Die Uniques! Die!!!

I was playing golf this weekend with a friend who does SEO. We were talking about how things were going, joking about how slowly big companies move, and the conversation came around to how, at the end of the day (because of how the company is structured, not because it’s his decision), he was just trying to, “bring in as many uniques as possible.”

Will uniques not flipping die?

The world doesn’t care about uniques, and we prove that time and time again. Uniques don’t care about uniques: that is to say people are not influenced by a binary function of whether they have or have not been exposed to a message. People need messages repeatedly; we are conditioned to ignore things because we are constantly bombarded with them, and it takes a lot of the same message to get us to get the point of conversion.

And, funny enough, advertisers don’t care about uniques either! Yes, it’s true.

All you need to do to prove this is look at how people advertise. Yes, they will buy networks with high numbers of uniques that fit certain demographic or behavioral needs. But what happens when they buy the SECOND placement? Wait, SpikeTV has millions of young, outdoor sports loving men (uniques) who love beer and nacho cheese. So why did we then also buy Versus during the Tour? And MTV during Jackass? And, and, and… aren’t those likely the same people?

Clearly, advertisers don’t give an airborne feces about the uniques, either; they care about the exposure, dare I say the REPEATED exposure, to a target market. Advertisers like Coke and Pepsi hit you hundreds of times a day. They are not trying to reach more people. They are trying to reach the same people over and over again, because that’s what it takes to make a brand stick and get you to be a good customer and a hopeful advocate.

Yes, advertisers to need to ensure they’ll have some reach. But if you have millions of visits, reach is implied. Obviously, you don’t have 8 people in their mom’s basement racking up a million visits. So get over the “safety net” of uniques=reach and start realizing that high visits with high value per visits (like a visit to CNN that has a high content velocity / high amount of content consumption) is the only thing you need to care about as an advertiser or publisher. These visits give advertisers and publishers alike multiple opportunities to make money, and brands with more engaged people (trading uniques for visit engagement) will be better customers, always.

On the web, we have opportunities in the form of visits. Visits are what matter. Each visit is a chance to move someone through a stage (or all stages) of their purchase funnel or other experience (like content consumption for media companies). Each visit is your chance to address their specific need (intent) at that moment. If you get a visit, you have no guarantee of that person coming back, or any reason to believe their next visit will have the same intent. Their previous click may have been a research intent, and this one may be a purchase intent, for example, and while the product may be identical over these two visits, the experience the consumer wishes to have is radically different. It doesn’t matter if they are a new person or someone you’ve seen before. That is your one chance to do what they need, or you can be damn well sure you’ll never see them again.

And we also need to look at it this way: is a single customer with two purchases worth more or less than two customers with one purchase each? Depending on the nature of your business, your answer may buck the trend, but we wrongly assume that most of the time, two customers are better than one. Not true. The customers who repeat business with you are usually far more valuable and far more profitable, and any businessperson with at least one lobe working will tell you that. The incremental cost of acquiring a new customer normally far outweighs the margin in their first purchase. You need loyal customers.

But wait, those loyal customers are uniques, right? WRONG. Loyalty is developed and maintained at the visit level. Each visit is, again, your one chance to do what that consumer needs. The web has zero barriers to your customer bringing their needs to your competition on a per-visit basis. I’ve bought stuff from Best Buy for years. Where is my next TV coming from? I honestly have no clue, and would be surprised if it was Best Buy. It’ll depend on what happens that visit (and I already happen to know they don’t have nearly the best price or return policy out of my consideration set).

Lastly, we know that every single system that we have in place today fails miserably in measuring uniques. Web analytics? No. Competitive intelligence tools? No. Toolbars? No. Logs? No. Login-based systems? No. Every single system that humankind has come up with to measure uniques online is utterly and completely wrong.

So, in summary, here’s what we know about uniques:

  1. Unique customers don’t care about uniques
  2. Advertisers don’t care about uniques
  3. The vast majority of businesses derive less value (cash flow) from reach than they do from depth
  4. The numbers are completely wrong. Always. With zero exceptions.
  5. We can’t understand intent at the visitor-level (what is the intent of an Amazon.com unique?)

And here’s what we know about visits:

  1. They are accurately (much more accurately, I should say) counted
  2. A large percentage of the time, we understand the visit intent
  3. We can find gaps in our web/business offering against the array of intents that make up traffic on our site
  4. They teach us about our customers’ needs, anticipated and unanticipated
  5. They reveal operational problems in our business process or business model (see 2 and 3)
  6. We care deeply about visits, because they are actionable (they either cause cash flow or [should] cause change)

Yes, there is business value in segmenting your visits between new and repeat in terms of understanding how these segments can be subdivided by intent, allowing you to hopefully make a good first impression. This reporting relies on the same methods as unique reporting, but actually has value. But that’s a whole other post.

No Trackbacks


  1. In the past, I’ve fought this battle many times. The reason I choose visits over visitors is that I am concerned about “an opportunity to convert.” The visitor is not an opportunity, the visit is. Each time someone visits a site, a store, a whatever, that is a unique opportunity to convert that person.

    If you focus on unique visitors, you instantly lose visibility into those unique opportunities to convert.

    Posted December 7, 2010 at 7:28 pm | Permalink
  2. The dilemma is that there is value in understanding what an individual is doing – what advertising they engage with, what they do on your site, what do they buy and how long does it take them – but the unfortunate reality is that a “visitor” is not a person, yet it is treated as such. My preference has also been to evaluate visits over time – more stable, understandable, comparable over time, and accurate. Where I use visitors, it is in context with other metrics, and with an understanding (and disclaimers) that it’s a poor substitute for the measurement of a person.

    Posted December 8, 2010 at 12:03 am | Permalink
  3. Good point, although I find that most visitor reporting exists at the extremes. On the one hand, you have an incredibly valuable tool like tealeaf that literally gives you each visitor’s experience. On the other end, you have Analytics data on “average” visitor behavior. But for most sites, there is no such thing as an average visitor. Someone who looks at 4.85 pages across 3.81 site sections and converts on visit 2.48. You can break this into a histogram to see relative frequency, and that is helpful, but more from an educational standpoint than an actionable one (which is valuable). But beyond this, insight is difficult. Visitors are difficult to segment into valuable groups. It’s a lot like pathing, where there are so many forks in the road and such a complex layering of use cases, that insight is rare. That said, the education is necessary and smart, but to improve sites, I’m a believer in visits, where I can compare experience to intent and act. Sites are opportunity engines, and rarely are good enough or appropriate for developing deep relationships with individuals, so I try to make it a better engine.

    Can you walk through some of the things you have looked at and gleaned insight from?

    Posted December 8, 2010 at 9:53 am | Permalink
  4. Hi Evan,

    Great stuff as always. I had a meeting earlier this week with several executives who kept throwing around acronyms like ‘UV’, ‘TV’ and ‘RV’ to talk about the types of visitors they get to their site – it wasn’t until 20 minutes in I was able to ask them “So what are you trying to accomplish online exactly?”

    The only question I would ask you is about your thoughts on the new breed of pan session analysis tools coming into the marketplace. Unique Visitors is a useless metrics in a traditional clickstream tool like Google Analytics, Omniture, etc, but when you start modeling visitor metrics in a tool like KissMetrics/Convertro/ClearSaleing you can get some serious (and awesome) visitor level analysis. (Please note that web analysts don’t use creepy analytics tools that take personal information or stalk all your web use – but I digress). If proper pan session tools get more adoption, doesn’t unique visits become a legitimate metric?



    Posted December 8, 2010 at 10:47 am | Permalink
  5. raf

    interesting post, and I will agree to an extent. And definatly agree that intent and conversion are probabay the most important thing to try to do.

    I would also argue that vitually all visit/visitor metrics need to be put into context and not viewed in a stand alone situation, however, I have done this in an analytical methodology which can contradict what you are saying

    I have done econometric time series models and regressions which would prove you right and wrong depending on the situation. (to be fair when I typically have measured unique I use registered users)

    I do agree that you have to look at the situation and that every visit is an oppurtunity, but the problem is you are just confirming a general marketers comment I have heard many times. “Any A/B, multivariate testing is a waste because any hold out is a wasted oppurtunity.”

    In some respectss you are not respecting the analytics, which I don’t think is the intent.

    It is important to note the abuse of blind metrics as you imply and ignoring important metric for no reason is dangerous.

    full disclosure I do work for SAS

    richard foley
    Product Manger Text Ansalytics
    Director Emeritous WAA

    Posted December 8, 2010 at 10:58 am | Permalink
  6. Absolutely agree. As you say, essentially it comes down to the numbers – the measurement of “visitors” is not as accurate, and even if they were, they’re still affected by the well-known issues: they’re browsers, not individuals and individuals use more than one browser, with browsers often used by more than one person. Couple this with the fact that individuals have multiple needs from a site, and the whole thing descends into a farce.

    Posted December 8, 2010 at 11:08 am | Permalink
  7. Hi Evan,

    Interesting post. Your metrics are only ever as good as what you are intending to do with them. Mainly uniques are usually given as ways of working out penetration figures rather than working on solutions to the marketing/website. In this case they are useless because uniques =/= people. In sales processes you are usually more likely to want to think about visits because they are a better metric for improving your marketing/website.

    But if you can find a valid use for uniques and will be able to do something about it, then they can be a useful metric. In all the examples you give above they aren’t. But as Richard says above, in some situations you do want to track your users over many visits to work out your sales process and improve it. This CRM is probably well above most companies though.

    These situations are rare though – far too frequently it is because someone wants to know how many people have visited the website.


    Posted December 8, 2010 at 11:44 am | Permalink
  8. YouKnowWho

    Oops; when I said “Uniques” on the golf course, I meant “Visits”, so in other words….



    Posted December 8, 2010 at 2:54 pm | Permalink
  9. Thanks everyone for the great comments. Let me try to address everyone:

    Great points, and I love tools like these. My point is that cleaning up the multi-touch or pan-session “funnel” involves really just one thing: making each of those touches or sessions more successful. Each of those sessions came with their own intent, some of them identical across two or more sessions. As each session is more successful, the likelihood of the next session happening increases, and if the next session is serendipitous (another media interaction, for example, without the user remembering your site from last time), it builds a platform upon which greater success in that session can be realized.

    But, unfortunately, I do not believe that many sites are dynamic or interactive enough to actually establish an intent-agnostic relationship with the user (they won’t just go there for the hell of it), thusly their pan-session experience is only as good as the sessions that comprise it. Even among news, media, sports, etc. sites with fantastic loyalty and repeat traffic, the switching costs to another provider are low and a sparkly new thing can gain new loyalty seemingly overnight. It is a legit metric, for sure. But from a “what can we do about it” standpoint, it feels to me it still boils down to the micro use cases of each session…

    I really appreciate the points you have here, and I can’t refute that econometric modeling does reveal a lot (clearly, since I am talking about lifetime value). But what we have happening here is people training for Delta Force combat when they don’t even know how to turn the safety off of their gun. This type of sophistication is not where the next win is for the average site, and it’s probably not where the next 50 wins are.

    I would love to see some of these models built into analytics tools in the near future, and have little doubt that pan-session and/or visitor-level expected future value (or dynamic discounted cash flow) modeling will be available soon. I’m sure you’re doing something similar, and the payout of adopting these models is enormous for large sites like the kind you’re used to dealing with. But going from where we are to this degree of analysis is not the right trajectory.

    Lastly, I absolutely and completely disagree that this implies anything about A/B or MV testing, both of which are tied far more closely to sessions and use cases than visitors or people. I am a rabid testing evangelist. I’d like to better understand how you arrived at this conclusion, because I couldn’t possibly disagree more with that comment, and I think any marketer or other person that would recite what you quoted marketers as saying is a complete idiot (and I would probably make them a shirt that said, “I am a complete idiot” that they would be required to wear at all times) πŸ™‚ Please let me know who said this nonsense about testing being a waste so I can be sure to never become friends with them πŸ™‚

    Very true stuff. But I would argue that penetration (in terms of reach) is increasingly useless, and there aren’t any modern marketing efficacy studies that refute this: past a certain mass (obviously marketing to 10,000 people is better than marketing to 50), there is an outrageous diminishing marginal return / opportunity cost to reaching incremental uniques vs remarketing to your existing audience (not existing customers, existing reach).

    As far as sales process, I still think that there are too many gaps in the funnel of a sales cycle that the site plays a role in, from a measurement standpoint. We can’t assume that the measurable phases don’t have offline or alternate site intermediate phases, so long-cycle, multi-session funnels really aren’t all that actionable for the majority of businesses; not nearly as actionable as use case based analyses, although people should have some operating knowledge about cross-session intent refinement or paths.

    We see the same issue, even in things we can measure today: what is the conversion rate of a keyword, for example? This metric is readily available in your adwords reporting, but the truth is, a keyword doesn’t have a conversion rate. A keyword generates traffic to a landing page. That landing page generates various paths through a site. Those paths will result in product detail views, which send items to carts, which send visitors into the checkout process, which then finally results in a conversion. By changing the site, we can radically alter conversion rates across any and all keywords. So what does that say about the intrinsic value of a keyword? There is probably an upper-limit conversion rate for a given keyword, but no: in no way does that keyword HAVE a conversion rate in its current state. There are simply too many phases in between.

    Posted December 8, 2010 at 3:13 pm | Permalink