Tuesday, February 06, 2018

Intent and execution in Go's dep tool

I'm starting to understand how the golang dep tool works, especially in terms of adding constraints to a dependency.

Taking the sample case from https://engineering.pivotal.io/post/go-dep-intro/, if we wanted to specify that we were using the dep-changes branch of github.com/fatih/color, we have two choices:

  • manually add a line with branch = "dep-changes" to the entry in Gopkg.toml, followed by a dep ensure to update Gopkg.lock and the vendor directory.
  • assuming we weren't already using the package when we did dep init or our most recent dep ensure, then on the command line enter dep ensure -add github.com/fatih/color@dep-changes, ensuring we add the code reference before we execute our next dep ensure.

Notes from UKUPA talk on UX and ROI

An excellent evening at Sapient Nitro, herewith some notes and a few thoughts. There was also a good Twitter stream for #UXROI, and @yahnyInLondon did beautiful sketchnotes.

There were four speakers. First up was Joe Leech (@MrJoe) of CxPartners in Bristol (slides here), who pre-emptively trumped all the numbers to follow with a figure of $3.6 billion, the money that flows through the systems he's involved with, each year. He pointed out that the climate for UX practitioners was much more competitive now, with easy wins being a subject for nostalgia, and also mentioned the idea of MVT (multi-variate testing) as "a threat to UX", an interesting idea which coloured my perception of the rest of the evening.

Joe distinguished between Visionary Metrics ("what and why",- emotional, attitudinal factors such as Net Promoter Score) and Strategic Metrics (operational stuff such as conversion ratios, as provided by MVT / AB testing).

His checklist for metrics to be useful, rather than just analysis paralysis:

  1. Does the metric have a timescale?
  2. Does it have a benchmark?
  3. Does it have a reason to be reported?
  4. Does it have an associated action?
He then gave us two key metrics to think about:
  1. Net Promoter Score: asking how likely a customer is likely recommend your product, site or service to a friend of colleague, in order to judge their overall perception of quality. Very popular with management, in some cases included in personal targets.
  2. Conversion Ratios: important to attach a dollar amount to drop-outs at each stage, makes it easier to justify work on that area. What's a good conversion rate? Really hard to answer in general - it depends, but likely to be between 2% and 7%. How much can I expect conversion to improve? It depends, but "by about 0.25% to 0.5% for a full end to end UCD process"