170 likes | 443 Views
Crowdsourced Testing: A Quick Overview. I first heard of it (and the broader crowdsourcing topic) at the Jul 2008 TMF in a talk by James Whittaker (who said it was part of Microsoft’s vision of the future).
E N D
Crowdsourced Testing: A Quick Overview • I first heard of it (and the broader crowdsourcing topic) at the Jul 2008 TMF in a talk by James Whittaker (who said it was part of Microsoft’s vision of the future). • Piqued my interest as I was at the BBC wrestling with extensive browser/OS coverage requirements triggered by the BBC Charter. • uTest.com the only player in the game back then. • uTest had raised about $12m at that stage (now $37m), all of it after the credit crunch.
Crowdsourced Testing: A Quick Overview • Unlike crowdcasting (e.g. logo design) where the winner takes all, in testing everyone can contribute a piece of the overall solution. • That makes testing ideally suited to engaging a motivated crowd. • Go to the brilliant Zappers event run by TCL (uTest’s UK affiliate) and you’ll meet excellent crowd testers who are motivated by non-monetary factors (more on that shortly). • The existing maturity of test management tools has provided an excellent basis for the (essential) crowd management.
Introducing Utest.com • US-based company, pilot system launched in Feb ’08. • Circa 49,000 in the crowd (circa 3000 in the UK). But how many active? • Sales team in US (at least some of them), handing over to TCL (uTest’s UK affiliate) to project manage the cycles. • A functional management interface (but not commensurate with the funding received or the time since launch). • iPhone app available (but bugged as of last summer). • Fixed price model for the client, testers paid per bug or artefact on the other side of the equation. • TCL PM drives crowd selection, but with a client steer. • Daily reporting (if desired) with an optional test lead to sanity check submissions at minimal cost (approx £8 per day). • Offers functional and non-functional (although I’m not sure about the repeatability of the latter). Covers mobile, web and desktop.
Alternatives To uTest I’ve briefly looked at, but not used: • Bugfinders (UK company, client pays per bug). • Centre4Testing (Leverages C4T’s candidate pool, UK crowd, formally engaged to deliver coverage. In quoting for my case study they wanted a lot more info and lead time than uTest. About 10% dearer than uTest on the 1st cycle but about 50% cheaper for subsequent cycles. Also slightly cheaper on 1st cycle if you exclude uTest’s new client discount). I’m aware of, but have not looked at: • Bugpub • 99tests (Seemingly not associated with 99designs.com) I’m sure there are (and will be) others.
Crowdsourced Testers: Some Motivations • Practising their skills, breaking things, trying the latest software, having fun, increasing their market exposure through rankings, networking*, etc… • For many (most?), money is not a material factor and they have a day job for subsistence. * I’ve since met the uTest team lead for my case study and would happily engage him in a full-time role.
Crowdsourced Testing: Some Pros • Significantly more eyeballs on the case £ for £; compares extremely favourably with a single contractor. • Externalise accommodation costs (desks, PCs etc). • Ultra rapid ramp-up time (vis-à-vis recruitment) and feedback loop (potentially get bugs within minutes). • Flexible engagement model for reactively rolling testers on and off the project (e.g. for post-release testing). • Mitigates costs of test environment provision in the face platform proliferation. • Evening and weekend productivity at zero marginal cost. • Cheap form of usability testing if you know how to frame the survey.
Crowdsourced Testing: Some Cons • Lack of direct accountability (some ‘soft’ sanctions like star ratings and feedback to platform operator). • System knowledge vested in un-contracted resources. • Could be unsettling to in-house testers. • Could be seen as de-valuing testing within the organisation. • If it’s a desktop application and it gets leaked – a watermark system may be scant consolation. • Testers may (probably?) care less than their internal counterparts. • Need to provide strangers with access to your test instance. • Could confer unwarranted confidence; relies on stakeholders understanding these downsides.
Crowdsourced Testing: Applicability • Start-ups that can’t sustain in house test resource. • Organisations that aren’t exclusively reliant on the crowd (not withstanding my point above). • Agile teams that want out-of-hours productivity (especially for smoke testing purposes) and/or to mitigate the effect of having a single embedded tester working in mini-water falls. • Public facing systems that need to undergo testing out in the wild (corporate systems involving sensitive/financial data are much less appropriate). • Organisations where the testing workload is highly bi-modal. • Places that lack the time/environments to test in-scope OS/browser combinations. • Environments where exploratory testing has been culturally accepted. • Environments that may want to target users in specific geographies to enable, for example, localisation testing out in the wild.
Case Study: The Context • One of Europe’s largest price comparison sites with profits measured in the millions. • Wholly reliant on TDD (within Kanban) save for 1 over-worked manual tester located on the continent. • Requirement was for post-release testing of a UK-only web proposition in public beta. • No time or money to recruit contract testers. • 3 week window of opportunity for testing but with scope for it to slip. • No meaningful collateral from which to derive tests (just a load of binned post-it notes). • Engaged by the Programme Director who wanted to de-risk the testing and was open to suggestions. He knew about uTest but had no time/patience with working through the relevant questions. • I wanted crowdsourcing experience so I offered take this off his plate.
Case Study: The Solution • I became the uTest interface and clarified the commercials, legals, process etc. Need to read the terms – uTest can quote you as a case study unless you opt out. • uTest sales team willing to progress paperwork over the weekend. • Commissioned a UK-only crowd for 3 calendar weeks of exploratory testing at a cost of $3000 which factored in a $600 new client discount. We paid $250 to have a team lead (in addition to the TCL PM) to sanity check submissions. $3,250=£2,075. • uTest provided a fixed IP to enable test traffic to be removed from merchant billing. • Coverage: Windows: Chrome, FF, IE 7/8/9 (+ Mac/Safari). • Testers were given a minimal briefing with goals comparable in detail to a charter within session-based test management. • TCL PM provided lightweight daily reports tailored to our requirements (weekdays only).
Case Study: The Outcome • TCL invited 35 testers of which 17 accepted with 10 submitting defects. Around 80 defects (from memory) with a rejection rate circa 10%. Some useful site review feedback was also provided (at no extra cost). • The reporting and defect sanity checking worked well and made the internal triage process more effective. • Bug detection continued during weekends and evenings through to the early hours of the morning. • Rightly or wrongly, the client was delighted with the results. • Programme Director – who is a contractor - has vowed to “try and use uTest in as many future roles as possible as it worked brilliantly”. • Whilst recognising that the bar had been set low for adding value (post-release, no internal testers, non-complex web app etc) I also felt positive about the experience. • I felt it was too cheap; I wonder what time horizon the uTest backers have in mind for turning a profit.
Conclusions • A rich and burgeoning topic area in which venture capitalists are quite active. • Still a young area with massively un-tapped potential, but established enough to not be dismissed out of hand. • uTest looks dominant but there are other options. • In the right hands and circumstances it can be a very powerful addition to your context-sensitive toolbox. • My uTest experience was positive but the bar was set low. • Its disadvantages will be unacceptable for some clients. • In my view, crowdsourced testing does have legs which will see its adoption rise over the coming years.