You sponsored an expensive conference and had a conversation with a promising lead. You added the lead to your marketing database, started sending her newsletters, and she started commenting on your LinkedIn posts. A few weeks later, one of your customers recommended your product to the lead. A day later, she opened your newsletter, clicked one of your Google ads, submitted a demo request, and eventually signed up as a new customer. Congrats!
Which of these marketing campaigns (conference, newsletters, LinkedIn, customer referral, Google ad) should be credited for bringing in the new customer?
Attribution, for the purpose of this post, is the process of allocating credit to the marketing campaigns and channels that created a new Sales Opportunity, in order to analyze their effectiveness and ROI (Return on Investment).
If you have some experience dealing with attribution, you probably heard or experienced first hand that it’s a complex process that doesn’t accurately describe reality. In this post, we’ll try to simplify attribution by recommending a certain attribution model which I believe is appropriate for most B2B startups.
It’s important to note that our recommended model isn’t perfectly accurate, either. Instead, we believe in the 80/20 rule, and in trying to maximize the value we can get out of an attribution model, while minimizing our investment of time, money, and mental effort.
If it’s so complicated, why should we invest in attribution at all?To compare performance across and within marketing channels and to understand what’s working, especially for expensive channels and campaigns.
It’s important to note (again) that because attribution is never going to be perfectly accurate, it’s better to focus on comparing metrics across channels or over time vs. focusing on absolute numbers. For example, if our conference ROI is 0.9x, it’s not very effective to obsess over this number. It’s better to observe that in the past year our overall conference ROI went down from 2.1x to 0.9x, and to realize we have an issue with our conference strategy or execution.
Cool, so what should we do?
You may have heard of various attribution models, starting with the simple Last Click and First Click, to the (seemingly) more sophisticated Time Decay or Position Based. Google Analytics has a concise article describing the different models. And beyond a singular model, there are even whole products claiming to build the perfect mathematical attribution representation to fit your unique setup with 100% accuracy.
Our attribution model recommendation aims to maximize the following factors:
- 80/20. The model won’t be perfect, but it’ll reasonably explain 80% of cases – with 20% of the investment. You’ll be getting high ratio of benefit vs. cost.
- Simple, quick, and easy to understand and implement. It could be that a super complex, expensive model will be more accurate, but is it worth the extra investment? Will we be able to explain it to the CEO and the sales team? Will it take months to implement, during which we won’t be able to get any information about our campaign performance?
Our recommendation for B2B startups (especially for those just starting out with marketing, but not only them) is to go with Last Click attribution. Namely, the last channel or campaign a lead engaged with before turning into a Sales Opportunity gets the credit.
Why Last Click?
- It’s intuitive. Think about it – our goal is to credit the marketing activity that transitioned the lead from passive to active interest in our product. It makes sense that the campaign the lead engaged with just before she became an active lead was what tipped the scale and made her agree to a sales call or a demo. Of course, this isn’t always the case, and there are more complex situations which we’ll cover in a bit. But if we go back to the 80/20 argument, there’s a reasonable chance that we’ll be right for 80% of cases.
- It works well for smaller startups. Early on, there isn’t a ton of marketing activity going on. So it makes sense to assume that if you publish a new blog post once a month, the opportunities that open in the following days come in thanks to the post, and not thanks to the trade show you attended six months ago. When your company is bigger and in one week you attend a trade show, publish a post, and run multiple display retargeting campaigns, it’s harder to argue that the post brought in the opportunity (more about this later).
- It is relatively easy to understand and implement. It’s easy to explain to other people in the company how the model works. You don’t need dedicated attribution products or even a sales or marketing platform like Salesforce, Hubspot or Marketo to implement Last Click. In the very early days, you can simply add a “Source” column to the spreadsheet you use to track sales opportunities, and use it to document the channel or campaign that brought in the opportunity – e.g. “LinkedIn post about product launch,” or “Referral – David.” If you are using the tools I mentioned, they all offer similar functionality. Initially, you can tag opportunities manually, and later on implement automations (e.g. bulk-tag all trade show leads, and assign the trade show as the opportunity source once an opportunity is created).
For all these reasons, if you chat with other companies (and not just startups), you’ll find that the vast majority of them use Last Click. When I worked for the world’s largest e-commerce company whose name rhymes with Glamazon, we used Last Click all the way. When I later worked for a company that sold software to hundreds of large e-commerce retailers, we discovered that despite a lot of talk about sophisticated attribution models, everyone (literally every single company) was using Last Click.
When should we maybe not use Last Click?
- When you’re running more channels. Like I mentioned, when your marketing activity expands and you will simultaneously run multiple campaigns across channels, you might need a model that credits more than one campaign.
- High ACV (Annual Contract Value) or deal size. An expensive product (say >$50-100K / year) usually means a long, complex sales process. We might need to invest in several campaigns to get to the Opportunity stage: For example, we met the lead at a conference, she received a newsletter, she attended a webinar a couple of months later, and after a week she responded to a prospecting email and agreed to get on a demo call. It makes sense to credit more campaigns beyond the final prospecting email. In contrast, for lower price products, it’s pretty common for a lead to agree to a demo call after one cold email – so the conference from six months ago doesn’t need to be credited.
In that case, what to do if not Last Click?
Our recommendation is Linear / Uniform Distribution – meaning equal weight / credit for every campaign the lead engaged with. In the prior example, the conference, webinar, newsletter, and email will each get 0.25 credit for the opportunity.
Wait, but why equal credit? If we had an hour-long conversation at the conference, shouldn’t the conference get higher credit than a newsletter the lead skimmed for ten seconds? Yes, maybe. And still, in the spirit of KISS (Keep It Simple, Stupid) we still recommend uniform distribution.
One of the companies we worked with switched from Last Click to Uniform Distribution. To prepare, we reviewed a few dozen opportunities with the VP Sales. We heard the “story” of each one, and we gave precise credit to each campaign according to the story. In the previous example, we could have given the conference 50% of credit, webinar – 30%, email – 15%, and newsletter – 5%.
After summarizing all campaign credits, we discovered the results of this detailed and “accurate” exercise were very similar to uniform attribution… in other words, the differences in importance between the different campaigns averaged out across all opportunities – sometimes the conference was a major driver, other times it was minor, and ultimately it contributed equally across all opportunities.
The results weren’t identical, but again, in the spirit of 80/20, we decided we’re better off going with Uniform Attribution that would accurately describe 80% of reality for 20% of the investment. It’s much easier to implement an automation that evenly distributes credit vs. a process that attempts to assign every campaign the specific percentage point it “deserves.”
The bottom line
Attribution isn’t fun, but it’s important. Our recommendation? Go forth with Last Click. Start by manual implementation, and add automation over time. If you start feeling like there are too many cases where the model doesn’t represent reality, consider switching to Uniform Attribution.
