Common Mistakes in Website Performance Optimization

Enterprises make a lot of mistakes when testing the performance of their websites. High-speed Internet access and ever more powerful servers, have given a good excuse to many IT managers and web producers to become a little bit lazy when testing and optimizing their websites.

Nevertheless, the benefits of performance testing are unparalleled: speedier website loads, lower hosting costs, more sales, increased equipment durability, happier customers and faster backups are but a few of the many advantages. Even when presented with these attractive gains, most project managers skim on testing to deliver projects on time or to avoid further delays. In this scenario, managers forget to fit performance testing in their projects timelines and then run into all sorts of problems. Here are five mistakes that enterprises make when testing the performance of their websites.

#1: Unrealistic Traffic Estimates
The first mistake is not emulating the real stress their websites will endure by failing to adequately estimate traffic. Adequate capacity planning involves estimating regular traffic conditions as well as predicting traffic spikes during busy days and geographic distribution of traffic and service loads. Some countries and regions will place higher demand on particular sections of your website and thus, contracting CDN services in those regions might make a lot of sense for your organization. However, none of this is possible without an adequate traffic estimation that takes into account seasonalities, regions, online services and timezone differences.

Traffic can be estimated by adding up the expected search engine traffic with the social media referrals and email lists traffic. Several free tools are available online and, along with historic server logs, they should be sufficient to estimate your web traffic adequately.

#2: Not Isolating Database Testing
A big no-no is failing to test your databases separately. Isolation of database components makes for easier and more accurate testing. This is even more important today, because of an ever increasing avalanche of data. You should pay especial attention to batch processes and remote web services because, believe it or not, there are the most expensive processes of most major data stores and are often overlooked by performance tuners.

Sadly, on the quest for the fastest launch possible, most companies skim on database testing and optimization and make their sites slower to read and index. This translates into decreasing traffic charts and slower search engine indexing.

In order to avoid this problem remember to comprehensively test and optimize every query, stored procedure and remote service but do so in isolation using the specialized tools provided by your database vendor.

#3: Uneven Unit Testing
In many companies, unit testing is not part of the daily routine. This is a major blunder, units should be tested and stress tested before making it to production or else! Unit testing makes your team accountable, fosters communication and levels the information technology understanding of your junior programmers.

#4: Neglecting Client-side Testing
Most companies don’t test their websites from a client perspective. They forgo slow Internet connection simulations and the like. But without a realistic idea of how your customers experience your website, little can be done in terms of caching, code optimization and CDN implementation to better customers’ experience.

The swift identification of poor user experience and bottlenecks can save you time and money, as well as improve sales. Furthermore, client-side testing allows to rapidly identify malfunctioning webpages in different browsers, platforms and mobile devices.

#5: Test Environment Differs From Production
Finally, enterprises fall prey to a very basic mistake: forgetting to test on machines comparable to their production servers. While you don’t need to test the on the actual production servers, servers with similar specs and connectivity help to highlight obscure issues related to the OS and hardware of the production environment. Complexity created by operating systems, processors, libraries, connectivity issues and the like make a very strong case to simply replicate your infrastructure for testing purposes. Specially when testing complicated websites, you should be as thorough as possible in simulating real operational conditions to make your tests as accurate as possible. This discipline minimizes surprises and allows making last minute changes to further optimize operations.

Although performance testing requires additional investment, time and discipline, your company will benefit greatly while avoiding many technical pitfalls. Google, Facebook and Amazon created impressive online businesses by “paying attention to the milliseconds” and optimizing their websites as much as they could. If larger scale and increased revenues are on your company’s horizon, give performance testing the attention it deserves. You won’t regret it.

The Great Educational Divide / La Gran Brecha Educativa

We need more people in the world that understand this!! (Cauchy’s Inequality)

Regarding my collaboration with Math Multipliers, I would like to share some important documents regarding basic education in China and Latin America. The contrast in math education between this two regions is quite staggering. China has clear guidelines to use math as a tool that enriches the individual’s mind and brings forth a higher level of societal understanding thru abstract thinking, to this extent, China executes an educational program three times the size of its latino equivalent.

For more on Chinese Math Education refer to Math in China, Mathematics Curriculum Standards, Prepared by the Ministry of Education of People’s Republic of China

A particular weakness of the Latin American system is forgoing training in inequalities and maximization problems. Hence, we end up with a society that lacks significant high-level understanding of maximization processes, statistical analysis and optimization procedures. In particular, very few Latin Americans now about the Cauchy-Swartz inequality, a very important, albeit somewhat advanced, mathematical concept.

To learn about the Cauchy Inequality refer to Starting With Cauchy.

Yes, the situation is worrisome, but programs such as Math Multipliers, sponsored by the CGIU and the Iberoamerican Math Olympiad seek to improve the situation by making this knowledge accesible to young students and fostering math studies with international math olympiads and other prestigious competitions.

Learn more about the tremendous benefit that Math Competitions bring to education.

——————————— In Spanish / En Espanol ———————————

Necesitamos más gente en el mundo que entienda esto!! (Desigualdad de Cauchy)

Con respecto a mi colaboración con el proyecto Multiplicadores Matemáticos, me gustaria compartir algunos documentos importantes de la educación en China y Latinoamérica. El contraste en la educación matemática de estas dos regiones es francamente sorprendente. China tiene lineamientos claros para usar las matemáticas como herramienta de enriquecimiento individual que le trae al individuo un mayor entendimiento de la realidad social a través del pensamiento abstracto y, en ese tenor, China ejecuta un programa educativo tres veces mayor al latino en cuanto a educación matemática se refiere.

Consulte (en inglés) Matemática en China y Estándares Curriculares de Educación Matemática – Preparado por el Ministerio de Educación de China para obtener una perspectiva más clara del sistema educativo chino.

Una debilidad particular del sistema educativo latino es la ausencia de entrenamiento en desigualdades y problemas de maximización. Por ende, graduamos a una sociedad a la que le falta un entendimiento significativo de los procesos de maximización, el análisis estadístico y los procesos de optimización. En particular, muy pocos latinos conocen acerca de la desigualdad Cauchy-Swartz, una de las más importantes en las matemáticas, aunque para algunos, un concepto muy avanzado para ser explicado en la educación media.

Para aprender acerca de la desigualdad de Cauchy visite Demostración de la desigualdad de Cauchy.

Sí, esta situación es inaceptable, pero programas como Multiplicadores Matemáticos, apoyados por la CGIU y la Olimpíada Iberoamericana de Matemática, organizada por la OEI, buscan mejorar la situación haciendo estos conocimientos accesibles a jóvenes estudiantes y fomentando el estudio de las matemáticas a través de prestigiosas competencias internacionales.

Descubra los beneficios que ofrecen las competencias de matemática a la educación. (En inglés)

Si no esta familiarizado con el tema le recomiento que descargue el documento Entrenamiento Básico en Desigualdades.

AgencyRank – A New Way to Reward Creative Professionals

State of the Art in Digital Agency Ranking

Our industry is a very subjective one. A client likes your design today and hates it tomorrow. You enter a Behance* profile and find very nice looking portfolios with sometimes, too much curation and end up with an under representative sample of work that only shows you the very best productions of a particular professional.


The last group in the graph is essentially award-winning work, the middle 95% is most likely representative of the usual performance of the professional, or agency, while the low 2.5% can be thought of as “unpublishable blunders”. The problem in the web design industry is that, right now, there is absolutely zero visibility into the blunders and very low visibility of the typical 95%. Agencies and designers only show their very best work and this makes it harder for us to pick the most consistent players. Both as employers and as end customers of digital agencies.

Agencies and sometimes designers are ranked primarily based in the number of awards they have obtained. This rank, while a subjective measure is also a very unfair one because it makes it harder for the unestablished designers to move up and get better clients. However, rankings such as these use only award counts to rank agencies. In conclusion the current of affairs as awards can be subjective, can over-celebrate brand association and can reward friendships and particular design styles.

Enter AgencyRank

A better way is to rank professionals and agencies based on the success of their work and the importance of each production. In this case, the judgment should be performed by clients not judges, so that we capture the very important 95% piece of productions.

The way to calculate this is very simple, we first need a catalog of sites in a given industry, say energy for example. Then we produce a public list with the sites and key public figures provided by clients: page rank variation (or logarithm variation to account for an exponential raise in page rank), bounce rate change, average time spent on page, and so on. You can click the following image to see a sample calculation**.

AgencyRank Calculation

In this particular competition, the judges assign more weight to bounce rate reduction because, probably, page rank is not naturally distributed in this particular industry. Or maybe this is a web design and not a web marketing competition. Making these equations public, giving clients the ability to self-report the numbers and come up with a more reasonable set of criteria can produce better rankings and help the whole industry rank creative professionals in a more reasonable manner.

A similar metric can then be self-reported in oDesk and the like to give us principals are more accurate view of the capabilities of each professional or group of creative types. Giving clients the wheel in this scenario also helps in capturing more of that 95% of the performance that still remains in the dark.

* Behance is a showcase site for creative professionals. A good tool for sharing and discovering the best work of creatives but not a particularly enlightening one when it comes to hard-figure analysis.

** I don’t account for page rank’s exponential nature. Feel free to make adjustments to the formula and consider including logarithms in your page rank calculations to adequately account for big jumps in page rank.