screenshots and the links you then have to track.
Budget the editing hours first. The data bill is the smallest line, and it stays small even when the list grows quickly.

A second hypothetical: ten times the size
Now imagine a larger invented site, ToolAtlas, with 50,000 listings. Scoring all of them takes 25 bulk_dr calls, which is 25 credits or $0.0025.
Adding referring domains for every listing at 2 credits and 1,000 rows per call needs 50 calls, so 100 credits, or one cent. Add anchors and top pages and you reach about 300 credits.
Under a dime a month. The work that grows is on your side: storage, page templates and checking that scores still match the right domains. Plan for the listing fields before the calls.
Reading the bill from the response itself
You do not need a separate billing report to know what a month cost. Every authenticated response tells you the cost of that call and the balance left after it.
Log those two values with each run, and the monthly total becomes a sum you can check rather than an estimate.
Read the balance after each run
The _credits field carries the cost of the call and what remains. Store it beside the batch date, so a later editor can match a run to its cost without opening the API log.
If the balance falls faster than your plan says, stop the job and look for the extra calls before running it again.
Top up before the balance runs low
When credit runs out, calls return HTTP 402. A scheduled job that meets a 402 should stop and keep the old scores, not write blanks over them.
Top-ups are paid in USDT (TRC-20) from the account page. Topping up once a quarter with a margin is simpler than reacting to a failed run.
Before you set a monthly budget
- Every job has its own line, including the hand lookups.
- Calls are counted per month, and rows are never used as the call count.
- The refresh schedule is written down, with the number of runs per year.
- A margin is set aside for a repeated run or a late addition.
- The
_creditsvalue from the first run is logged and compared with the plan.
Money questions
Is the $100 credit really free?
Yes. Signing in with a Google account adds $100 of credit, which is 1,000,000 credits, and no card is asked for. When it runs out, calls return HTTP 402 until you top up with USDT.
Does a bigger response cost more?
No. Each call is billed once whatever the number of rows, up to 1,000 rows per call. Deep pagination is capped at 50,000 rows. That is why asking for the full page of rows is almost always the cheaper choice.