What growth-driven design statistics exist, and what can each one establish?

Luke Summerfield, who created growth-driven design at HubSpot, told the Build a Better Agency podcast in October 2017 that "at the six-month mark, Growth-Driven Design clients were getting 16.9 percent more leads than those that did traditional. And 11.2 percent more revenue from their website" [3]. Market Veep repeated the 16.9% figure in July 2018 and attached a separate claim of a 14% traffic increase to the same survey [5]. The methodology's own site now publishes 14.34% and 12.56% for those two metrics over the same six-month window [1].

Leads move down by 2.56 points between the two versions and revenue moves up by 1.36 points, which rules out a single restatement error running in one direction. The current site version also carries a qualifier the interview did not: the figures describe agencies that used the method on HubSpot websites against WordPress websites [1]. That wording changes what the comparison is, because it puts a CMS change and a method change on the same side of the line, and nothing in the published material separates them. The underlying report is a gated PDF marketed to agencies as data for prospect pitch decks, so the raw tables are not public and nobody outside the survey can settle which version is right [2]. If you quote one of these figures, name the publisher and the date alongside it, since the two versions disagree and no correction appears on either page [1][3].

The difference is 48 days, or about seven weeks [1]. Summerfield described the 60-day figure as covering the strategy phase and the launch pad website together, against 108 days for a traditional design that then gets set and forgotten [3]. Both sides are survey averages reported by agencies, and neither is a measured build.

A launch pad is a reduced first site built around the highest-impact pages, with the rest added in later cycles, so reaching launch at 60 days is not the same event as completing the website the 108-day project delivered [3]. The method's own launch pad page shows what individual builds look like next to that average: Teazzers launched in 96 days, Sprocket Rocket in 100, and the Pittsburgh Parks and Integrate builds in 120 each, with Distil Networks at 40 [8]. Four of those five published builds took longer than the 60-day survey average, and two took longer than the 108-day traditional average, which is a fair check on how much weight the 60-day number can carry for any particular project.

growthdrivendesign.com carries the footer "2017 Growth-Driven Design (dba of HubSpot). Made by the Lean Labs GDD Team.", so the site publishing the survey results is a HubSpot doing-business-as name built by an agency that sells the method [1]. The Agency Management Institute page is a podcast transcript in which the host interviews the person who created growth-driven design at HubSpot, which makes it a second publication of the creator's own account, not outside coverage [3]. Apart from one A/B testing paper, every figure on this page reaches the reader through HubSpot, through the method's creator, or through an agency that sells the method [10].

Market Veep's "2 weeks late" figure cites "the 2017 State of GDD survey responses" by name, so the statistic most often introduced as outside corroboration is the same survey arriving through a second domain [5]. Market Veep's "Only 49% of website redesign projects finish and launch on time" links to a HubSpot file, science_of_website_redesign_presentation.pdf, whose PDF metadata gives a creation date of June 13, 2011 [6]. Inside that deck the 49% appears on a pre-webinar holding slide under the header "Did you know?", with no sample size, no denominator, and no year attached [6]. The deck's methodology slide describes its whole data set as "Surveys of 100+ consumers, 100+ marketers, 100+ agencies" [6]. A third Market Veep claim, that 54% of companies say a redesign will take over six months, links to a HubSpot blog post that today reports a different statistic: 57% of more than 6,000 surveyed customers planning a redesign that year, which is where the 3,500 count in Market Veep's sentence comes from [5][7].

The 2011 deck's cost table reports an average of $54,596 against a median of $10,000 across a sample of 60 projects, with a high of $1,000,000 [6]. Removing that single largest project drops the mean to $38,572, still close to four times the median, so the $55,000 figure that gets quoted describes the tail of a small 2011 sample and not a typical project [6]. Its timeline table reports an average of 5.1 months and a median of 4 across a sample of 116, with a range of 1 to 24 months [6].

Ron Kohavi, Alex Deng, Roger Longbotham, and Ya Xu, writing in KDD 2014 from experiments run at Amazon, Booking.com, LinkedIn, and multiple Microsoft properties, state that "for web sites like Bing, where thousands of experiments are being run annually, most fail, and those that succeed improve key metrics by 0.1% to 1.0%, once diluted to overall impact" [10]. The same paper puts Bing's success rate on ideas at "about 10-20%" and reports a prior probability of success of one third as the average across multiple Microsoft experiments [10]. That paper is the only source cited here with no commercial interest in growth-driven design.

A six-month lead difference in the low double digits is not implausible next to a program of many small changes, because the 0.1% to 1.0% range describes one diluted experiment and not half a year of accumulated work. The same paper's third rule is aimed directly at readers doing what this page invites, which is reading someone else's published percentages: "your attempts to replicate stellar results reported by others will likely not be as successful" [10]. The authors list four reasons: case study quality varies and is rarely peer reviewed, what works in one domain may not work in another, novelty effects fade when tests run too short, and effects get attributed to the wrong factor [10]. All four apply to the growth-driven design case studies, which were written by the agencies paid to run the projects and published without baselines [8][9][11].

The only adoption figure in circulation is 940 agencies across 50 countries, and it appears in Luke Summerfield's speaker bio on the October 2017 podcast page, describing the method's first twelve months [3]. It is a self-reported count from the person who built the program at HubSpot, and no count has been published since.

growthdrivendesign.com/report still offers only the "2017 State of Growth-Driven Design" Agency Edition, gated behind a form, with the page stating that "350 GDD Agencies shared their responses" [2]. No 2018 or later edition is offered there, a search for one returns only articles restating the 2017 figures, and the methodology pages themselves still carry a 2017 copyright line [1][2]. Anyone publishing a present-day adoption or performance number for growth-driven design is extrapolating from this material or reporting their own client base.