The State of First-Touch, Last-Touch, Self-Reported Attribution and Revenue
In this report, we analyzed over 8,500 self-reported attribution answers from 36 B2B SaaS companies to discover the differences in attribution models and what works and what doesn't in B2B SaaS.
All HockeyStack Labs reports are done using anonymized HockeyStack customer data. We did not partner with anyone on the creation of this report and it was not sponsored by a vendor.
In the beginning of this year, we decided to publish our first Labs report on the relationship between last touch and self-reported attribution, knowing the results would help us better understand the effectiveness of our GTM motions.
Following the initial report, we decided to create a second chapter. Here, we will examine not only first touch, last touch, and self-reported attributions but also their translation into revenue.
We know that attribution will never be perfect. However, this doesn’t preclude the use of or trust in attribution; it simply means that different attribution models need to be applied case by case. The insights from attribution must be considered alongside self-reported attribution. We cannot rely solely on one or the other.
It’s not a simple equation. In psychology, the concept of ‘selective recall’ illustrates how individuals may emphasize certain memories over others. This is a phenomenon where people remember or focus on certain parts or information while forgetting others. Think of relationships, you might reminisce about certain aspects and forget the other parts after a break-up – depending on the emotional state, coping mechanisms, and the narrative you’ve constructed about the relationship, selective recall plays a significant role in how we remember our past experiences. This principle of selective memory is not just limited to personal relationships but extends into the realm of B2B SaaS marketing as well.
In marketing, we can call this concept “recall bias” — a cognitive bias where individuals fail to accurately remember past events or experiences. Recall bias undermines the reliability of self-reported data, as people tend to forget, oversimplify, or distort their memories over time. This highlights the crucial role of integrating multi-touch attribution with self-reported attribution to gain a comprehensive understanding of the customer journey.
Therefore, leveraging multi-touch attribution in conjunction with self-reported attribution is essential for fully capturing the customer journey. For this reason, this report is structured into four parts. The first part examines the relationship between first-touch and last-touch attributions. The second part incorporates self-reported answers into the analysis. In the third part, we analyze the journeys of the self-reported answers; and in the last part, we correlate this data with revenue metrics.
You might notice discrepancies with the numbers presented in the initial chapter of this report. The reason for this is my intention to refine our data grouping approach. Consequently, I conducted a detailed analysis of the responses individually to determine if some could be reclassified.
The results of this reevaluation are evident in this second part, where the contribution from social media has been adjusted from 19.58% to 13.68%. Furthermore, new categorizations have been introduced, enabling clear distinctions between channels such as LinkedIn, Facebook, Twitter, YouTube, and TikTok.
Methodology
Traditional attribution models: single-touch attribution, first-touch, last-touch.
PPC: Google Ads & Bing Ads.
Search: This term is used to categorize the self-reported attribution answers which could be Direct, Organic, or PPC (since we can’t really know what people mean when they say “Google,” “Bing,” unless they specifically say “Google Ads,” which they rarely do).
Referral: G2, Capterra, Partner websites.
MQLs: High-intent demo, pricing page, contact us submissions. Basically every hand-raiser on the website. Ebook form submissions, lead-gen stuff, webinar registrations were not counted as MQLs. In the previous report, some people got a bit heated with us for saying MQLs, but I won’t be changing this, because I like the sound of it.
Data sample: Over 8,500 answers from 36 B2B SaaS companies.
Date range: Feb 1st, 2023, to Dec 31st, 2023.
Part I: What do we see in First-touch vs. Last-touch?
As expected, in both the first-touch and last-touch results, we barely see anything. This serves as a great example to show that traditional attribution models are far from enough when dealing with long sales cycles and multiple stakeholders.
In both models, we’re seeing that more than 80% of the results are either Organic, Direct, or PPC. From first-touch to last-touch, we see that Direct decreases from 30.44% to 26.56%; Organic decreases from 30.84% to 23.39%; and PPC increases from 22.31% to 32.90%.
I believe what we’re observing here is that after the initial discovery, most people search the brand name on Google, then click on the first result, which in most cases is the brand ad, and then submit the demo form. Hence, we’re witnessing a significant increase in PPC at the last-touch level, leading to the perception of an exceptionally high ROI (!) from brand campaigns.
Relying on these traditional models essentially shows us that 8 out of 10 conversions are Organic, Direct, and PPC – leaving us clueless about the effectiveness of other paid channels or whether sponsoring an event was beneficial. It means we’re essentially flying blind, opening the doors to run random experiments, test theories without hypotheses, and spend money just for the sake of it – a perfect world for mediocre marketers.
Thankfully, we get to see some variety among the rest of the sources.
LinkedIn, in total (both ads and the company page), accounts for around 9% of the first-touch conversions and 12% of the last-touch conversions. I’m not sure how we should classify the company page conversions; are they organic or paid? If you have a strong employee advocacy program and your employees constantly post about your company, then your audience might end up visiting your website through the company page and convert—this would make it organic. But also, if you are promoting your company page posts and people end up visiting your website through the company page and convert, this would make it paid. In any case, though, we’re seeing that LinkedIn comes as the fourth most popular source in both attribution models.
Next, we see referrals and review sites—so this might be your partner websites, Capterra, G2, etc. This is a bit interesting because we’re seeing a huge discrepancy between the first touch and the last touch results here; while referrals are bringing in website visits and definitely influencing conversions, they don’t directly impact the number of MQLs. We’re observing a similar pattern in Display Ads and YouTube data. These channels influence conversions, but they don’t bring direct results.
On the other hand, email marketing seems to have an opposite pattern. On the first-touch side, we observe only 0.43% of conversions attributed to the email marketing; however, on the last-touch side, this number nearly triples. My assumption is that the email nurture workflows contribute to this increase; your audience visits your website, possibly downloads something, signs up for a webinar, or leaves their email in some way but doesn’t directly convert. Then, they’re included in an email nurture track and convert after receiving some emails.
I think this needs to be taken into account together with what we found in the touchpoints report where now we have the data to prove that having an email marketing strategy actually decreases the number of touchpoints required to generate an MQL and speeds up the sales cycle.
Apart from these, we’re seeing pretty similar numbers in both models when it comes to Facebook, Twitter and Reddit.
Part II: First Touch & Last Touch & Self-Reported Attribution
Now, what happens when we blend this data with the SRA answers?
When it comes to SRA, it’s important to highlight that some data cannot be categorized under the categories we have for the first or last touch models. For example, while Direct, Organic, and PPC constitute 80% of the data in traditional models, since people don’t typically type “organic,” “direct,” or “PPC” into “how did you hear about us” fields, we can’t really match this data. For instance, only 0.5% of people mentioned “Google Ads,” “Adwords,” or “PPC,” but many people said “Google” and “Bing.” Does this imply these people discovered the company organically on Google? Possibly not.
Therefore, similar to the first report, we categorized those answers under “Search”—this includes all explicit mentions like “search,” “Google,” “Bing,” “keyword search,” “internet,” “Adwords,” etc. This accounts for 45% of the SRA data, which could be either organic, direct, or PPC. We’ll delve into this later.
Let’s start with the stuff that we can match.
LinkedIn Ads: When we look at the SRA answers, at first, it appears more effective than what the traditional attribution models show. However, we need to keep in mind that there was literally no answer saying “Linkedin company page”; therefore, this 13.59% most likely includes the company page conversions as well. Hence, although there’s a 1.5x difference between first-touch and last-touch, we don’t really see a difference between last-touch and self-reported attribution data.
Referral/Review Sites: We definitely see more answers mentioning G2 or Capterra in SRA than what the last-touch data shows. This part is interesting though, because there are 306 results at the first-touch level, then just 34 results at the last-touch level, and 154 results at the SRA level. Yes, your audience discovers you through review sites and through those categories, but this number is apparently not as significant as one might expect.
Facebook Ads: This is interesting to me. While the first-touch and last-touch data were pretty similar, showing 166 and 195 submissions respectively, on the SRA side, only 20 people mentioned “Facebook.” The main reason for this could be that Facebook is mostly used for retargeting; your audience doesn’t typically discover your brand through Facebook, but rather, they convert through Facebook; but this still doesn’t explain the first-touch side. But this still doesn’t explain the first-touch side. Another explanation could be the SRA answers; maybe people refer to Facebook as “Social Media” more than they refer to LinkedIn as “Social Media”? We’ll examine this in the third part.
Display Ads: We’re seeing a similar pattern here as we do on Facebook, but there’s a slight difference: on Facebook, we’re actually seeing more last-touch conversions than first-touch influence; on Display, we’re seeing more first-touch influence than last-touch conversions. This might mean that display ads definitely bring traffic to the website, and this traffic, in a way, converts into MQLs—we’ll see if these are actually qualified or not in the next part.
However, most people actually don’t remember seeing display ads. Some might argue that this phenomenon is similar to what we see on the PPC side, where people don’t necessarily say “Google Ads,” but this doesn’t mean that PPC doesn’t work. When they say Google, it might mean PPC; but PPC natively happens within Google Search. Display doesn’t necessarily happen on Google; it could be on other external websites, therefore it wouldn’t be a fair categorization to combine Display and Google.
Email Ads: Similar to LinkedIn ads, the last-touch data and SRA data align closely. This validates the hypothesis. Your users most likely leave their emails on your website, then they are included in the nurture tracks, and subsequently, they convert. Since those emails effectively convince them to take action, we observe a similar volume of SRA responses.
Here’s where it gets interesting.
Let’s circle back to the start. Traditional attribution models showed that more than 80% of the conversions were classified as either Direct, Organic, or PPC. Now, blending this with the self-reported attribution answers reveals a new perspective.
We see that Search, Content, and Website account for 47% of the conversions—a significant figure, indeed, almost half. Yet, this is significantly lower than the 83% we see in the traditional models. Essentially, there’s a 36% discrepancy between the traditional attribution model data and the self-reported attribution answers.
So what is this 36% about?
Hence, based on more than 8500 conversions, we can assume that dark social constitutes 36% of your top-of-funnel conversions.
Now, let’s spice things up a bit. How reliable is this data, really? Like, do we actually trust these answers? Can we really think everyone remembers exactly how they found us, or is it just the last bit they remember, or maybe just the bits that stuck out to them?
Obviously, word of mouth, events, or podcasts might have influenced these journeys, and I don’t mean to underrate them – but could there be something else?
36% of the total answers correspond to slightly above 3k people, so 3k form submissions.
What else can we analyze here?
We can analyze if these people had been seeing ads on LinkedIn throughout their journey…
And yes,
Apparently, more than 90% of these conversions actually had seen LinkedIn Ads (at least 100 impressions) within the last 30 days before submitting their demo forms. So okay, they say “word of mouth,” “podcast,” even “outbound” – but for most of them, LinkedIn has been a major part of their conversion journey. Interesting, isn’t it?
These impressions might as well be retargeting; maybe they already visited your website and then got retargeted on LinkedIn. So, LinkedIn didn’t cause the conversion, but it did speed up the sales cycle; or they were actually in the cold audience and were seeing your ads – but they didn’t necessarily remember your brand until they heard it in a Slack community, or in a podcast.
In a nutshell, there are two things to highlight in this section:
1- On average, 36% of your conversions are from dark social.
2- LinkedIn still plays an integral role in your dark social conversions.
Part III: The Relationship
Okay, so 36% of the conversions could be attributed to dark social and cannot be easily tracked through traditional attribution models. In this part, we’re going to look at what the first-touch and last-touch attributions show compared to what people say in the self-reported answers.
Let’s start with the word of mouth part – this part contains all answers including but not limited to word of mouth, Slack, communities, Facebook groups, etc. Out of over 8,000 answers, 747 of them were word of mouth.
According to first-touch data, 90% of these conversions were either direct, organic, or PPC; on the last-touch side, the results are similar, at 88%.
The PPC side is worth mentioning, where we see an average of 27% “word of mouth” conversions attributed to PPC – there’s a big chance that most of these were brand conversions, like someone hears about your company in a community, googles the name, clicks on the first result. Brand campaigns always have ridiculously high ROI for a reason. When we look at this data, it feels like if you turn off your brand campaigns, your audience is still going to find you.
There’s a caveat though, if you’re in a highly competitive market and your competitors are bidding on your name. Letting them have the first position might mean that your audience will discover them as well. Like if someone searches for your brand and sees another name, then they will most likely check them out as well. That’s a risk.
Almost all of the YouTube SRA answers were either Direct, Organic, or PPC according to the first-touch and last-touch data. The same applies for Events, Podcasts, Colleague, ChatGPT, and Reddit answers.
However, we’re seeing a much wider variety in the Social Media answer, which corresponds to almost 14% of all the SRA data.
This time, Direct, Organic, and PPC correspond to 55% of the total traditional attribution data; followed by LinkedIn with 28% on average, and Facebook 15%. I believe this is important because this data validates that although only 0.24% of the SRA answers mention Facebook, in reality, this figure is way higher than what people say or remember. Similarly, although we’re seeing around 14% of the SRA answers being LinkedIn, we can easily assume that in reality, at least 20% of the total conversions are influenced by LinkedIn.
Lastly, I want to look at the SRA data that contains LinkedIn.
Similar to what we often saw previously, 70% of the conversions that contained LinkedIn in their self-reported answers were actually attributed to Direct, Organic, and PPC. More importantly, on the last-touch side, 25% of these conversions were attributed to PPC. Hence, relying on the traditional last-touch model would have basically shown LinkedIn underperforming.
Another interesting point is the referral part; 3% of the conversions were attributed to referral on the first touch. So, probably someone first heard about your company on G2, visited your website, didn’t convert, and then was retargeted several times before finally converting.
Fun fact: Seems like all of the unqualified SRA answers were from display ads…
Part IV: Revenue impact
Okay, we’re seeing all these different answers and stuff, but what do they actually mean? What brings in revenue?
As you might remember from the Recap report, we found out that the average MQL:CW rate was 5.94% among our customers, and the ACV was $20.8K.
Let’s do some basic math. We’ve analyzed 8500 answers, with a 5.94% win rate, this means over 500 closed-won deals, and with a $20.8K ACV average, this leads to $10.5M in added revenue.
So what percentage of this $10.5M revenue was generated from what?
According to the self-reported attribution answers; previous customers make up only 1.5% of the total MQLs; but on the revenue level, this 1.5% of the total MQLs actually equals 7.06% of the revenue.
This means that not only do these previous customers have significantly higher conversion rates but also significantly higher ACVs.
We’re seeing a similar impact when it comes to email marketing, LinkedIn, word of mouth, review sites, events, and Facebook.
While only 1.16% of the SRA answers contained email, at the revenue level, email marketing makes up 2.24% of the total revenue.
Only 13.59% of SRA answers contained LinkedIn, but this accounts for nearly 21% of the total revenue.
Word of mouth answers were 18.1% of the total data, but this corresponds to 23.77% at the revenue level.
Although only 0.5% of the self-reported answers contained “Events”, the ROI seems to be remarkable. With more than 3x ROI, events seem to have the second-best ROI after previous customers.
Search seems to have a negative ROI— here it’s not only PPC but it might also mean direct and organic—although Search brings almost half of the conversions, it brings just a bit more than a quarter of the revenue.
When we add all the numbers together, dark social makes up 36% of the total conversions and accounts for almost half of the revenue (48.4%).
Search accounts for almost half of the total conversions but makes up only quarter of the revenue. (27.5%)
LinkedIn is expensive but it works.
Events have more than 3X ROI on the revenue level.
Past customers have more than 5X ROI on the revenue level.
Conclusion
When I took over writing the second chapter of this report, I was kind of hoping it would be very straightforward, but I’m pleasantly surprised with the amount of data that we can pull and analyze.
This report once again proves that traditional attribution models should only be used to understand certain parts of the journey but definitely not for understanding the complete journey. On top of this, self-reported attribution answers are crucial, but they definitely don’t show the complete journey; multi-touch and SRA should be used together (that’s what we do when analyzing journeys on HockeyStack).
Similar to what we saw in the previous reports, where we analyzed a bunch of different stuff, search is getting less profitable, and LinkedIn is becoming more and more crucial for B2B SaaS – more importantly, though, this report also validates the importance of not solely relying on paid channels as we can clearly see the impact of previous customers, events, and communities.
This doesn’t mean that you need to start a community; if you’re selling to marketers or salespeople, there’s an inflation of communities. Just make sure that you’re in these communities and help people out rather than building your own – unless it’s a customer community. However, if you’re selling to other personas where there aren’t many communities, then starting one might be a good option. Apart from that, having a strong event marketing and customer marketing framework seems to be really good ways to improve revenue.
It’s easy to say “let’s increase the budget by 2x and expect to see a 2x in revenue” when you’re trying to scale, but this approach is not sustainable and has never been sustainable. Increasing revenue is not just about paid – and this is coming from someone who gets paid by doing paid – but it’s about building a strong growth marketing strategy and tracking this with multi-touch and self-reported attribution together.
Once again, self-reported attribution by itself is not the answer, as we can clearly see that people don’t really remember their actual journeys, and they won’t ever remember. Think of the word-of-mouth self-reported answers and how 90% of them were actually served LinkedIn ads in the last month. We need to keep in mind that people tend to remember only some parts of our journey, it’s the recall bias.
I hope you enjoyed this report and let me know if you have any suggestions for what the next reports should be about.