11 BEST PRACTICES

FOR PROGRAMMATIC
ADVERTISING
…and the setup tips that marketers need

PRINCIPAL AUTHORS
ROLAND SIEBELINK,
HEAD OF CAMPAIGN QUALITY & PRODUCTIVITY, ROCKET FUEL INC.

JARVIS MAK,
SENIOR VICE PRESIDENT, CUSTOMER SUCCESS, ROCKET FUEL INC.
2

INTRODUCTION
This whitepaper is about
the practical ways to make
programmatic advertising more
effective. It focuses on how to
set up programmatic advertising
programs for success, and how to
coax the best possible performance
out of them.

People who will benefit from

this whitepaper include media
planners, marketers, ad-operations
experts, and media analysts whose
primary focus is to actually run an
This always-on approach is a vast
PLEASE NOTE:
advertising program, as opposed to
improvement to what marketers This whitepaper does not focus
buying it.
have long considered best on what programmatic advertising
practices in traditional advertising.
HOW TO AVOID As in every paradigm shift, a little
is, what makes it different from
traditional media buying, or how
COMMON extra knowledge can go a long to select a good programmatic
PROGRAMMATIC way toward helping you gain a buying partner. It assumes that
ERRORS sustainable advantage over your the buying decision has already
competitors. been made. If you are looking for
Clearly, marketers have a lot of
answers to those pre-buying-
questions about programmatic
BEST PRACTICES decision questions, we strongly
buying, and a need for
recommend downloading the
simple answers. The world of FOR PROGRAMMATIC
whitepaper, “10 Questions on
programmatic buying is evolving EXECUTION Programmatic Buying... and the
so quickly that what might have
This whitepaper is a checklist Answers Marketers Need.“
been accepted general knowledge
of simple tactics that can help
last month could be completely
you make your programmatic
obsolete next month. Rocket
advertising programs more
Fuel‘s “always-on“ philosophy of
successful. We’ve also included
programmatic advertising lets
examples of actual use cases
us test, optimize, and optimize
throughout to help illustrate the
advertising programs every minute
best practices we describe.
of the day, every day. As soon as a
program can be improved, it is.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 3

11 BEST PRACTICES
ONE: INTEGRATE PROSPECTING AND RETARGETING ............................................ 04

TWO: COLLECT DATA ON EVERY PAGE ..................................................................... 06

THREE: OPTIMIZE TOWARD REAL BUSINESS GOALS ............................................ 08

FOUR: RELAX YOUR CONSTRAINTS ......................................................................... 10

FIVE: TEST DATA MODELS HEAD TO HEAD ..............................................................12

SIX: TEST MULTIPLE CREATIVE VARIATIONS LIVE ...................................................14

SEVEN: ENTRUST DAILY OPTIMIZATION TO THE ALGORITHMS .............................16

EIGHT: EMBRACE VIEW-THROUGH CONVERSIONS ..................................................18

NINE: SHARE RESULTS IN REAL TIME .......................................................................20

TEN: RANK VENDORS OPENLY ..................................................................................22

ELEVEN: SELECT A PROGRAMMATIC PARTNER

THAT EMPOWERS ALWAYS-ON MARKETING .................................................. 24

Artifical Intelligence. Real Results.

4

ONE:

INTEGRATE
PROSPECTING
AND
RETARGETING
Programmatic advertising works best
when it’s allowed to combine upper-
funnel and lower-funnel efforts SHARE VALUABLE MODEL
LEARNINGS
in the same programs. The best
Having access to all funnel stages multiplies the amount
practice is either: of data generated, which greatly speeds up learning
cycles and increases effectiveness.
1. Do not delineate funnel stages at all,
Assigning certain vendors only to specific parts
and let programmatic vendors compete of the funnel, such as using separate vendors for
fairly on full-funnel effectiveness, or prospecting and retargeting, means each vendor
needs to build up its own data about consumers,
2. if the advertiser insists on delineating interests, actions, and behavior.

funnel stages, let each programmatic The best practice here is to let vendors work on this
vendor choose which stages they want funnel end to end. This enables them to leverage
more data and capture more consumer actions in their
to compete in. algorithms, thus deriving far better funnel models for
the advertiser‘s product.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
SPLIT THE RETARGETING POOL splits the retargeting pool. The more flexibility
you have to adjust the allocation, the better.
It used to be a best practice to prevent bidding wars
for retargeting impressions by assigning retargeting 4. Remove all other vendor pixels from the
exclusively to one vendor and prospecting to others. advertiser’s website.

This is no longer the case.

5. Your retargeting pool will now split cleanly

First, splitting the retargeting pool between different between the vendors defined, according to the

vendors prevents bidding wars more efficiently than allocation you determined.

exclusivity does, with far more day-to-day control.

6. After an initial evaluation period, begin adjusting

Second, letting programmatic vendors compete full- your retargeting allocation in favor of the

funnel provides more insights and a better use of the better-performing partner. You‘ll want to confirm

most advanced technology. Ultimately, this provides a sustained performance before switching over

better ROI than letting “specialized“ vendors cherry-pick completely to a single partner.

the easiest conversions without replenishing the funnel.

7. Don‘t rest on your laurels. Implement another A/B
test six months to one year after you begin so
Finally, retargeting can be so expensive that it is crucial
you can ensure your partner of choice continues
to keep comparing the ROI—not just between this month
to be the strongest competitor.
and last month, but also between competing vendors.

USE THE SMARTITIONER SCRIPT  FOR EXAMPLE:

DOUBLED PERFORMANCE FOR A
Setting up a simple script on the advertiser‘s website
CREDIT-CARD CLIENT
will help split the retargeting pool fairly across several
programmatic advertising partners. A major credit-card company wanted us to split their
campaigns between prospecting and retargeting goals,
1. Download the Smartitioner script, which with 50% of the budget devoted to each, and using a
Rocket Fuel has contributed to the open-source separate creative for each stage.
community.
http://www.github.com/rocketfuel/smartitioner When initial CPA results were disappointing, the
agency recommended running one end-to-end tactic
2. Pick the vendors that you want to split the including both prospecting and retargeting. This let our
retargeting pool between and assign a percentage algorithms decide how to balance the stages, and which
of the allocation to each of them. For example, creative to serve, at which moment.
you can give one partner (the incumbent) 75% of
the retargeting traffic, and another partner (the Almost instantly, we saw the prospecting/retargeting

challenger) 25%. balance move away from 50% /50% to 85%/15%, and
the CPA for the overall program drop by more than half.
3. Implement the script on the advertiser‘s website, Once clients try integrated prospecting/retargeting
or use a tag-management system that cleanly tactics, they rarely return to separate tactics again.
6

TWO:

COLLECT DATA ON
EVERY PAGE/URL

Programmatic advertising works

best when it knows as much
as possible about consumers
previously exposed to the
advertiser‘s owned or paid
media. This is why advanced
programmatic vendors ask an
advertiser to not just place a pixel
on a conversion page (or a few key
pages on their website), but on all
possible pages and other digital
media that users may access.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 7

LEARNING FROM The third most common error is reusing the same pixel
for vastly different marketing activities, without passing
NON-CONVERSION ACTIVITY a parameter where the pixel was placed. This can be
Marketers often believe that pixels only matter for confusing to the programmatic partners‘ algorithms.
retargeting. The truth is that information about the
consumer is valuable at every stage of the funnel. PLACE A UNIVERSAL PIXEL AND A
CONVERSION PIXEL
The earlier that programmatic advertising vendors can
build a profile of an anonymous consumer, the more data A good programmatic advertising vendor will propose

they can gather about the conversion funnel, and the at least two pixels: One specific conversion pixel for the
conversion page and one universal landing pixel for all
easier it is to optimize the entire advertising program
other pages.
toward the funnels that have proven most ideal.

It is a best practice to implement those pixels as per the

recommendations of the vendor:

• The universal landing pixel in the footer that gets

attached to each page

• The conversion pixel on the conversion page

In addition, another best practice is to have separate

pixels for all the different marketing channels.

 FOR EXAMPLE:
QUADRUPLED PERFORMANCE
OVER DIRECT COMPETITOR
Placing a pixel on every page can drive up performance
2–4x, a much bigger factor than many marketers might
DON‘T LIMIT EXPOSURE OF A guess.
VENDOR PIXEL
We can illustrate this with an interesting case of two
The most common error is to limit pixels to the
telecom advertisers—let‘s call them ComTel and TelCom.
conversion page. The few seconds of effort webmasters ComTel and TelCom offer similar products and services
save by not placing a universal pixel could well result in at similar prices. They focus on similar segments, and
an ROI that is hundreds of thousands of dollars lower run similar advertising programs. The only difference is
than it could have been. ComTel had a pixel on its conversion page only, whereas
TelCom had a universal pixel that gathered data on
The next most common error is to place the pixels of
every page.
competing vendors inconsistently. Fair competition
requires a fair exchange of consumer data, which can The gap in their performance was striking. ComTel‘s CPA
only be achieved if each competitor has the same was invariably less than half of TelCom‘s, and—more
number of pixels in the same places. typically—closer to a quarter.

Artifical Intelligence. Real Results.

8

THREE:

OPTIMIZE TOWARD REAL

BUSINESS GOALS
Programmatic advertising drives the best results for advertisers when
agencies and marketers are able to move beyond the short-term
advertising goal. If they can identify the underlying business goal, and
link that goal directly to the programmatic efforts, marketing results will
suddenly become much more relevant for the advertiser, making the
contribution to the advertiser‘s bottom line rise exponentially.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 9

ROI IS HIGHER WHEN DRIVING DO SHARE THE REAL BUSINESS

BIGGER-PICTURE RESULTS RESULT AND UPDATE IT
An example is moving from a CPA focus to a customer REGULARLY
lifetime value (CLV) focus. Suddenly, the program will A best practice is to share the real business result, or
ignore the same old repeat customers and drive new even to ask for it if the advertiser did not include it
consumers to the advertiser—people who are highly in the original briefing. Allowing the most advanced
likely to buy multiple times and at a high value. programmatic advertising companies to optimize
directly toward this business metric will generate
The perceived ROI of the advertising program can go up
enormous success. It will also show a much clearer
by several factors, and will rise higher, the closer it is
differentiation between the kinds of optimizations
tied to the advertiser’s strategy and bottom line.
that different programmatic advertising platforms are
capable of.
For example, an advertiser growing its business in
new markets, given the difference in awareness levels
alone, will have little chance of success if it expects
 FOR EXAMPLE:
the same CPA it enjoys in an existing market. But focus MAXIMIZED CASH INFLOWS FOR
the same program on the discounted cash flows of NEW CHECKING ACCOUNTS
future customers, and the new markets suddenly look
far more attractive. After the financial crisis hit, banks were scrambling
for consumer dollars and advertising their checking

DON‘T REDUCE BUSINESS accounts aggressively. A leading online bank found that
its programs persuaded many consumers to open bank
GOALS TO LOWEST COMMON accounts, but that those consumers did not necessarily
DENOMINATORS fund those accounts with much cash. We worked with
the agency to shift the focus from accounts opened to
Being fully aware of the business goal for the client,
aggregate dollars flowing in.
but “translating“ that into a pure marketing goal for the
programmatic advertising vendors is a common error.
This optimized program generated as much as a 20% lift
in bank accounts that are profitable for the bank, thus
This is partially based on habit, and partially on
helping the bank to drive much higher ROAS.
insecurity about programmatic advertising vendors not
being able to measure something outside of the core
advertising metric toolbox.

Artifical Intelligence. Real Results.

10

The best DR marketers allow the programmatic

models to discover these new data points and

FOUR: generate conversions even in unexpected corners of

RELAX YOUR
the market, rather than constraining advertising to
traditional segments.

CONSTRAINTS
IMPROVES LEARNING AND
REDUCES WASTE
Traditional marketing tends to avoid waste by limiting
Programmatic buying works best
spending to closely prescribed target segments, then
when it can make full use of all selecting media that claim an acceptable degree of
reach into that segment.
the consumer data available, then
Advertising programs, however, buy every single
amplify successful conversions to impression individually and have millions of data points
new audiences that exhibit many to predict whether an impression will ultimately lead to
a conversion. They minimize waste by the very act of
of the same data points. optimizing toward the program goal, and no longer need
the traditional trinity of segments, content, and reach.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 11

DON‘T FORCE A TRADITIONAL WAY reach more Spanish-speakers into “only advertise on
Spanish-language websites.“ The resulting acquisition
OF SEGMENTATION costs were typically 2-3x higher than the goal
Limiting programmatic DR programs to only impressions originally intended.
within a certain segment or on certain publications will
often be counterproductive because this will limit both: A major power company started out with a troubled
advertising program too, defining micro budgets for
• Conversions that happen to fall just outside of each DMA. Once we removed this restriction and
the segment focused on the same geography at a state level, the CPA
dropped by more than half.
• Learnings for more advanced modeling of the
consumers most likely to respond to the program

We have seen many examples of programmatic programs

BEST PRACTICE
with “hard constraints“ on segments or site lists under-
performing because the algorithms could simply not learn
enough about this limited set of consumers to discover
the program‘s additional market potential. FOR BRAND
DO SPECIFY DESIRED SEGMENTS MARKETERS
AS STARTING POINTS Trying to discover new segments does
The most effective way to use audience segments in not fit with brand programs. Brand
a programmatic DR program is to specify them as “soft marketers have audience composition
constraints“: and reach as core performance
measurements, so the approach must
1. Set a clear goal and a numerical constraint for the be different for those programs.
program performance
The good news is that programmatic
2. Limit hard constraints to geographical
advertising programs are excellent
constraints only
at driving a high reach within a very
3. Specify all demographic, psychographic, and other specific audience composition. The key
targeting criteria as soft constraints to success is to optimize the program
4. Let the campaign perform and optimize within the for that measure and not for an
constraints and build on the starting points given inaccurate proxy measure such as clicks.

 FOR EXAMPLE: Advanced programmatic advertising

HALVED ACQUISITION COST FOR A companies have permanent
partnerships with brand-measurement
POWER COMPANY
companies like Nielsen and comScore to
Relaxing audience characteristics can double or triple integrate the programmatic advertising
the impact of an advertising program. program with tried-and-true brand
metrics, and even to optimize toward
A major automotive marketer translated its intent to
these brand metrics in-flight.

Artifical Intelligence. Real Results.

12

FIVE:

TEST DATA MODELS

HEAD TO HEAD
Marketers gaining experience with programmatic buying start to
notice how their advertising becomes less dependent on the sites
and apps where ads are served, and more on the right consumer data
underpinning the selection of impressions. This is why it makes sense
to add more data.

On the other hand, buying third-party data models and and then try to find more consumers that “look” like the
investing in the quality of first-party data segments is consumers that ultimately converted in these segments.
expensive. A simplistic programmatic buying program
But because nobody can reliably predict the validity of
may shift a large chunk of the efficiencies gained
data, this approach is generally inefficient. It takes the
in media buying toward data buying, unless it tests
validity of the third-party data models for granted, and
aggressively which data actually works.
pays for their use no matter what their influence on the
marketing results is.
The best approach for marketers is to handle data the same
way they test vendors: Try out as much as possible, but
never commit large budgets until performance is proven.
DO INVEST IN DATA THAT IS
PROVING ITS WORTH
DON‘T FOCUS ON SPECIFIC DATA The best practice is to suggest all of the possible data
MODELS DIRECTLY sources to the programmatic advertising vendor, but to
consider them just a starting point.
Simplistic programmatic advertising programs buy
multiple third-party data models, maximize reach and Advanced programmatic advertising algorithms will run
frequency among the consumers in these segments, massive tests of many combinations of data sources,

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 13

messages, and context, and will quickly figure out which on showed large differences in brand lift. At the
combinations of data points are most successful and recommendation of the agency, it decided to test all
rise to the top. data models against an algorithmic-synthesized model
of their intended audience.
Many other data sources will remain below the surface
and not show significant lift for a given program While two out of eight third-party data models
objective even though they may have performed well for performed better than the synthesized model, six
a different program or client, or may do so in the future. out of eight did not, dragging overall brand lift
down 45%. Naturally, the client decided to move
 FOR EXAMPLE: the budget into the performing data models and the
MODELED AUDIENCES synthesized model.

SEEDED WITH DATA MODELS

A major automotive manufacturer saw an equally
OUTPERFORMED BY 2.5X
impactful difference in a head-to-head test for
A major CPG advertiser noticed that the third-party behavioral targeting: the dynamic model did almost 2.5x
data models that they had based their program better than the data models, with a 59% lower CPA.

Artifical Intelligence. Real Results.

14

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 15

SIX:

TEST
MULTIPLE
CREATIVE DON‘T TRY TO PICK THE SINGLE

VARIATIONS
BEST CREATIVE UPFRONT
One of the worst errors in programmatic advertising

LIVE
occurs when advertisers try to develop a single
“perfect“ ad rather than keeping a few variations on
the table. Programmatic advertising programs do better
when they start with a few flavors of an ad and then
allow the market to decide which one(s) works best with
While programmatic ads need to which consumers and in which context.

look like a small print ad to be DO TEST VARIATIONS AND ADD

effective, the big difference with POLISH LATER
digital is that there is no need The goal during the first week (or maybe even the first
day) is simply to learn as much as you can about which
for “final“ creative. If the goal is creative elements work best. The algorithms will do

to learn and optimize as much as massive testing that yields a massive amount of data
about the best message, the best image, the best
possible during the program, then promotion, and the best call to action.

it is actually counterproductive to
 FOR EXAMPLE:
have “final“ creative at any point. INCREASED PERFORMANCE AND
LEARNING
MAXIMIZE LEARNING AND A well-known financial services niche publication

EFFECTIVENESS started running just one creative. While the client was
happy with the initial performance, tests with additional
Programmatic advertising is most effective when it creatives kept increasing results. As a result, more
can test and revise (or “learn“) across many cycles. This creative variations were added.
learning should not be limited to tactics and settings;
it should also include the messaging. The reality of The client is now at a total of 47 creative variations,
programmatic advertising is that very subtle differences across all creative types, sizes, and funnel stages.
in ads can result in strikingly different response rates. Average performance is now 82% better than when the
Let the data from the response rates be your guide. program first started.

Artifical Intelligence. Real Results.

16

SEVEN:

ENTRUST DAILY
OPTIMIZATION
TO THE ALGORITHMS
Many marketers, especially when working directly for the advertiser
rather than at an agency, are very hands-on when it comes to
managing their budgets. They are used to making constant changes
to the budgets based on the performance of the most recent
few days. While we applaud this hands-on management in
traditional digital media or in search advertising, it can be
counterproductive in programmatic buying, especially in a
fully algorithmic environment.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 17

MANUAL TWEAKING HURTS DO LEARN FROM THE MASSIVE

PERFORMANCE TEST RESULTS
The programmatic system already takes the variability Instead, a best practice is to set a clearly measurable
of the environment into account and adapts to it
program objective and let the algorithms do their job.
constantly. Changing the goals resets the learnings and
can make previous learnings less valid. • Be honest about the real program objective and
avoid “padding“ that number for better performance.
DON‘T OVERIMPOSE OR
OVERCORRECT • Let the algorithms run, and provide the vendors
with honest feedback about performance (see
• The first common error is to correct manually and
“Rank Vendors Openly“ on p. 22).
reset the learnings before the algorithms have had
a chance to correct themselves.
• Avoid making manual changes to content, segments,

• The second common error is to correct for naturally objectives, and budgets until the algorithms have
occurring variability of the kind that an algorithm run at least a week with stable settings.
can recognize and manage, but stressed-out human
marketers might overreact to. • Delve into the insights that the algorithms’ massive
tests provide, and adapt other parts of your program
• The third common error is to overcorrect. Marketers to the new learnings gained.
may set a lower CPA than they can really afford,
just to “have the result reach the middle.“ In such a
 FOR EXAMPLE:
constrained system, the outcome may be that the
partner‘s algorithm is only able to deliver a fraction 45% IMPROVED PERFORMANCE AT
of the volume at the more aggressive goal. 68% HIGHER BUDGET
We worked directly with the marketing team of a well-
known women‘s-apparel retailer. The marketing team
was extremely hands-on, prescribing exact daily and
weekly budgets and swapping out creative every single
day. This threw off the automatic calibration of the
models so much that the client missed its KPIs by as
much as 50%.

A major eyewear e-tailer had a marketing team that

was very hands-on, resetting budgets on a daily basis.
Looking for better performance at a higher scale, it
allowed our algorithms more freedom to move budget
between channels and between days of the month.
This flexibility made such a difference that CPA dropped
45%, even with a 68% budget increase.

Artifical Intelligence. Real Results.

18

EIGHT:

EMBRACE VIEW-THROUGH
CONVERSIONS
Consumers typically see ads numerous times: during a Google search, on
Facebook, as a display ad, in an email, or on an affiliate site—sometimes
all of the above. Therefore, it makes little sense to give full credit to
the last time a person clicked on the ad before converting. Effective
advertising programs draw in customers every step of the way.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 19

STRONGER FUNNEL GROWTH DO CHAMPION STATE-OF-THE-ART

Apportioning credit to multiple touch points before a ATTRIBUTION
person converts is not that difficult, and it pays huge We recommend implementing a robust multi-touch
dividends. Embracing the effectiveness of view-through attribution framework that credits every stage in the
conversions is a key step to capturing the full benefits funnel appropriately, and updates programs with data
of programmatic advertising. in-flight (see “Share Results in Real Time“ on p. 20).

Experienced marketers understand the importance of You can build a robust attribution framework by
“top-of-the-funnel“ efforts designed to plant a seed in a experimenting iteratively with the mix of partners,
customer‘s mind. For most consumer products, only 14% channels, and attribution allocations across a few
of consumers are said to be open to alternatives at the programs. Most likely, you can reach the same bottom-
actual purchase stage. This means that at least 86% of line results much more efficiently with fewer partners
the purchase decision is made before the last click. and fewer touch points.

DON‘T OVERSIMPLIFY YOUR  FOR EXAMPLE:

ATTRIBUTION 99% OF ORDERS RESULTED FROM
Multiple analyses also show that people who click on ads VIEW-THROUGH
represent a tiny, and generally unattractive, demographic
We work with one of the biggest fast-food delivery chains,
segment of Internet users. One program seeking 18-24
driving both in-store sales and increasing the effectiveness
year-old city students of wealthier backgrounds, for
of TV advertising with programmatic support.
example, found that a majority of clickers were retirees in
poor households and rural environments. Initially, both the advertiser and the agency were
convinced that they should give credit only for
Marketers that insist on click attribution will ultimately purchases attributable to a click. However, Nielsen
find their marketing returns collapse. This is simply research proved clearly that there is “virtually no
because they are focusing on the skewed audience that relationship between clicks and brand metrics or offline
clicked on an ad in order to convert. sales.“ ComScore supported this finding, saying that
only 16% of consumers ever click on ads, and that 8% of
users account for 85% of all clicks.

In a program we ran for a major restaurant business,

clickers were also very different from converters.
Converters tended to be 18–44 years old, whereas
clickers skewed toward 65+/retired. Converters tended
to be interested in movies and video games, whereas
clickers preferred golf and ladies’ leisure.

Further research with the agency showed that repeat

customers (those with high lifetime value) were 8.8x
more likely to be view-through converters than click-
through converters. After we showed that click-based
revenue represented less than 1% of total revenue
generated, the advertiser quickly embraced view-
through conversions.

Artifical Intelligence. Real Results.

20

NINE:

SHARE
RESULTS IN
REAL TIME
Some of the low-hanging
fruit in programmatic
advertising is to share
performance data
with vendors on a
programmatic basis. With
just one login to your ad
server or analytics system,
you can create massive
performance improvement
solely on the basis of how much
the programmatic system can learn
from granular results.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 21

The most common performance tracking

SPEEDING UP THE systems are the third-party ad servers. But,
if your “source of truth“ for marketing performance is
LEARNING CYCLE
a web-analytics system like Omniture, then it‘s critical
Programmatic buying is designed to respond to changes to provide that feedback to your partners as often as
in supply and demand in microseconds. To do this well, possible. If the web-analytics data conflicts with the
it needs up-to-date feedback on how well it is doing for third-party data, but it‘s the web-analytics numbers that
the advertiser. Direct programmatic access ensures that matter to senior executives, then that‘s what all partners
the programmatic advertising algorithms can respond to need to be using. Otherwise, their systems are optimizing
these fluctuations instantly. off the wrong data.

SHARING A PDF ONCE A MONTH  FOR EXAMPLE:

IS AN OPPORTUNITY LOST 2X PERFORMANCE OVERNIGHT
This means that it is not sufficient to tell your We should mention that the first benefit of real-time
programmatic buying provider about your results on data sharing is less work for the agency or marketing
a weekly or monthly basis. You also shouldn‘t do it team. Once clients move to real-time data integration
in a format that is hard to import. And sticking with with their ad server or analytics system, they notice
manual processes causes frequent errors and delays in that campaigns are start to auto-optimize without any
importing the performance data back into the system, manual intervention. This productivity gain alone is a
which results in inconsistent calibration and reporting. massive benefit for busy marketers.
Allow your vendor to pull data from your system
This optimization, of course, also provides a
By giving a programmatic vendor real-time access to performance boost. After a premium automotive brand
the same data that you use to track their advertising decided to provide us with a login to its ad server, the
results, the vendor can optimize their program in real automatic optimizations kicked in and performance
time rather than weekly or monthly. doubled overnight.

Artifical Intelligence. Real Results.

22

TEN:

RANK
VENDORS
OPENLY
A second aspect of sharing
performance is easy to accomplish
and highly effective too: Simply tell
each partner how they perform
against their competitors.

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 23

ENABLES CALIBRATION OF
PERFORMANCE METRICS
The very nature of programmatic advertising is that it
has advertising programs running on automatic pilot.
The software to run these programs works best when it
can regularly calibrate itself to the performance that the
client sees.

DON‘T CANCEL A VENDOR

IMMEDIATELY
This is why it is better to share relative performance
openly before taking a vendor off the plan. Whereas  FOR EXAMPLE:
performance with traditional media vendors is typically
ALWAYS SHARE INFORMATION TO
fixed for any given program, programmatic buyers have
at least a few levers they can pull to push performance.
HELP ENSURE HIGHEST ROI
Cancel them too early, and you may miss out on a part of In one particular case, our algorithms were beating the
the program that could have performed better. internal client goal, but the client did not inform us and
we could not see the same performance in our numbers.
Smart programmatic vendors strive to be the optimal long-
We therefore recommended that they cancel with us
term performer for the advertiser. Their goal is to drive
in order to work with another partner (not knowing we
performance at a level that is better than competitors’,
were the best performer on the plan). If our system
but to also be scalable to a high volume at any given point
had been re-calibrated on the basis of the performance
during the program and over time. Calibration can help
that the client saw, we could have scaled the campaign
facilitate a switch to a shorter-term perspective, when that
significantly, delivering a much higher return on
is what the advertiser really wants.
advertising spending than the client ultimately
obtained. It was an opportunity lost for both of us.
DO TELL EACH VENDOR THEIR
RANK WEEKLY On the flipside, a leading online music-distribution
service initially declined to provide relative performance
We recommend sharing relative performance with all
feedback. Since our program beat the official CPA target,
partners on (at least) a weekly basis.
the team did not see the need to tweak performance
versus scale and sustainability.
1. Rank all partners on the plan by the one metric
most important to the advertiser.
After the client informed us that performance was more
important than volume and that another vendor was
2. Tell each partner what place they reached in the
driving better CPA results, albeit at a lower scale, some
ranking, and indicate if the gaps between them
and their rivals are large or small. tweaks quickly drove a 15-20% performance improvement.
More importantly, this discussion also generated more
3. Ask the best-performing vendors how much insight into the client‘s real business goal: driving better
more budget they could take at similar levels of prospects with a higher potential lifetime value. After we
performance. The best programmatic vendors can learned that, we quickly optimized the program, to much
predict this with a high degree of accuracy. higher satisfaction of the client.

Artifical Intelligence. Real Results.

24

ELEVEN:

SELECT A PROGRAMMATIC
PARTNER THAT EMPOWERS
ALWAYS-ON MARKETING

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 25

We hope that this whitepaper has shown many actionable ways to make
your programmatic advertising run better.

While some of these tips can be applied liberally in

a self-service setting, the reality is that the vast
majority of programmatic advertising runs through a
full-service relationship with one or more specialized
vendors. Therefore, our final recommendation is to
be critical about vendors and to pick the ones that
ultimately deliver the best ROAS (return on ad spend)
to the advertiser.

Obviously, advertisers may want to take into account

buying criteria like transparency, data access, low
margin, and a beautiful user interface. But, in the end,
dozen or more seemingly comparable partners “duke
we find that the best practice is to focus on attaining a
it out“ without utilizing some recommended best
sustainable ROAS first and foremost. If you do that, the
practices. Instead, choose one or two partners that
rest will then fall into place.
have long and proven track records, and the expertise,
the technology, and the roadmap to build a long-term
TWO VENDOR “MUST-HAVES”:
strategic relationship with your company.
1. SPECIALISTS WITH PROVEN TRACK RECORDS
Choosing the right partner is important because Think of the vendor-selection process almost as a
programmatic advertising is changing much faster, technology decision—one that is more far-ranging than
and thus is much less user-friendly to operate running a media plan for one month. With that in mind,
than, say, search advertising. it’s clear that it should take plenty of time to investigate
various vendors for the best long-term fit. The best
2. DON‘T HAVE MULTIPLE VENDORS ON THE PLAN programmatic marketers think about adapting their
“DUKE IT OUT“ strategies for an always-on marketing world. Rather
Choose a number of partners for your plan, and than thinking about campaigns with fixed starts and
give each every opportunity to show you their ends, they think about marketing as a 24/7 practice, and
best stuff—letting them know how they are doing advertising that learns and improves doesn’t need to
against (anonymized) competitors. stop or sleep to improve ROAS.

Since programmatic depends so much on learning and That said, you should also run a head-to-head test at
strategic data integration, the worst way to choose scale to ensure that each partner produces outstanding
programmatic advertising partners is to have a half results time and time again.

Artifical Intelligence. Real Results.

26

11 Best Practices For Programmatic Advertising More questions? Contact us at 888-717-8873 or [email protected]
 27

The best way to run traditional

advertising is often not the
best way to run programmatic
advertising. This whitepaper has
outlined eleven best practices
to get the most out of your
programmatic marketing efforts,
and we hope you will be able
to apply most if not all of these
practices across all of your
programmatic advertising programs.

Choosing the right partner is

CONCLUSION crucial. As the most successful

programmatic advertising platform,
Rocket Fuel is laser-focused on
delivering optimal advertising
results, day after day. Our
“always-on marketing“ approach to
programmatic means your program
results are tested, reviewed, and
optimized all day and all night.
With this strategy, we can take
your advertising results and
marketing ROAS to new heights.

Artifical Intelligence. Real Results.

ADVERTISING
THAT LEARNS
®