Enrichment is priced per row. That single fact means the order of your steps is your budget.
The order most teams use
Enrich everything, then score, then filter.
This pays full price for every row, including the ones a free condition would have removed in the first second. It is also the intuitive order, which is why it is so common.
The order that costs a fraction
1. Free filters first. Geography, company size, obvious disqualifiers, anything already a customer, anything already in your CRM. Costs nothing and usually removes a large share of the list.
2. One cheap enrichment whose job is to disqualify. Not to enrich. To remove. A single inexpensive signal that eliminates a meaningful percentage pays for itself many times over before anything expensive runs.
3. The expensive waterfall on what survives. Waterfall logic already helps here: ask one source, and only pay the next one if it fails. Coverage goes up and cost goes down.
4. The model column last, on the shortlist. It is the most expensive per row and it produces better answers when the other data is already there.
Same table, same outcome, a fraction of the spend.
Two rules that keep the columns honest
One column, one question. A column that answers three things cannot be verified, and unverifiable enrichment is money spent on something you cannot use.
Mark unknown as unknown. Anything the source does not say should be recorded as missing rather than guessed. An invented value costs more than a blank one, because it gets said out loud to a prospect.
The controls that prevent a bad month
- A cap per run. A hard limit on rows processed.
- An alert on daily spend, not monthly. The expensive incidents are loops and accidental re-runs, and both are visible within a day.
- A test on fifty rows before any full run. Always. It catches the column that is wrong before it is wrong five thousand times.
The number to watch
Not credits used. Cost per qualified account.
Credits used falls when you optimise and also falls when you enrich less and qualify nothing. Only the second number tells you whether the table is worth running.
Track it monthly against the alternative: what the same qualified account costs through any other channel. That comparison is the one that survives a budget conversation.
When enrichment is the wrong answer entirely
When the data you need is already in your CRM and merely dirty. That is cleaning, not enrichment, and normalising what you own is far cheaper than buying it again.