1 second of download acceleration gave a 10% increase in conversion: Staples case

    Staples is one of the world's largest office supply chain chains with more than 2,000 stores in 26 countries. The company also conducts business online, its online store is one of the largest in terms of turnover in the United States. The company has always recognized that the site’s UX is extremely important, and that the site’s loading speed is one of the main factors affecting the UX.

    By the time Staples decided to pay more attention to speeding up its own site, the data was already known that with increasing site loading time, the bounce rate increased (according to mPulse / Soasta and Air.rf):



    And the conversion was falling:


    For work on the site’s speed, the company has created a cross-functional team consisting of representatives of several departments:

    • Management: its representatives determine what functionality from the point of view of the user should be present on the site.
    • IT: defines the ways in which the required functionality will be implemented and its subsequent implementation.
    • Marketing: responsible for the design and optimization of the images and videos used.
    • Merchandising: determines how the product will be visually presented on the site.
    • Analytics: links everything together and provides data analysis.

    The work began with the cataloging of all the various functionalities that are supported on the site.

    The company established the rules: every week the team met and discussed 1-2 site functionality. The speaker was the so-called "owner" of this functionality, that is, the person from the company for whose business tasks it works. The team discussed these features in terms of speeding up loading and the justification for abandoning it or some aspect of it in order to speed up loading the site.

    Progress


    One of the important decisions that had to be made was whether to consider the target KPI as the full load time (FPL - Full Page Load) or whether it was necessary to focus on improving the time until the user was able to interact with the page (Time To Interactive). The team eventually decided to focus on full load time, since this is a more specific indicator. While the indicator "time to the possibility of interacting with the site" is somewhat vague and allows different interpretations in the sense of what features should be available to the user at this moment.

    Website Banners


    A lot of banners were used on the site, and in many cases they were too “heavy”, and the image optimization parameters were chosen arbitrarily by different designers.


    Very visually similar images, different versions of the same banner could “weigh” 25 KB in one case and 250 KB in another.

    The team further optimized the existing banners and implemented the process of creating banners in order to avoid this problem in the future.

    A / B testing


    The practice of conducting A / B testing in the company was ineffective. The company used two different A / B testing tools, each of which added its own script to the site with external calls.

    The team abandoned one of the A / B testing tools, and also implemented the process of ending the experiment to make sure that the testing script that was no longer needed was removed from the pages where the testing ended.

    Calls to other third-party services


    I had to put in order a decent amount of calls to third-party services, of which there were a lot on the site: analytical services, advertising systems, recommendation and feedback services, buttons / widgets of social networks.

    The team deleted some of them, but notes that further optimization is still possible in this regard.



    Database Access


    Some pages of the site made 200 accesses to the database. The team optimized them to 10 per page.

    Javascript


    The script code on the site has been substantially rewritten, their size has decreased by 48%, and by 52% in the lines of code.

    CSS


    Styles on the site have also been optimized, in volume they have been reduced by 83%, and in lines by 88%.

    Lessons learned


    1. Commit to significant changes


    Initially, the team tried to work according to the “sprint” model and to make some improvements with each new release of the site version. However, it turned out that such minor improvements were almost imperceptible. There was a feeling that the work was underway, but there was little or no result.

    At the same time, due to the ideology of the "sprint" there was no time to make really extensive changes that could produce a significant, significant result. As a result, the team switched to introducing big changes, albeit not immediately and not with every release.

    2. Pay attention to specific pages.


    Quite quickly, the team realized that in order to achieve significant results, it was necessary to focus on specific pages that are visited by most users - the main page, product card, search, login, basket.

    results


    • Median loading time of the main page of the site decreased by 1 second
    • Download time was reduced for 98% of users. At the “slowest” users, it was reduced by 6 seconds.
    • Conversion up 10%

    Note on the interpretation of the results from experts from Soasta, the authors used to optimize the mPulse user behavior tracking tool:

    These results do NOT mean that ALL users began loading the main page 1 second faster.

    For someone, acceleration did not happen. Since the site is visited by thousands of users with a wide variety of device configurations and network connections, the spread in the loading time of the home page for different users is very significant. Therefore, this case uses the median load time as an indicator of overall acceleration.

    It also does NOT mean that ALL users have started to convert 10% better.

    Those users for whom pages actually loaded faster gave such a good conversion that on average for all users it improved by 10%.

    In conclusion, we add that such a case is not unique or unusual. UX professionals and conversion optimization specialists note that this is a fairly stable result when a decrease in the site’s loading speed leads to an improvement in conversion.

    Here are a few studies that support this statement:

    • A study by Soasta showed that speeding up site loading for mobile device users gives a conversion increase of up to 27%.
    • Tests conducted by Radware showed that only 2 seconds of additional delay in loading during the checkout process gave an increase in the percentage of incomplete transactions of 20%.
    • According to a BlueTriangle study , accelerating the loading of a product’s card in online stores by 0.1 seconds gives from 1.2% to 3.5% of revenue growth.

    Therefore, if the site loading speed is within your competence, pay close attention to this indicator. Perhaps you will be the person who will provide the increase in your company's revenue right now.

    Also popular now: