Waiting for perfect attribution costs more than deciding with imperfect attribution
No marketing attribution is fully accurate. See why, what a good enough number is and which budget decisions it already supports.
Published on
5 min read

No marketing attribution model gets everything right, and anyone who waits for the perfect number before deciding spends the quarter deciding by gut feeling. An imperfect number, always calculated the same way and with known limits, already supports most budget decisions. What it cannot do is change its rules at every meeting.
The difference between models, such as first click and last click, was the subject of the article on how to choose an attribution model. Here the point comes before the choice: why none of them closes the full picture and how to decide anyway.
Why perfect marketing attribution does not exist
The buying journey passes through places that tracking cannot reach, and each one takes a piece out of the number. The most common are these:
- Cookies and consent: a person may decline cookies on the site's banner, and then their next visit looks like someone new.
- Multiple devices: they see the ad on their phone and buy from their work computer, and the two visits look like different people.
- WhatsApp: the conversation has no page address to carry the UTM, and the source is only recorded if the click identifier of the WhatsApp ad is saved at the first message.
- Long cycle: in a sale that takes months, the click that brought the lead may fall outside the conversion window configured in the platform.
Add to that the fact that each platform counts its own contribution. Google Ads sees the clicks on Google and Meta sees the clicks on Meta, which is why Meta Ads data does not match the CRM.
Even the platforms' rules change. In the official Google Ads help on attribution models, Google states that it stopped offering the linear, position-based, first-click and time-decay models. What remains are last click and the data-driven model, which is the default for most conversions.
What counts as good enough attribution to decide
Good enough marketing attribution is consistent over time and has its limits written down. Consistency means the same model, the same window and the same UTM rules every month. When the error is similar from one month to the next, comparison between periods and between channels remains valid, even if the total does not add up.
Imagine, in a hypothetical example, that 30% of sales arrive with no identified source every month. If that share stays stable, a channel that went from 20 to 30 attributed sales truly grew. If the share with no source jumps to 50% in the same month, the comparison loses its value until someone finds out what changed.
The written limits fit in one line of the report. They keep anyone from treating the number as exact:
- the model and the conversion window used;
- the share of sales with no identified source in the period;
- the channels left out, such as referrals and WhatsApp conversations without an identifier.
Resource · In PortuguesePrompt Guide for Marketing and Sales70 prompts with questions about marketing, sales and behavior to ask your data.Download the guide
Which decisions can withstand an imperfect number
Decisions based on large differences and on trends withstand an imperfect number well. Those that depend on small differences call for more care, because the known error may be larger than the difference.
Decisions the imperfect number supports
- Reallocating budget between channels when the cost per sale of one is much higher than the other's.
- Cutting a campaign that spends a lot and generates no opportunities by either first or last click.
- Tracking whether a channel grows or declines over the months, with the same rules.
Decisions that call for more care
- Choosing between two channels with close results, when the difference falls within the known error.
- Cutting top-of-funnel channels, such as reach and content, which last click tends to underestimate.
- Raising a channel's budget by a lot, because the average return does not show the return on the next dollar invested .
- Judging a new channel that still has few sales.
In these cases, it is worth cross-checking the platform's number against the source recorded in the CRM or running a short, time-boxed test before touching the budget. It is faster than waiting for a better model.
The question the C-level should ask instead of “is this number 100% right?”
Ask what the number leaves out and whether that gap changes the decision. The question about 100% always has the same answer, and the meeting ends without a decision. The questions below lead to a data-backed decision, even with the imperfect number:
- How much of the sales has no source, and has that share changed since last month?
- Are the model and the window the same as in the previous report?
- If the error is double what we estimated, does the decision change?
If the answer to the third is no, the marketing attribution you have is already enough and the decision can be made in the same meeting. These questions add to the business questions that the marketing and sales report still does not answer. For whoever puts the material together, it is worth following a paid media report that ends in a decision, with the limits of the number written on every page.
Frequently asked questions
What is marketing attribution?
It is the method that splits the credit for a sale among the channels and ads the customer interacted with before buying. The attribution model defines the rule for that split, such as giving everything to the first click, to the last one, or distributing it across the history.
What is the best attribution model?
There is no single best one for every case. The best one is the one that answers the question behind the decision at hand and that the operation can keep the same from one month to the next.
How much of sales with no source is acceptable?
There is no universal number. What matters is that the share stays stable and is reported in the report, because a sudden change in it distorts any comparison between periods.
How can you improve marketing measurement without waiting for the perfect model?
Standardize the UTMs, save the click identifiers in the lead's record and register the source in the CRM at the moment of conversion. Each gap closed reduces the share with no source without changing the comparison rule.
To get the formula and a calculation example for every funnel metric, from cost per lead to CAC, download the Laiki Metrics Guide.


