What Is an Assisted Conversion?
An assisted conversion is a conversion where a given channel appeared somewhere on the customer's path but was not the last touchpoint before the sale.
An assisted conversion is a conversion where a given channel appeared somewhere on the customer’s path but was not the last touchpoint before the sale. It is not an additional sale. The same conversion can be assisted by four different channels at once, which is why assist counts across channels can never be added together.
Why assists change the argument
Last-click reporting flatters whatever sits closest to the transaction. Paid brand search and email get credited for demand that something else created weeks earlier, and the channels doing the introducing look like cost centres. Cut them on that evidence and you find out, two quarters later, that the cheap conversions were never cheap.
Assist data exists to stop that mistake. It answers a narrow question well: did this channel show up on the path at all? What it cannot tell you is whether the path would have completed without it, and that is the question a budget decision actually rests on. The distinction only stays visible when your reporting shows its working, including the model and the window that produced every number in it.
How the number gets produced
- Touchpoints are collected into a path. Every recorded interaction that precedes a conversion, in order, forms that conversion’s path. Anything the analytics platform cannot see, such as a session on a device it failed to stitch, is simply absent from it.
- A lookback window trims it. Only interactions inside the window count. Widen the window and assist counts rise across the board, with no change whatsoever in customer behaviour.
- The last touchpoint is separated out. That one takes the conversion under last-click. Every other channel on the path is recorded as an assist.
- The model decides the credit. Assist counts and credited value are different outputs, and how the credit gets divided is a modelling choice, not a measurement.
- Direct is handled by a rule. Platforms differ on whether a direct visit can take credit or is passed over in favour of the previous known source. That single rule moves assist totals more than most campaign changes do.
- Nothing here proves causation. Presence on a path is correlation. Only a holdout test, where you withhold a channel from a comparable group, gives you evidence that it caused anything.
Two of those steps are settings rather than facts. The lookback window and the direct-traffic rule are chosen by whoever configured the property, and neither one appears on the slide reporting the result. The same month can therefore produce two very different assist totals from identical raw data, without anybody having done anything wrong.
| Recorded path | Last-click credit | Recorded as assisting |
|---|---|---|
| Organic, email, paid, purchase | Paid | Organic and email |
| Organic, purchase | Organic | Nothing |
| Paid, organic, direct, purchase | Direct, or paid, depending on the rule | Whatever the rule leaves behind |
| Organic, organic, purchase | Organic | Nothing, despite two visits |
| Untracked referral, organic, purchase | Organic | Nothing, because the first touch was never recorded |
The blunt version
Assisted conversions are the number an agency reaches for when the last-click number looks bad. That does not make the metric dishonest. It makes the timing of its appearance the thing to watch: if assists were never in your deck until the channel underperformed, you are being shown a better-looking view of the same month.
Work the arithmetic. Say a store records 100 conversions in a month. Organic is the last click on 12 of them, and appears somewhere on the path of 60. One slide says organic drove 12 sales; the other says organic contributed to 60. Both are true, both describe the identical 100 conversions, and only one gets presented in a month where organic is under review. Add paid’s 45 assists and email’s 38 to organic’s 60 and you have 143 assisted conversions out of 100 sales, which is how a channel mix can appear to be over-performing on paper.
So ask three questions of any assist figure: which model produced it, what lookback window was used, and whether the same number was reported last month on the same basis. A supplier who cannot answer all three from memory is quoting an export, not an analysis. That test is the fastest way to judge a reporting dashboard properly without becoming an analyst yourself.
Example
Say a B2B software firm reviews its content budget. Last-click shows organic responsible for a small share of demo requests, and the finance team proposes a cut. The path report shows organic present on most of the paths that end in a demo, usually as the first recorded touch weeks before the form is submitted. That is a genuine argument for keeping the budget, but it is not proof, because those buyers may well have arrived by another route had the articles not existed. The defensible next step is a holdout: pause promotion in one comparable region for a quarter and compare. Assist data narrowed the question. It did not answer it.
FAQ
Can I add assisted conversions to my last-click total?
No, and doing it is the most common reporting error in this area. One conversion can be assisted by several channels, so summing assists counts the same sale repeatedly. Totals built that way routinely exceed the number of orders the business actually took, which is a useful red flag.
Why did our assisted numbers change without any campaign change?
Usually the lookback window, the attribution setting, or the platform’s handling of direct traffic changed underneath the report. Consent rates and cross-device stitching also move these figures. Ask what the settings were on both dates before you accept that customer behaviour was what moved. Configuration drift is the more common cause.
Do assisted conversions prove a channel is worth funding?
They prove the channel was present, not that it was necessary. Presence on a path is correlation. The only evidence that survives scrutiny is an incrementality test: withhold the channel from a comparable group, run it long enough to matter, and compare conversion rates between the two.
Related terms
- Attribution Model — the rule set that decides how much of each sale a channel is credited with.
- Retrieval-Augmented Generation — how AI assistants build answers, and why some paths now begin somewhere your analytics never sees.
- llms.txt — a proposed file for AI crawlers, relevant here because AI-referred visits often land in reports as direct.
If the assisted number only appeared in your deck the month last-click looked bad, it is a rescue metric, not a finding. Ask which model and which lookback window produced it, and ask before you approve the budget.