Excellent Analytics Tip #26: Every Critical Metric Should Have A BFF!

YellowThere is unlimited amount of data thrown off our digital existences. (Or to use sexy term du jour , we have big data!)

Our leaders (companies, agencies, teams) have to deal with an incredibly complex landscape, and they don't have enough time.

The very natural outcomes is this ask of us: "Can you make it simple? What's the one thing I should care about?"

And we oblige: "Conversion Rate, that's it." Or "don't worry about anything except Facebook Likes." Or, "I read this blog, Bounce Rate is the only one!" Or, "Profitability, it is so sexy, just focus on Customer Lifetime Value, no, sorry, I mean Profitability."

To be fair, it is not just our leaders. Because the combination of complexity, limited time and available time, everywhere in the organization people want the one thing to watch for.

Honestly, who can blame them.

But the problem is that single golden metrics hide valuable insights and, more often than not, drive bad behavior. Especially in medium and large size organization because responsibility gets fragmented pretty quickly. (In small organizations there is a lot more end-to-end ownership amongst the few employees, and if something is going awry things hit the fan pretty fast. This is a great behavior correcting mechanism.)

So, how do we fix this problem in a responsible manner?

Here's my proposal: If you are pushed to have a single golden metric, give it a partner. For each metric deemed to be critically important, identify an immediately adjacent contextual / OMG we are on the right track metric that will give more context while incentivizing the right behavior.

The key is the immediately adjacent part. The BFF metric you find should not be one that is very far away. It should be immediately adjacent.

The reason is that it is easy for every discussion to come down to: "Well, all we care about is Profit. Why not just measure Profit?" That is right, we will measure it. But as we strive to improve the many things that result in a massive digital success, we need to look at multiple facets of our existence and we need to look at a cluster of critical few metrics. For many of them, Profit is not the metric that will give valuable context.

You'll see this in action in this post.

Let's look through ten specific strategic and tactical examples that will help internalize the value of the approach I'm recommending. The examples cover elements we optimize for in our acquisition (what are we doing to attract traffic), behavior (what happens once they land on our website) and outcomes (did we end up making money, were the customers satisfied) strategies.

1. Click-through Rate <-> Bounce Rate.

There are many good acquisition metrics including impressions, clicks, delivery rate, share of voice, and on and on. One of my favorites is Click-through Rate (CTR).

I like it because CTR it immediately discourages spray and pray strategies so prevalent in our industry (particularly in display advertising). It says, you must also get clicks at a certain rate. That incentivizes a focus on the targeting strategy, the content in the ad, recency and frequency capping, and other such things. Better, more relevant ads will get more clicks.

So, great metric. Dare I say, a key performance indicator.

The problem is that it does not provide any incentive to the marketing team to ensure the rest of the experience for the user is great. How do we get them to care and not just dump people on your site (mobile or desktop)?

Simple. Give CTR a BFF. Find it the immediately adjacent contextual metric. I suggest Bounce Rate.

click thru rate bounce rate

So the marketer is now incentivized to get lots of the right people to the site (better more relevant ads!) and get them to the right landing page that deliver on the promise made in the ad.

If users are sent to the right page, and they don't bounce, the Marketer should get her bonus.

See the magic? The BFF fixes a gap in the incentive/org structure.

Let's try another one.

2. Visits <-> Visitors.

(Or Sessions – Users)

You definitely want a lot of Visits. It delivers happiness!

But an obsession purely with Visits drives very short term thinking. You can get lots of terrible Visits on your site. Get a bit number. Up and to the right. But is the business really doing well? Are we adding value with our efforts?

That is unclear with Visits.

So. What's the immediately adjacent contextual metric? Visitors.

How many people did we manage to get to our site? Now things get interesting.

Consider this scenario: 50,000 Visits, 50,000 Visitors. And 50,000 Visits, 10,000 Visitors.

The people who see the data will ask very different questions.

For the first set perhaps they will ask… How come each person only visited once? Is that what we are solving for? How will our business survive?

For the second set perhaps they will ask… Wow, that is cool, each person visited five times on average. I wonder what the distribution looks like? Are there some outliers? What did the people who come back most consume?

Think of this scenario: 50,000, 60,000, 70,000 Visits.

How intriguing would be with the immediately adjacent contextual metric… 50,000 Visits and 50,000 Visitors, 60,000 Visits and 50,000 Visitors, 70,000 Visits and 50,000 Visitors.

Cool right?

visits visitors

The numbers could go either way, but having them together allows the recipient to just see the right amount of complexity. That is great.

Let's consider a more controversial, and behavioral, metric.

3. Time on Site <-> Page Views per Visit.

Time on Site is not a great metric in almost all circumstances. Time on the last page of a visit is not recorded (that also means time for bounced visits is not recorded).

But I still see it used, so let's get off that topic. (But think carefully before you use Time on Page or Time on Site .)

Say you communicate that the Average Time on Site is 60 seconds, 150 seconds, 98 seconds in the last three weeks (/months/days/years).

How much context is there to be able to separate the good from the bad? Not a whole lot.

What's the immediate adjacent contextual metric we can use? Why not use Page Views per Visit? Another indicator of activity during the Visit!

Now you might see 60 secs and 5 PVV, 150 secs and 20 PVV, 98 secs and 5 PVV.

OMG! What happened?

Try different combinations above and you'll see how these two BFFs works very nicely together.

From an incentive perspective, this is also pretty cool. Is a lot of time spent on the site good? Is very little time good? It is a difficult question to answer, but having the number of pages seen during the visit gives us immediate enough context to understand what is going on and where your initial focus should be.

That's it. That's all we are solving for.

[Bonus: While these are not as apparently adjacent, if you use Time on Page as you metric, try Page Value as its BFF. They are particularly good together!]

Let's switch to some outcomes metrics.

4. Conversion Rate <-> Average Order Value.

Our returning champion, Conversion Rate! Everyone loves Conversion Rate!!!

And they should. Conversion Rate is money, sometimes directly as revenue and other times indirectly via Leads collected.

But a pure obsession with Conversion Rate can incentivize sub-optimal behavior (not on purpose, but people react to incentives).

For example, a Marketer can focus on getting lots of simple, lower value, conversions because that will boost the rate up. Or they might prioritize fixes in the site design or experience that get people to a particular cluster decisions that will make Conversion Rate looks better but not solve for the longer term for the business.

conversion rate average order value

One simple way to solve this is to use of my favorite immediately adjacent contextual metric, Average Order Value.

2% Conversion Rate, AOV = $26. 2.5% Conversion Rate, AOV = $14. Ouch.

2% Conversion Rate, AOV = $26. 2.5% Conversion Rate, AOV = $40. Goodness, bonuses all around!

Simple fix, right?

The purists amongst you might notice that I'm really using Revenue as the BFF. You are right.

5. Conversion Rate <-> Task Completion Rate.

Two lessons.

The immediately adjacent contextual metric you choose will really depend not just on the type of business you have, but also the people you have, the size of your company, the incentives currently in place, and a number of such factors. So this blog post should simply serve as inspiration. Take the spirit, apply it to your unique circumstances.

Your immediately adjacent contextual metric can be a qualitative metric. At least some of the times, it likely should be!

Where I've implemented a simple where you able to complete your task qualitative data collection mechanism, I always pair Conversion Rate with Task Completion Rate.

Two simple reasons.

Conversion Rate solves for the company and Task Completion Rate solves for the customer. Such a delightfully nice approach to take.

Conversion Rate only tells you how a very small fraction of your users, who came to buy, did. Task Completion Rate shows you how 100% of your audience did, were they all successful regardless of why they came to the site.

When you see 2% Conversion Rate and 14% Task Completion Rate you will cry. Everyone in the company will cry. And they they will ask how come only 14% of the users completed their task! That will lead to a broader obsession by the digital team, almost always leading to big wins.

6. Revenue <-> Profitability

I have to admit this is a hard one.

None of the digital analytics tools make it easy to measure true profitability. And if you can pass that barrier (with, say, dimension widening using universal analytics), it is very hard to find this data inside the company (Finance department?), at a level of aggregation or granularity you need, and send it into your digital analytics tools.

The whole thing is so painful. But it is incredibly rewarding and if you want your digital analytics practice to reach the analytics.ninja state you need to do it. (Ask an authorized consultant to help you, you will get there faster: ).

Revenue is the ultimate goal. Lots and lots of Revenue!

But of course it is entirely possible for you to make lots of Revenue and go bankrupt. Simply sell products that are loss leaders or don't cross high enough above the hurdle of the Cost of Goods Sold.

The immediately adjacent contextual metric for us is Profitability.

Now when you report at a business/site level you can show that $54 mil in Revenue resulted in $40k in Profit. Or, at a campaign level you can show that while Twitter brings $5 mil in Revenue, that only results in $5k in Profit and while the Email Revenue is $1 mil the Profit from those campaigns is $700k because how how remarkably your campaigns are targeted.

As long as there is even $1 in Profit you should spend all the money on Twitter, but when it comes to making strategic decisions for the company, you might make different ones now that you know the profitability of email, and other acquisition efforts.

Think of how much fun it will be have this pair for the products you sell, the geographic locations of your customers, and so much more.

Revenue, meet your new BFF Profit!

[Bonus: Another sign you are an analysis.ninja is that you not only measure Profitability – session-level short-term metric – but pair it up with the immediately adjacent contextual metric of Customer Lifetime Value – person-level long-term metric. Shoot for the above first, then, if you are successful, get to this one because it is, obviously, much harder even if it is disproportionately more impactful.]

Let's step outside our owned platforms and on to rented platforms .

7. Video Views <-> Subscribers

Raise your hand if you've heard this: "How can we make our video viral?" Too often, right? And you know the moment those words are uttered you are dealing with a video/effort that is most definitely never going to become "viral."

Sad, but true.

And yet, the number of Video Views is a metric often elevated as the thing to solve for, the single golden metric for video powered digital initiatives. You won't be able to get away from reporting views, so why not find a partner for it?

The immediately adjacent contextual metric that works marvelously well is the number of new Subscribers.

youtube rsa subscribers

You can solve for the short-term with Views, but if you are not converting them into Subscribers you are not really building an owned audience that you can engage with over time. If your last one million Views video got you only 25 new Subscribers, was it really a success? Even if you got a temporary bump in publicity?

[Bonus: While Subscribers is my ultimate success metric for YouTube – I crave large owned audiences so that I can stop renting them from TV and/or Radio and/or Google and/or AOL – the other two that can also share key context are % Completes and Amplification Rate. Depending on your local circumstances, you could possibly consider those as well. Though if you want to make me happy, you'll choose Subscribers!]


People who don't know anything about Social Media use Facebook Likes to measure success.

There are so many of them, including your boss! Let me give you a virtual hug. There, there, it gets better.

First, be sure to mention that Likes simply represent people walking by us on the street who smiled at us. They meant nothing more. We need to make sure that we are creating content that is incredible and of value. That is what causes people who gave us a passing Like to come back again, engage with us, give us their precious attention.

Then we have to think about how do we give our dear boss, still obsessed with Likes, the immediately adjacent contextual that will help her/him make smarter decisions.

I recommend two different ones.

Likes are most commonly used at a page-level. For example, my Facebook brand page where I post daily analysis on an interesting topic has 19,789 Likes. The best immediately adjacent contextual metric for the page-level Likes is Talking About This. At the moment that number is 1,203. It is a decent tentative way to understand the engagement on Facebook.

The 1,203 is great context to have for the 19,789 for your boss.

[Bonus: For later reading, when you are attempting to be an analysis.ninja, see this post: Excellent Analytics Tip #25: Decrapify Search, Social Compound Metrics]

Likes are also present at a post-level. For example, my Facebook post on how to stand out from the crowd during an interview has 62 Likes. That is insufficient to indicate success because I not only want you to Like it, I also want you to amplify it to others so that I wonderful content (!) can reach others I can't reach myself. The best metric for that is Amplification Rate.

The above post only has one Share (the key ingredient in Amplification Rate). That is super-lame (and for such a good post!).

While another post, Global Views on Morality: Homosexuality, has 72 Likes and 40 Shares. Much, much stronger Amplification.

When your boss looks at the two posts he/she will now be able to recognize that one was more successful in terms of what's important (reaching new audiences) than the other. That's exactly why you want your metrics to have BFFs!

9. Mobile: Installs – 30-day Active.

We looked at YouTube, we looked at social, and so mobile can't be too far behind! Let's look at a quick one for mobile apps.

The most important metric our leadership cares about when it comes to apps? Number of Installs.

And it is important.

But if 80 to 90 percent of all downloaded apps are used only once, should we have an immediate adjacent contextual metric that will be more insightful?

I recommend also reporting 30-day active, the number of unique users who have been active during a 30-day period. You have some flexibility in how exactly you define it, but as long as you stay consistent it does not matter.

Now your boss is focused both on getting more new customers, and on keeping the ones you already have. Balance. It is what makes the world go round!

10. [By channel] Conversions– Assisted Conversions.

Let's close with a pairing for all of you analysis.ninjas.

It is common to segment Conversions, our beloved key metric, by the source of traffic. Earned, owned, or paid. Or, Google, Email, AOL etc. It helps your boss understand how best to optimize your acquisition strategy.

The challenge is every single analytics tool reports single-session conversions only (also known as last-click attribution). This is absolutely silly and leads to awful decisions.

The immediate adjacent contextual metric you need is Assisted Conversions – the number of times that same acquisition channel (earned, owned, or paid) was present in the customer journey that lead to a conversion but that channel was not the last-click.

Essentially, how often did that channel help with a future conversion?

assisted conversions

Now you have excellent context for making smarter decisions about the full-value of each acquisition channel in your portfolio.

For example, Organic Search delivered 119k last-click Conversions and also assisted with another 73k Conversions that were delivered via other channels when looking at a last-click view.

[Bonus: For more awesome goodness on this yummy topic check out this post: Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models.]

Ten short stories to help you internalize the incredible value of having an immediately adjacent contextual BFF for every critical metric you report to your management team.

Oh, and you can easily put all this together in a very simple dashboard…

digital analytics dashboard

Throw in your pretty pie charts (no!) and your stacked bar graphs and some lovely trend lines and you have yourself all the ingredients for creating an organization where data delivers the kind of insights that deliver big action!

Please consider the examples in this post, and the dashboard above, as a way of thinking I would love for you to embrace. The specific metrics you end up choosing will depend on many important factors. If you create your Digital Marketing and Measurement Model, you will know exactly what will go in the above dashboard for you and then all you have to do for each metric is find the BFF metric.

Friends don't let their KPIs not have BFFs. : )

As always, it's your turn now.

Do your current critical few metrics have an immediate adjacent contextual metric? Do you agree with the metric recommend in this post as the BFF metric? Would you have chosen something different for Time on Site or Visits or Likes? Are there metrics you are struggling with when it comes to identifying the BFF metric? Does your company dashboard provide all the necessary context to aid smart decision making?

Please share your insightful feedback, tips, omg don't do thats, and stories.

Thank you.


  1. 1
    Sridhar Kesaraju says

    Hi Avinash,

    I really suck at analyzing things. I always wonder where to start the analysis for my website, but now I know where to start exactly to find useful insights and drive action.

    Thanks again for a great post, as always.

    Sridhar K

  2. 2

    For social media, shares are the most important metric, since people are willing to state that they are endorsing your post.

    Comments are also a great sign since it shows that people had something to share, although comments are easier to manipulate.

  3. 3
    Ben Armstrong says

    Hey Avinash,

    Even though I'm not knee-deep in Analytics in my current role, having read a lot of your articles – I feel as if I have a very well rounded knowledge of best practices should I need to pose as a Ninja.

    The thing I value most highly about this post is the ease of implementing it (at least in some form!). It's also impossible to argue against the improved context of a BFF approach and it drives the company forward with an analytical/inquisitive approach rather than a black/white one.

    As you state, looking at a single metric promotes not understanding the data and thus judging it against some arbitrary measure of success ("21% Bounce Rate, sweet! Let's go and party" etc). With the dual metric, the Hippo's will be saying 'If this means that then why is this so low/high/bad/good/befuzzling" which is obviously a much more important discussion to have.

    I see your BFF approach as essentially forcing a discussion with data rather than simply presenting it and crying/jumping for joy.

    It's a more interactive way of getting everyone thinking on the same wavelength and promoting company-wide interest in the importance of data. For this I salute you!

    Kind regards, and keep up the great blogging.

    Ben (Potential lover, not non-flirt)

    • 4

      Ben: I could not love your comment more!

      You are absolutely right, it is all about forcing that discussion. And, I'm sure you got this as well, forcing more accountability on to the Analysts for not just, as you wonderfully put it, puking data out with some arbitrary measure of success.

      While this will be a new uncomfortable role for the Analyst, in the long run it will make her/him a much stronger part of business success.



  4. 5

    Hi Avinash,

    First of all thank you for slowing down this OMTM (One Metric that Matters) viral thing that has come after the Lean Analytics book. I do think we have to focus when analyzing but "one metric only"? Seems too superficial for me. OMTM + BFF metrics is a waaaaay better tactic. Anyway…

    I have a question about your line in "Each person visited five times on average. I wonder what the distribution looks like? Are there some outliers? What did the people who come back most consume?".

    When studying heavy and soft users I did a histogram to distribute the clicks frequency on a certain homepage. The results were as aspected: almost 70% of the clicks were done by a mere 15% of my users. But then I tried to do that with my visits. That's when I got frozen by our data team! "You can not distribute visits frequency, Google Analytics only works in a cookie 2 year time frame. You'll never get to set the time frame you want and know how many times your users visited your site."

    And that made me really sad. Knowing my clicking frequency by user on a page is just as important as knowing my users visiting frequency. Now the question I leave to you is: when viewing the report Frequency in Google Analytics, does the time frame chosen really apply? Or am I seeing a 2 year history data? Because one thing is certain: the number of visits in the Frequency report never matches the number of visits in the Audience report. =/

    What should I do?

    Thank you a lot and great post.

    • 6

      Rebecca: Ignore Google Analytics, just think of all web analytics tools and how cookies work (first-party and not the sub-optimal third-party), and how we all collect data.

      Two years is a very long time for a web analytics tools to track any person using a individual cookie. You'll always have some people in that bucket, usually a minority, but for the most part as you head beyond a few months, say beyond a year, your ability to have unique cookie based tracking to be totally accurate is on very thin ice.

      I'm not worried about privacy or other things, just the raw technical ability for that tracking mechanism to work with a very high degree of confidence is low.

      So, if you got two years count yourself lucky, assume the data is a little south of being perfect, use that as context for decisions you make.

      Ok, back to GA. Some variation in long-term numbers between Behavior and Audience reports is to be expected. See above. If you see massive differences it is prudent to hire a consultant to help you figure out what is up. I just looked a a bunch of accounts and the numbers look reasonable.

      Now back to all web analytics tools… the type of analysis you really want to do is Person-centric that does not simply rely on cookies. This can be done now with tools like GA allowing you to segment and report on Users. It can be made stronger if your company is willing to use things like setting a custom variable with the scope set at User that is tied to an login ID (using an anonymous hash key). That is real strong tracking, and you could likely rely on it for years of data and across multiple devices.

      The question is how much your company actually wants to do this. If they really want it, it is possible and just needs some technical investment and encouragement to your Users.


  5. 7

    Hi Avinash,

    Great post and great context around each metric, makes perfect sense.

    You didn't mention the off-line metrics to be included to make wiser decisions and its impact on your online activity & vice versa.

    This will off course be a manual process since there is still no primarily key between the two world.

    Thanks for sharing.


    • 8

      Waiss: An example did not come to mind when I was writing the post. But of course offline impact is critical to measure.

      One example I can think of from recent past was tying an increase in tech support visits on the site to a reduction of phone calls to the company support line. Nice combo.

      Another one was in the automotive category pairing up leads received on the site to visits to the car dealership (amazingly the former was less than the latter!).

      The primary key is indeed a problem, but solvable with some effort. (Multichannel Analytics: Tracking Offline Conversions. 7 Best Practices)


  6. 9

    Hi Avinash!

    This post made me smile! The BFF tactic is so smart!

    Thank you for this amazing post!

  7. 10

    Hi Avinash,

    I just wanted to say a really big "THANK YOU" for helping me to get better and smarted in this endless world of web-analytics :)

    You`re really doing great!

  8. 11

    Awesome insight as usual Avinash.

    My pairing of preference pertains to email campaigns, but the same approach could be used in any situation where attrition is possible.

    I love seeing visits (from email) unsubscribes. This very quickly shows that hitting a list over and over might drive traffic (short term dopamine fix), but also keeps the future risk in mind.

    There is a tipping point where the damage to a list outweighs an instant boost in traffic; and this pairing encourages segmentation, discretion and consideration of the value of content beyond "buy now!"

  9. 12

    Avinash, excellent excellent excellent advice. I've always thought of the ride-along BFF metric as context and that works really well for determining what's the best BFF metric to show.

    Here are some other metrics I like to couple when analyzing or reporting KPI's to a client:

    Email subscribers & opened in the last 6 months – shows the size of the email file, but also the active size of the file. Similar comparison to likes & talking about for social.

    Conversion rate & number of orders (or conversions) – conversion rate is missing something that shows the volume of what's happening, I like number of orders or visitors as a reference for that.

    Orders & new customers – not really a web analytics metric comparison you can make, but it's crucial to showing what portion of your volume is new business.

    CTR & Avg ad position – This is an PPC specific coupling. The CTR of an ad is so rarely viewed in it's right context. A new competitor may have come in to the game, you dropped half an average position and your CTR fell, it's not the ad text's fault that your CTR is down. Another metric that can also be used here to show context is top of page rate.

    Acquired customers & churned customers – Totally an offline thing, but insanely important. If I had to choose one pairing of data to represent the overall health of your company's marketing, it'd probably be this one.

    My recommended pairings certainly have a little eCommerce bias to them, since that's what I do, but hopefully they're still useful for the non-eCommerce company.

    • 13

      Dave: These are great examples, and will help everyone (ecommerce or not) to think about how to step up their game. :)

      I don't use Open Rates in email computations simply because the fact that an email is opened is a very fragile bit of data to collect. It massively under-represents what the reality might be. Perhaps, in the spirit of your recommendation, use something like Clicks-to-Delivery-Rate. (Email Marketing: Campaign Analysis, Metrics, Best Practices)

      Thank you again, great list!


  10. 14
    Ryan Finley says

    Another excellent post! Thanks for the always interesting insights into our craft.

    I'm about 7 months into a new position, where I am able to utilize my analytics know-how not just to drive quality Onsite Search, but also to manage a recommendations program. One issue that I've come up against is the "One Big Metric" related to recs that my new company has always relied on: Percentage of Total Site Revenue. (They also tend to fixate on the high-end of the typical average, but that's a separate issue.)

    Does % of total site Revenue still couple with profitability? Or is there a different companion metric that may be beneficial? I fear love for this one metric is too ingrained for me to change, but your BFF idea may work perfectly for me if I can find the right buddy.

  11. 15

    You say, "As long as there is even $1 in Profit you should spend all the money on Twitter,…"

    I think you mean to say that you should continue to spend money in Twitter so long as the marginal dollar spent yields positive profit, or to put it another way, spend until marginal revenue equals marginal cost. You would not want to spend $5M in Twitter to yield $1 in profit if only spending $4M in Twitter yields > $1 profit. It is possible that we spend beyond the point of diminishing marginal returns and profit begins to fall as we spend more and receive less return on the margin. I'm quite sure this is what you meant but I don't think it was quite clear.

    • 16

      Ryan: If you measure % of Total Site Revenue by various sources (Search, Direct, Keywords, Email, Social etc.) then it is a pretty good use of that one big metric.

      The BFF you need could be picked from #4, #5, or #6. That is to say, Conversion Rate is a good BFF because it will help you identify where the rate is very high but perhaps you are not getting enough visits or Profitability is a good BFF because you could get lots of Revenue but you might not be making equal profit from everything or, finally, I touch on this, the Micro-Conversion Rate is a good BFF because it will help you look beyond the few people that will buy to everyone that will come to your site and deliver some value.

      So three excellent choices for you, pick the one that is most relevant immediately to your company. Over time, you can even use all three!


      • 17

        I believe you meant this reply to Ryan Finley instead of my post which immediately followed his, though we are different Ryans. I like the post in general but the specific advice regarding Twitter struck me as odd. I'd love to see a follow up post addressing the ability to find the right level of investment by analyzing marginal returns at the publisher level instead of average returns.

  12. 18

    Amazing. Truly amazing.

    I really liked this post because it gives practical and immediate insights.

    Second, I would add two dimensions to the 1st strategy – Landing Page and Device.

    As you know bounce is very much depends on the landing page you sent the visitor, and CTR (and of course bounce) is super duper depending on the device the visitor came from (for example – in AdWords if you're not in the 1-2 place in mobile – you will be in the bottom, and then your CTR will be VERY low. From the bounce side – if the keyword was "roses" but the landing page is about anemones – the BR will be high, and if your site isn't mobile friendly – that will affect the bounce also).

  13. 19

    Nice post Avinash!

    I am constantly promoting "buddying up" metrics due to the need to put the data into context and I believe you have discussed this previously too :)

    • 20

      Bradley: It is nice of you to remember!

      I don't think I've talked about this on the blog, but it is a section is my very first book, Web Analytics: An Hour A Day all those years ago. :) Though it was not as well thought out as this post (all these new tools and marketing options!).


  14. 21

    Hi Aavinash,

    Truly an amazing post. A very interesting read.

    But, suppose if I have a single page website, then CTR and Bounce rate would not be BFFs.

    If I have only one page on my site, Google Analytics would not register multiple pageviews unless users reload that page. As a result, single-page sites will have high bounce rates.

    So, I could have a high CTR with a very high bounce rate as well.

    What is your view on this?

    • 22

      Syed: You are right, each person will look at their business and adapt the recommendations in this post appropriately.

      Before I can answer your question, we would together have to answer this question: Why does your one page website exist? If someone comes to your site and only sees one page and does nothing else (because if they do anything else that is not a bounced session) then how do you make money?

      You could make money online or offline. If the answer exists to that, we can find an appropriate BFF metric for advertising CTR in your case.

      PS: For more on bounce rate and single page websites, please see my post on bounce rate. The comments will especially be interesting.

      • 23

        My own website is not of one page. However, I have seen a lot of one page website. So I was just curious who would be the CTR's BFF for such websites. :)
        Btw your bounce rate article is also worth reading. Thanks Avinash :)

  15. 24

    Hello Avinash,

    I wish to add a riff of a product/service with an big online presence for leads capture but offers an off-line experience/purchase.

    I see a scenario of Conversion Conversion. Where Conversion (Leads through website), the BFF would again be Conversion (Offline-paying customers through follow-ups by Sales Reps).

    We are indeed inspired. :)

  16. 25


    Thanks for this very useful and interesting post.

    Businesses need to find the right people and tools to handle important tasks. You shared some relevant information and it will guide me in putting my website on the right track.

  17. 26

    As much as I hate analysis and statistics, I love the way you presented it here in this article. Clear, laid down in a logical order, and explained well.

    I now know the importance of each element and why we should measure it and how.

    Thank you very much.

  18. 27

    Hi Avinash,

    Thanks for the BFF approach. Sometimes data (or metrics) in isolation can mislead, but having another metrics along side would definitely help in making sense of those numbers.

    Thanks again for this great post.


  19. 28

    Great and complex information.

    Google Analytics is very powerful tool. I use it from almost 5 years and maybe know 30% of its possibilities.

    Thanks for your tips.

  20. 29

    Hi Avinash,

    Nice post, was a pleasure to read this one. I have some questions on using this strategy in GA:

    1. Visits and Visitors as BFF is a nice idea. It also depends on which dimensions I want to use this metrics. As a content/news site I would focus on unique pageviews when it comes to single articles. However there are no unique pageviews for content groups, or the automatic category reports (e.g. /sports/, /news/, etc.). Would you recommend to use visits/visitors in GA when it comes to content?

    2. In mobile apps I'd love to have a retention report: 30-day-active metric. However in analytics the only way seems to be to set a custom dimension on user scope and set it to installation date. Or is GA having a new feature I haven't come to know yet?

  21. 31


    I think I know why that article has only 1 share. It's an article that teaches you how to stand out in a crowd. Why would I want to share that with the crowd? I want to keep that a secret so only I stand out! :D

    Thx for the post!

  22. 32

    Another solid resource, Avinash!

    My favorite combination is blog sessions plus assisted/direct conversions for the blog.

    We create a special channel grouping in MCFs based on landing page and count anything where the landing page contains "blog." We can see if the new traffic we're getting for each article ultimately sits on the conversion path to account for both the popularity and impact of what we're publishing.

  23. 33

    Interesting idea for attaching each metric with another metric partner.

    By using the magnificent business dashboard you recommend, this really help me to identify what to improve on my website, thanks!

  24. 34
    Veepal Panchal says

    Hi Avinash,

    Excellent post, your post really help to users, easy and useful content to understand


  25. 35

    Great analysis and perfect for starting business as mine, when I don't really have so much time to spend in really deep analysis.

    Thanks as always

  26. 36

    Does google analytics calculate unique visitors by acquisition channel?

    Recently I have taken over data and reporting in my company and found a 3rd party tool that is pulling unique visitors by channel from google analytics. When looking on google analytics however, this option does not seem to be available.

    Is this being pulled by through the API? If so, how are is it being attributed? How would the unique visitor be counted if it was directed to a site multiple times by different channels.

    • 37

      Wil: Yes, it does. And it does many more things too. Please see the Multi-channel Funnels reports.

      The best way to get answers to your question specifically for your business is to hire a local GACP. You'll find a list here: http://www.bit.ly/gaac

      Good luck!


      • 38

        Thanks for the reply.

        The Multi-Channel Funnel reports seem to show the mix of different channels that a converted "user" (what GA defines as a Unique Visitor) comes from. As expected, this typically includes multiple channels for a single user over a period of time

        But in my case, the 3rd party tool is pulling the metric "unique visitors" directly from GA. The total number for the Unique Visitors match up w/ what GA reports as "users" for a defined time period.

        When applying a filter that's defined similarly to GA's Default MCF channel definitions, the tool is reporting a specific channel that the UV is coming from. The channel segments sum up exactly to the total UV's ("Users") GA provides. This would only make sense if each UV visits our site exactly one time, but this is not the case.

        When visiting GA's Acquistion/Channel report, you are not able to select User's as a metric.

        Does the GA API segment out UV's and associate it by a specific single acquisition channel? I'm not sure how this would be defined or how to access this through the GA interface.

  27. 39

    Hi Avinash,

    I totally agree about the part where you said Facebook likes are just like people who smile at us on the street. =)

    Thanks for pointing out many good ways to analyze metrics, this post has given me motivation to analyze more to improve my website!


  28. 40
    Joseph Boisseaux says

    Hello Avinash,

    It reminds me of an old lesson of geometry :

    "A line any straight one-dimensional geometrical element whose identity is determined by two points."

    No way of reaching a goal without a well-defined course.

  29. 41

    Exactly as you say, Revenue is the ultimate goal.

    It's useless to optimize your landing page till your eyes burn if the revenue stays the same or goes down.

    I have seen this happen multiple times & that's exactly why split tests should be done one by one.

  30. 42

    Great post Avinash, you've nailed the sweet-spot of metric analysis! You haven't undermined the importance of analysing our metrics; you haven't encouraged a blinkered-obsession with single metrics; and you haven't encouraged a vague, white-washed, gotta-catch-em-all approach to analysis.

    You've paired the metrics beautifully, and I'm looking forward to applying these principles to my own analysis. I already have the data – I just didn't have the context!

    Thanks again Avinash.

  31. 43

    Hey Avinash,

    Excellent post. Do you have any articles about choosing the right metric?

    For example I want to test a brand page. The control is the standard filter/navigation on the right hand side with products listing in the main content area. The test is a page that just show categories of products the brand sells to help people narrow down their selection.

    By nature the test has a lower bounce rate as people will click to a category and then bounce if they don't see the products they like on that page. As a result I'm using time on site as a kpi for the test, but this only works for people who land on the brand page.

  32. 45

    Great post Avinash. As usual things have been explained in a way that a lay man would understand.

    Will try to use the BFF metric methodology in my future reports.

    Very useful guide to explain to our clients the results of a campaign. I have faced these challenges personally, especially trying to explain attribution.


  33. 46

    Thanks for providing better understanding to the relationships of analytics.

    When you look at the terms visitors and ctr, we can easily see them as simple metrics, not what they imply about what is happenting.

    Your simple example, visits–>visitors shows that we should remember that every visit means we have a visitor. this highlights the personal aspect of metrics. We are measuring and understanding actual human behavior.

  34. 47

    Great post, Avinash. And great discussions around here. To add to it:

    – On the “one metric that matters” topic, I do not know a single person that could do her/his job with a single metric, but I strongly believe in setting our priorities straight in a merciless way, so…

    – As you say, these “partner” metrics come to the rescue whenever one of them alone cannot bring value by itself. How about aiming for a single formula that combines every couple into a single, time-saving KPI? Example: “Qualified CTR” (discarding pages that surpass a certain Bounce Rate threshold for CTR estimation). Or “Engagement” (sorry!) combining Time on Site and Pages per Visit? Another example is Cost per Lead, rather orphan without a total volume of Leads unless the metric is fine-tuned to discourage working with an unacceptably low amount of leads (eg. by inflating their cost if below a given number).

    (I can almost see you replying that you hate complexity and perhaps adding that the underlying data is far too fragile for us to mess with it – which are always good points too :)

    • 48

      Sergio: Please accept my apologies in being so late with my reply to your wonderful comment.

      Your example of pairing Cost Per Lead with Total Leads is a great combo. We can even consider Total Revenue, or Missed Opportunity (number of impressions/clicks left because you are only willing to pay x) or Profit Per Lead (how much flex is in model for you to get more clicks or people), are all great combos.

      The danger of aiming for a single formula is that we lose visibility to which element of the formula is working better or worse. I talk about that a bit more in covering Compound Metrics, I'm shy about using them because they muddy the waters rather than clarify.

      But, I do not believe, as you've defined, that Qualified CTR is a compound metric, it is simply a segment. Time on Site and Pages per Visit sort of get at the same thing, how long did people spend with us. Since Time on Site is not always perfectly measured (for bounced sessions and last page of the visit), I simply recommend keeping Pages per Visit. So that won't be compound as well.

      All great recommendations, thank you.


  35. 50

    Great and and well laid out article

    Google Analytics is actually a very handy tool that most people seem to use in the wrong way never benefiting from its power.



  36. 51

    Wonderful post, at first I used to get confused about this topic but Avinash has explained it not only in simple language but also with diagrams to make the whole thing easy to understand and grasp.

    Thank you Avinash


  37. 52
    Dien dan dung says

    Nice post Avinash!

    I am constantly promoting "buddying up" metrics due to the need to put the data into context and I believe you have discussed this previously too :)

  38. 53

    Avinash, what a wonderful post. I think you buried the best, though. #10 is the best on your list — should be 1 or 2.

    thank you again


  1. […]
    Once you know what to measure, you can turn to the newest article from Mr. Kaushik about how critical it is to give context to your simple reporting. If you have to have one “golden” metric, you cannot fully discover its valuable insight if you don’t pair it with an “adjacent” or “BFF” metric to identify if you are really on the right track.

  2. […]
    La lezione privata di web marketing di oggi è dedicata all’analisi dei dati. Questo post è liberamente tratto da un post del blog del mitico Avinash Kaushik. Anche quella di leggere, tradurre e riscrivere un post può essere una buona forma di studio, per lo meno per me. Questo post parla delle metriche più importanti di Google Analytics (e non solo) e di come cercare di leggerle nel modo più critico possibile.

  3. […]
    Excellent Analytics Tip #26: Every Critical Metric Should Have A BFF! – Occam’s Razor
    You can’t have a successful marketing campaign if you don’t have the right analytics in place to measure it. In this piece, Avinash Kaushik lists some great stats that pair nicely with one another to give you true insight into how your marketing is performing.

  4. […]
    Here is an amazing analogy from digital marketing guru Avinash Kaushik – “Likes simply represent people walking by us on the street who smiled at us. They meant nothing more. We need to make sure that we are creating content that is incredible and of value. That is what causes people who gave us a passing Like to come back again, engage with us, give us their precious attention.” [ Attribution: https://www.kaushik.net/avinash/excellent-analytics-tip-critical-digital-metrics/ ]

  5. […]
    Every critical metric needs a BFF
    More: read the blog post
    Avinash suggests matching your critical metrics with a BFF metric, to add depth to your insight. The brilliance of this approach is that it adds valuable context and meaning to each metric you use. Here’s the list of metrics & their BFFs from the post.

  6. […]
    There are a lot of debates going as to what to look at in your Google Analytics and the metrics to discard, but I think it’s increasingly imperative to look at the whole picture. I READ AN ARTICLE A COUPLE MONTHS BACK ABOUT HOW EVERY METRIC SHOULD HAVE A BFF MEANING, it’s difficult to look at just one measurement to determine success. The mentioned metrics above shouldn’t be measured alone, but they should be looked at in conjunction with each other. For example — your blog posts have a high average time spent on each one… but your readership consists of just your loved ones.

  7. […]
    Occam’s Razor is run by Author, Digital Marketing evangelist and CEO, Avinash Kaushik, and features a series of amazing tips and tricks for SEOs and marketeers. Avinash’s Analytics Tips are particularly brilliant. Featured Post: Excellent Analytics Tip #26: Every Critical Metric Should Have A BFF!

  8. […]
    Common examples of various KPIs might be, Leads, pipeline, qualified opportunities, closed revenue, website visits, friends/followers/likes, clicks, conversions, downloads, etc. For more examples of KPIs see this great post.

  9. […]
    [Further reading: Avinash Kaushik – Excellent Analytics Tip #26: Every Critical Metric Should Have A BFF!]

  10. […]
    More: read the blog post
    Avinash suggests matching your critical metrics with a BFF metric, to add depth to your insight. The brilliance of this approach is that it adds valuable context and meaning to each metric you use. Here’s the list of metrics & their BFFs from the post.

Add your Perspective