WebPerfDays follow-up: 36 questions about web performance tools, measurement, and best practices

Today’s post is kind of like those blog posts where people answer random questions about themselves. Just a bit geekier. 🙂

When I was at WebPerfDays last week, I was walking past the ubiquitous board of post-it notes and started thinking about the fact that at every conference, so many of those questions and ideas go undiscussed. So I decided to snap each one and have made my best attempt at answering every question.

Before you read these, a big caveat: With many of these questions, the answer really depends on the situation. I’ve noted this where it’s particularly relevant, and given general answers everywhere else.

1. Why are Google Analytics so slow? Watch page loads – it’s always waiting on Google.

If you use the right script it should not affect your users.

2. How much impact can be made by optimising CSS selectors and does this have a trade off with maintainability?

Very little for most pages.

3. Do mobile networks do their own TCP optimization to devices?


4. Can performance + webfonts coexist?


5. Experiences with Android web driver?


6. Will HTML5 make things worse?


7. How build and deployment shapes our software architecture at thetrainline.com?

Don’t understand this question. Anyone?

8. GeoDNS versus Anycast Servers


9. Can we make significant improvements in website performance without further improvement in browsers or protocols?


10. ASP.Net server side instrumentation, analysis, and interpretation

Ask Richard Campbell.

11. Pros and cons for loading 3rd party JavaScript in head/footer best practices

Footer, if possible.

12. Should I use Google or Microsoft CDN for my JQuery, JQuery UI, etc. references or host them myself?

Either. Shouldn’t matter much.

13. Offline and embedded web COAP

Nice and lightweight — functional enough?

14. Large companies and embedded web


15. Have there been any studies of how good Time To First Byte is as an approximation for true server time? e.g. DNS, connection time

Yes: here and here.

16. API performance


17. GEO – Distributed

Can be complicated. Do you really need it?

18. Responsive image format

There’s a group working on a standard.

19. Using web analytics to identify (and fix) web performance issues

I love it. See here, here, and here.

20. Metric-driven development

Does any other type warrant discussion?

21. No SQL Scale and Speed and survey of options

Not my area of expertise.

22. CDN vs. web proxy in some datacenters

Can someone clarify this question?

23. What is the best way to get customers interested in web performance when their website or web apps don’t sell stuff?

Find a metric they care about (e.g. productivity, retention, bandwidth), and I bet it connects back to speed.

24. Tonnes of great stuff at Velocity – what do you start with?

Measure your performance on WebPagetest. Iterate and measure again.

25. How many users is enough? When and what to deploy web KPI – new tech release schedules and users based?

More more more.

26. MVC JavaScript (backbone)

It works.

27. How many data centres?

Do you need the complexity of two, or is it for your ego?

28. Continuous delivery versus bureaucracy

Easy to say, hard to do. Do A/B development: put together a team of both and see who enchants the business.

29. Do the radios in 3G dongles behave the same as phones (from sleep to work)?

I would assume yes.

30. Fave perf. tools and what is missing?

See Steve Souders’s post for tools. Missing: Resource timings.

31. Discussion on what is the best timer to measure onload, first interaction, anything else? Multipage times

All of it. Correlate to biz metrics.

32. What are the best RUM tools?

Boomerang is a good place to start.

33. Has anyone used lo-fi RUM (stopwatch timing etc.) to validate their collected metrics/KPIs?

Yes. I should write a post about it.

34. Best way of building performance testing into CI build?

Script WebPagetest or HTTPwatch or Web Page Speed.

35. Continuous performance analysis tools

Take a look at WPT Monitor.

36. Should we expect FEO as standard feature of our CMSes?

Yes. Basic features by 2014. CMSes are usually two years behind.

For the sake of expediency, most of these answers are really short. If you want elaboration on anything, let me know in the comments.

Related posts: