The document discusses how programmatic advertising spending, campaigns, and impressions tend to increase towards the end of each month, quarter, and half-year due to many campaigns being booked on a calendar month basis and starting late in the month. This creates higher competition for inventory at month-end, leading to higher prices and less availability on preferred sites. Blurring the lines between calendar months by starting and ending campaigns on non-standard dates could help advertisers get ahead by avoiding busy periods and utilizing less expensive periods.
7. So today…
I would like to show you the effect that this way of
thinking results in when it is aggregated across the
industry and what/how you can get a head start
if only you are able to blur the line between
calendar months, for at least some of the
campaigns you work on, and start and end them
any day other than the
first and the last of the month.
19. What did we just see?
Spend
Campaigns
Impressions
= demand increases towards the
- end of the month
- end of the quarter
- end of half year
Why is this
happening?
22. To recap…
• Campaigns, in the era of programmatic buying,
are still booked on a calendar month basis
• Many campaigns start late & crammed into less
time to deliver
• The combination of these creates a pattern that
is purely the result of human intervention
• There is higher competition for the same pool of
inventory towards the end of the month (30-
50%+ increase)
23. Why care?
• Higher competition results in higher prices
• It also means less availability on sites you may
prefer to be on
• And finally, this way of booking also limits
algorithmic learning
24. To sum it up…
Yes, many times there is need for calendar based
planning, and for campaigns to run accordingly. But
if not, you could…
• Up-weight campaigns in the beginning of the
month
• Try to avoid the busiest periods is possible
• Challenge your clients on the end date
• Consider running campaigns longer to utilise the
‘cheap periods’
Editor's Notes
The following scenario may sound familiar to you:
You speak to a client and you tell them, it is time you think about programmatic media. It’s time you try this wonderful new way of media buying.
You tell them about the algorithmic learning, the optimisation, and the efficiency gain. And the client gets excited and says: ‘Yeah, I want to be a part of this!’
And you go: ‘Great, tell me about the campaign, and what is it that you want to achieve!’ And the client goes: ‘I want the campaign to run from the 1st to the 31st of January and to run only on these five sites.
And your dream.. [of getting the client to think programmatic quickly becomes ‘problematic’.]
[And your dream…] of getting the client to think programmatic quickly becomes ‘problematic’.
Because this way of thinking isn’t programmatic. For a number of reasons, but today I will look at one particular aspect. TIME.
Time & timing. – in terms of start and end date.
Thinking about it how the majority of the industry still thinks about it:
a media campaign that usually starts on the 1st and ends of the 30th.
Or starts on the 1st day of the quarter and ends on the last day of the quarter.
And what this means when aggregated on an industry level.
So let’s look at data.
Big Data.
The fact that everything is measurable in digital media means we can monitor and analyse patterns.
And some patterns make perfect sense. For example…
Christmas 2013
Chart 1: daily
- Red represents impressions- Purple spend ($), and - Green represents number of live campaigns
(the colours will continue to represent the same metrics throughout the presentation)
No surprises here: December is a month of two halves
Starts off really strong, peaks around mid-month, then just a few days before Christmas, volume of impressions starts tapering off, together with number of live campaigns and spend
Christmas 2013
Chart 2: weekly (taking wk 1 as 100%)
However, dig deeper and there are more patterns that don’t necessarily make sense. Or at least not at first sight…
Let’s look at the month of May ’14
First by day:
steady increase during the month across all metrics
Then by week:
volume of bids are 150% higher on first week than on the last
Interesting. No apparent reason for such a spike in May.
[How about we check another month. Let’s look at April ’14 ]
[How about we check another month. Let’s look at April ’14]
Again, we are seeing an increase towards the end of the month without any obvious reason.
A whopping 62% and 42% increase in spend between week 1 and week 3 and 4.
Let’s see if we can repeat this increase over a quarter…
Let’s look at Q2 2014
Obviously there are more fluctuations, but the trend line is pretty clearly on a growth trajectory between the start and the end of the period.
Bear with me for one last chart.
This time, let’s look at Jan-Jun 2014 period.
This time, let’s look at Jan-Jun 2014 period.
There are some fluctuations again, but it’s pretty clear that impressions and spend seem to increase towards
the end of the month,
end of the quarter, not to mention
the end of the Financial Year.
All the while supply remains relatively steady.
Let’s recap.
What did we just see?
Spend : up
Campaigns : up
Impressions : up
= demand increases towards the end of the month & end of the quarter
Why?
What’s going on?
Have you ever seen a campaign to start on the first of the month and end on the last day of the month?
And what about the actual campaign start dates?
Have anyone ever seen creative to be on time?
Ok. Let’s be fair, there are other reasons why campaigns can be late.
And they are.
Lots of campaigns start late, all the while they still needs to finish on the original end date.
So they are crammed into a shorter period of time yet forced to burn through the same budget.
So to summarise, here is what’s happening:
Campaigns in the era of programmatic buying are still booked on a calendar month basis en masse.
Campaigns start late – en masse.
The combination of which creates a pattern that is purely the result of human intervention:
There is higher competition for the same pool of inventory towards the end of the month (30-50%+ increase).