Traceable time
If marketers want to deliver on transparency and unlock the transformative benefits of AI and blockchain, then it is imperative that they get their timing absolutely right, explains Simon Kenny
With recent high profile media reviews being announced by major advertisers referencing transparency as a key criteria for appointment, it’s time for a shift in the level of accountability and verification of the digital data chain delivered back to advertisers.
The good news is that the technology to produce irrefutable and traceable data records is currently being implemented in Fintech, where it offers a gold standard of verification based on a new quantum of traceable time, fit for the AI future emerging fast around us.
When a single automated trading server can execute multiple decisions within a single millisecond, (and it takes many of them spread across different data centres to execute a single transaction), it is not surprising that monitoring and reporting a whole campaign of automated advertising activity is a big challenge for marketing practitioners.
Add to that a requirement to find ways to maximize the potential of new applications like AI and Blockchain, then you realise that everyone will need highly coherent data records to support this activity.
But unless there are specialised protocols in place, one of the most fundamental elements of your data records will probably be unreliable; the clocks on the different machines in your transaction chain will not be synchronized to the right time (in some cases by a few seconds) and therefore the timestamps in your data records will be wrong.
A few seconds may seem like a trivial sensitivity, but we live in a fast-paced instant gratification digital age, so the implications for the utility of machine data and transaction records for other applications are profound. Key metrics such as buy-side & sell-side reconciliation, time-to-match and time-to-serve are meaningless unless the clocks on all the servers agree.
Imagine a transaction process, such as buying and serving an advertisement, involving ten different servers, spread across several different data centres, which takes 10 milliseconds (a millisecond is 1,000th of a second) to complete.
If the last server in the process has a clock 100 milliseconds slow and the first server in the process has a clock 100 milliseconds fast, then the timestamps on the data records will say the completion of the transaction chain happened before the chain began.
Repeat this process thousands of times and it will be almost impossible to make sense of the sequence and intervals in the transaction process from the timestamps records.
You cannot use them for reconstructing events to confirm process or assign responsibility for a machine action; they are not a good source of data for an AI application seeking actionable insights and, if blockchain does become widely adopted, the timestamps encrypted in the ledgers will potentially not be valid.
Fintech precedent
Fortunately advertising is not the first industry to have this problem; the financial markets were very early adopters of automated trading systems and the European Securities and Markets Authority (ESMA) that regulates financial services has identified this issue and two years ago introduced regulations to address it.
Essentially ESMA requires that all market particpants, beginning January 2018, have accurate, traceable, synchronized timestamping at the application level in a trading server, to enable the regulator to reliably reconstruct trading events from these time-stamped data records.
In effect, ESMA will use accurate synchronized time to achieve transparent reporting of automated trading by making the market participants improve their collective timekeeping.
So how will this work and can it transfer to advertising?
There are different models, but there are three basic steps to achieving accurate timing across an inventory of servers:
– a connection to a primary source of Universal Time (UTC) such as GPS, GLONASS or Gallileo, to which you synchronize three highly accurate clocks
– a client application on the server which needs to be synchronized that will take the accurate GMC time and use it to steer the local system clock
– a measure of the granularity of the timing within the server (which varies depending on how the server is loaded) and logging of the resulting timestamp so that it is traceable.
Traceability is important, it is what shows you have the right time. A system may have a connection to a time source, but the timestamp is only traceable if an event in the log can reference an unbroken chain of comparisons from the log, back to the primary timing source, to prove it is right.
Taken together they enable a virtuous cycle of accuracy that transforms the utility and reliability of data records without a need for significant new infrastructure or disruption to existing applications.
Can it transfer to advertising?
It can transfer because the digital/automated trading advertising industry uses the same basic architecture as Fintech so the new timing technology, once installed, can accurately timestamp digital advertising activity, make the data records coherent and provide a much more reliable base for new applications that will require orderly timing.
It will also enable the traceable reconstruction of events, and potentially provide the basis for ‘timing signatures’ on processes that define the average timing pattern of repeated activities, plus standard deviations, so enabling the identification of anomalies that might indicate a process was not completed properly.
The advertising industry has made an enormous commitment to digital advertising. It must take advantage of this innovation, already being utilised and available now in other industries, in order to claim a return on that investment.
Simon Kenny is CEO of Hoptroff London