An attribution model is the rule your reports use to decide which marketing touch gets credit for a lead or a sale. Pick the wrong rule for the decision in front of you and you will cut a channel that feeds the pipeline, or keep funding one that only collects credit at the end. This article answers two linked questions: which marketing attribution model should a 10 to 250 person service business use for each decision, and how do you build the reports on the levelpartner.ai platform so every number says which model produced it?
The attribution models in digital marketing that matter
Every model divides one sale among the touches that came before it. None is objectively correct. Each carries a bias, and the skill is choosing a bias that suits your sales cycle and the decision you are making.
- First touch: all credit goes to the channel that introduced the person. It favours discovery channels such as organic search, content and social.
- Last touch: all credit goes to the final interaction before the enquiry. It favours email, branded search and retargeting.
- Linear: equal credit to every recorded touch. It favours channels that appear often in the middle of the journey.
- Time decay: more credit to touches close to the sale. Think of it as a softer last touch.
- Position-based: most credit split between the first and last touch, with a small share for the middle. A compromise for longer cycles.
Data-driven models estimate weights statistically. They need far more conversions per channel than a firm closing a few hundred deals a year produces, so the weights move with noise rather than with real change. Know they exist and leave them alone at this size.
What the platform records
The platform keeps two attribution touchpoints on each contact. First attribution is the contact's first recorded interaction and never changes. Latest attribution is the most recent one and updates whenever a new tracked interaction arrives. You can read both on the contact record, and every Contact or Opportunity widget on a dashboard can be filtered or grouped by either one, using session source, medium or campaign values.
That gives you first touch and last touch directly. The activity history on each contact holds the touches in between, so linear, time decay and position-based credit can be calculated from an export when a decision needs them. The linear attribution model guide walks through that calculation step by step.
One decision, one model
The real mistake is choosing a single model for the whole business. Different decisions ask different questions.
- Where should next quarter's awareness budget go? Use first touch, because you want to know what introduces people to you.
- Which campaigns close demand that already exists? Use last touch, because you want the touch that turned interest into an enquiry.
- Is a nurture sequence pulling its weight? Look at the touches between first and last, where a linear view helps.
- Is a channel worth keeping at all? Look at both ends and keep it if it scores well on either.
A working default for service firms
For agencies, consultants, mortgage brokers and real estate firms, start with first touch for budget and last touch for tuning campaigns. Both come straight out of dashboard widgets, both are easy to explain to a partner or a board, and between them they answer the two questions owners actually ask: where do clients come from, and what makes them act?
Move the budget model to position-based only when your cycle is long enough that the middle clearly matters, such as a consultancy selling programmes that take months to agree. Most firms of this size never need more than the pair of single-touch views plus a quarterly position-based check.
Assume a median cycle of 34 days and six tracked touches per won client. Under first touch, suppose organic search earns £140,000 of the year's won work and email £20,000. Under last touch the order could flip: email £110,000, search £45,000. A position-based split might give search £95,000 and email £70,000. On those numbers, position-based would suit the budget model because neither channel gets written off, with last touch kept for weekly campaign tuning. All figures are illustrative.
Before you start
- Forms, calendars and chat running on the platform, or the platform's tracking code on your external site.
- Campaign parameters on every paid and email link, using one naming convention that never changes mid-campaign.
- Opportunity values entered and updated when deals close, so won value means something.
- Your median sales cycle in days and the rough number of touches a won client has before signing.
- A short list of the budget decisions leadership will make with these reports, and one named person who decides the model.
How to set the reports up
- Write each budget decision next to the model it should use.
- Go to Dashboard and open the dashboard you want to customise.
- Enter Edit mode and add an Opportunity widget, such as Won Opportunity Value.
- Go to Conditions, add Attribution and choose First Attribution.
- Open the Group By control and select Session Source.
- Review the visualisation and save the widget.
- Add a second Opportunity widget and repeat with Latest Attribution.
- Click Save Changes to keep the dashboard layout.
- Title each widget with its model and the decision it serves.
- Compare both widgets each quarter before any budget change.
Keep all conditions inside one filter group. Adding a second group removes session source and medium from the grouping options, which is the usual reason a widget suddenly refuses to group by channel.
What changes when you hold two views
Seeing both rankings side by side is usually the turning point. Content and social tend to rank higher on first touch. Branded search and email rank higher on last touch. Neither is wrong. A channel that scores well on first touch and poorly on last is introducing people who later convert through other routes, and cutting it shrinks the pool every other channel draws from.
Say last touch credits branded search with £400,000 of won value and LinkedIn content with £50,000, while first touch gives LinkedIn content £260,000 and branded search £120,000. A firm reading only the last-touch view might plan to cut content. With both views in front of it, the better move would be to keep content funded and trim paid search on its own brand name. These amounts are invented.
What most advice on models gets wrong
Most writing on attribution presents the choice as a search for accuracy. It is not. Every model is wrong in a known direction. The value lies in agreeing, before anyone sees the results, which wrong you will live with. Once that is written down, budget arguments stop depending on whoever picked the view that flatters their channel.
Changing model every time the results are uncomfortable, or never noticing which model a report uses at all. A model becomes useful only when one quarter can be compared with the next under the same rule. Pick one, write down why, put the model name in every widget title and leave it for at least two quarters. If you want another view, run it alongside, never instead.
The trade-off you accept
Using two simple models instead of one sophisticated one means you never get a single number per channel. You give up the comfort of one ranking. In return, every manager can understand the models and check them against individual contact records, which matters more at this size than statistical elegance.
Time it takes
Choosing the model takes one two-hour session with the owner and the sales and marketing leads. The dashboard takes under an hour to build. After that the work is quarterly: an hour before each budget review to compare the views and note why each change is being made, plus one to two hours if you also run a calculated position-based split.
When not to rely on a formal model
If most new work comes from referrals and existing relationships, the touches are not recorded and the model matters less than capturing the referral source on every enquiry. If you close only a handful of new clients a quarter, list them by name with their source and skip the model. Small volumes let one deal swing any ranking, so judge trends over several quarters.
Troubleshooting
First and latest attribution are identical for most contacts. People enquire on their first visit, or later visits are not recorded. Check the tracking code is on every page of an external site.
Direct traffic dominates the latest attribution view. Returning visitors type your address. That is normal. Use first attribution to see what originally brought them in.
Each team prefers a different model. That is the bias problem in action. Agree the model before looking at results, not after.
Results swing sharply between quarters. Volume is too low for the model chosen. Widen the date range to a rolling six months.
Frequently asked questions
Which attribution model is best?
There is no best model, only a best model for a decision. Use first touch for awareness budgets and last touch for demand capture, and look at both before cutting anything.
Does the platform support multi-touch models?
It stores first and latest attribution on each contact and keeps the activity history, so multi-touch credit is calculated from an export. Multi touch attribution explains how to capture the touches those models rely on, and the case for last touch covers when the simplest view is enough.
Can we use different models for different channels?
No. Credit has to add up to total revenue, and mixing models double counts. Use one model for budget decisions and treat other views as background.
Which model do HubSpot, Salesforce and Pipedrive users typically rely on?
Teams on any CRM tend to start with first and last touch, because those are the simplest to report. The model choice does not depend on the CRM. What matters is that sources and revenue sit in the same system, which is the groundwork described in marketing attribution basics.