As someone who has worked for a performance agency for the last five years, I have learned that almost everything we do needs to be based on data, knowledge, and past experiences. But only almost because sometimes we get faced with things we’ve never seen before. It’s rare, but occasionally we have to act on impulse and instinct.
That is a scary place for marketers. What happens if we’re
wrong, if we make a bad decision, or we don’t see the results we expect? A lot
of us go into some kind of internal meltdown, blaming ourselves for not doing
what we should have done even though we didn’t know what it was that we should
have done in the first place.
As much as we hate to admit it, there are always going to be
situations where we don’t know the answers to everything, so the next best
thing we can do is try and minimise the time we spend in those scenarios and
get through them as quickly and effectively as possible. So how do you do that?
Work through your issues.
Say you need to order a pizza; before you do you need to
decide what kind of pizza you want, how you want to order it, whether you want
any sides, what size you want, do you want a drink, do you even want pizza?
Trick question, everyone wants pizza.
But what happens when it is something more serious than
pizza? Another trick question, nothing is more serious than pizza.
For argument’s sake, let’s say there’s an issue issue with a client’s website where performance just isn’t where you expected it to be. The client is questioning it and what you’re going to do about it, but you aren’t sure what it is or what to do to fix it, and you know you’ll probably need to do a bit more than usual because the client’s developers have a massive backlog to work through. What do you do?
You work backwards. You find all of the possible blockers to
getting this performance fixed, and then you figure out which aspects of those
blockers are in play and what you do next:
What potential blockers could you encounter in this
There are probably more, I just like round numbers. Now,
which of these can you influence?
Hardcore marketers might argue that you can influence
demand, but that is a whole other post.
Now you’ve got a list of things that a) might be causing
issues on your website, and b) you can do something about.
2. Identify aspects of blockers causing
The next step is understanding why these things might be performance blockers, which aspects specifically are at play? For this, you can follow a similar thought process to below:
Poor indexing setups can impact the crawlability of a site,
which in turn impacts how crawlers understand your site. If your client’s
developers block the entire website when they are doing maintenance, chances
are over time it is going to hugely impact the ability of crawlers to
understand your architecture, internal linking, and authority. If you’ve
accidentally blocked an entire category from being indexed, it isn’t going to
Although probably less of a concern than it was a few years ago, high levels of duplicate content will leave the website battling with itself to rank the correct pages. Although you won’t be “penalised” for it, instances of duplicate content will make it very difficult for either content set to rank effectively. Lower rankings = less traffic = less performance.
The go-to excuse for every marketer at some point in their career, “oh, my performance can’t be doing that, it must be the tracking”. Most of the time it isn’t, sorry. On the odd occasion, however, it is. Whether something has been pushed live in GTM that changes your configuration, or a site migration has resulted in duplicate tracking codes being in place, if you’re really struggling to figure out what has changed, speak to one of your data experts. But also check out this deck from Emma Barnes where she looks at how to report on data from Google Analytics (and how you can diagnose issues).
Ah, keywords. While Google is getting better at understanding the semantics around search, almost everyone you speak to will have a way to determine which keywords should be associated with which pages and where you’re missing out. That’s fine, we’re moving a bit more towards key phrases than individual words, but we’ll roll with it. Incorrect keyword targeting can result in cannibalisation as well as a poor experience. Why would a service page on penetration testing mention things like colocation? Oh, but it has more search volume, well of course include it then! No, you wouldn’t, because it isn’t a relevant. A quick ranking report overlaid with a data export from GA will help you identify where keyword targeting might not be the best on your site.
Almost impossible to look at objectively, UX is a much harder one to diagnose. You might have different opinions to other people, how you use the website might be different to how your friend uses it. To really get to the bottom of whether you have a UX problem, you’ll need things like heatmaps that you then analyse and turn into A/B tests, that you then analyse and turn into conclusions, and likely tweak and retest. UX improvements are never final, but give yourself a starting point.
A very broad issue, usually put down to demand/changes in user behaviour. Great, so nothing you can do then? Wrong, if it is a change in user behaviour you can check by looking at how users are accessing your site, what terms are they coming through on, has bounce rate increased, are you answering the queries properly? Or, is it simply your checkout doesn’t work properly on the top browsers being used to access your site?
The holy grail. Landing pages are your best friend and your
worst enemy. How many times have you had clients implement “PPC landing pages”
because your normal ones aren’t converting paid traffic? When it comes to
landing pages, you have got so much data available to you it can be difficult
to know where to start, but ultimately if they aren’t ranking you look at
keywords, and if they aren’t converting you look at relevancy. These should be
the same thing.
3. Determine impact scope of identified
Once you’ve drilled down the potential broad issues into
aspects, you can start weighting the aspects in terms of priority to fix (PTF).
For this, you need to look outside single channels – because fixing things from
a single channel perspective doesn’t fix anything – what overall impact is this
having? You can generally break this down into four areas:
For simplicity, here’s how that would look at a broad level:
And then within these 4 areas you can start looking at
overall performance – if it is page-specific, are there particular channels
that perform better/worse than others? If it is custom, are there specific
setups impacting how data is reported?
This is your impact scope. With all of this information, you can finally break it down into next steps.
4. Determine requirements for
Site-wide and custom issues will most likely require
client-side resource to assist with any fixes, e.g. development team, whereas
section and page-specific fixes can often be implemented agency-side.
Everything you do should be prioritised on an effort vs. impact system:
Low effort, low impact = medium priority
Low effort, high impact = high priority
High effort, low impact = low priority
High effort, high impact = medium priority
It will be very rare that you are able to fix everything
yourself, but if you know that the client’s developers are working through a
massive backlog, try and do everything you can your side while you wait for
them to get to a better place.
Resolving issues isn’t about highlighting all of the things
wrong with a website and then passing off the fix to developers. They have
other work to do, which is often more “business-critical” than your requests.
So, when you send them a list of 10 things to do that you could do 40% of in a
day or two, they are going to be in no rush to help you when you’re just adding
to their to-do list.
All of your recommendations get added to a backlog, and then
you blame them for not implementing things fast enough to turn performance
around while they are complaining about you for sending them work that isn’t
When you reach the end of this process, you should have a
clear understanding of:
All the different aspects that are causing
issues with performance
How much of an issue each aspect is
How much damage each is causing
How you need to fix it
When you’re going to fix it
Who is going to fix it
And what marketer doesn’t want that as the outcome of every
bit of analysis they do?
Dos and Don’ts
Finally, a quick wrap up of dos and don’ts to take with you
to your next daunting scenario:
Approach every issue with a structure, no matter how many times
you’ve seen it before.
Try and look at everything at once.
Break issues down into Blockers and Aspects.
Underestimate the impact of small changes.
Categorise each Aspect into an impact scope level.
Assume that your previous recommendations are still live, especially
if the site is updated very regularly.
Prioritise everything based on effort vs. impact.
Put every performance issue down to a tracking issue – your data team
will hate you.
Help developers. Do what you can instead of putting everything to
them. Call them, go and sit with them to implement things, do screen shares
to talk them through. Do whatever you can to make their life easier.
Give recommendations (even in passing) without looking into the issue
in detail. Clients and developers will lose trust in you, especially if they
waste time looking into implementing solutions without proper guidance.