CurrencyFair

How we increased performance by 500% for a fintech company

"Delivered excellent results!" - Simon Walsh, Marketing Manager

CurrencyFair.com is a currency exchange company offering cheap foreign currency exchange through a peer-to-peer model. Within three months of working with Segmatic, its paid search (Non Brand) performance increased almost fivefold.

Segmatic achieved this increase despite the fact that rivals were spending huge budgets on above the line spend; traditional banks were pushing back, and the product, while significantly cheaper, required a significant amount of information on registration. So, despite these issues, how did CurrencyFair.com increase its paid search performance so dramatically?

In this article, we will show you how we did it.

JavaScript chart by amCharts 3.19.6 1 2 3 4 5 6 7 8 9 10 11 20 30 40 50 60 70 80 90 100 0 50 100 150 200 250 300 CPA TARGET JS chart by amCharts

Controlled advertising builds a virtuous circle

Account structure is the bedrock of a successful paid search campaign

Paid search advertising is not just about ad copy and bidding strategy. At its most fundamental level, online advertising is about controlling who sees your ad, what ad they see and when exactly they see it.

Control is at the core of Segmatic’s approach to paid search advertising. Control allows us to understand what our ads are doing, develop actionable insight, and use these insights to improve performance. Our virtuous circle gives clients confidence that their marketing budgets are providing Return On Investment (ROI) and encouraging growth and re-investment.

When Segmatic took over the Currency Fair account, the client was unable to exercise an appropriate level of control over their advertising. This lack of control manifested itself in two primary ways; confusion of campaign segmentation and misunderstanding of match-types.

The importance of effectively segmented campaigns, and how to build them

The real power of paid search advertising is that there is a large amount of data available to advertisers which allows for effective testing and optimisation. The value of this data, however, can be lost if companies begin to view their customers as simply numbers on a page – forgetting that behind each search is a human with a different level of purchasing intent or brand loyalty. These potential customers need to be treated differently to maximise the possibility that they will complete an online purchase. This is why campaign segmentation is a vital part of any account build.

Any search advertising professional should tell you that there are two basic types of campaigns: branded and non branded. Non branded campaigns can be further segmented by distinguishing between competitor and product terms.

As you can see form Table 1, these three campaign segments have radically distinct characteristics – a potential customer who searches for the term “Send Money with Currency Fair” needs to be treated very differently to one who searches for the term “Send Money Internationally” and differently again to a potential customer who searches for “Send Money with Bank ABC”.

Segment
Branded
NB - Product
NB Competitors
Example
Currency Fair
Bank Transfer
Bank ABC Currency Transfer
Description
The search terms related to the Brand
Search terms related to the products
Search terms related to Competitors
Click Through Rate
High
Medium
Extremely Low
Conversion Rate
High
Medium
Low

Understanding match types saves money and builds value

While Google offers four different match types, well-run campaigns should only use Exact Match and Broad Match Modified.

Google offers four different match types with an example of the type of search term that they would appear for.

Match Type
Broad
Broad Match Modified
Phrase
Exact
Keyword
Womens Hats
+Womens +Hats
“Women’s Hats”
[Women’s Hats]
Search Term
Buy Ladies Hats
Hats for Women
Buy Women’s Hats
Women’s Hats

That’s why separating your brand and non-brand programmes is an essential first step in analysing and revitalising your Paid Search programme. This is a reflection of a simple truth - brand Paid Search is easy. Sure you can mess it up but the skill in PPC is getting non-brand to work for you. That is why Segmatic have completely different billing models for brand and non-brand search.

Key Lessons:

  1. You can’t trust Google to identify relevant synonyms by using Broad Match. | A keyword like Womens Hats will result in ads showing search terms as varied as baby clothes, scouts caps, top hats, etc. Broad Match Modified (BMM)This is what Google claims Broad Match is intended to be, but it covers significantly more keywords that you realise.
  2. Phrase and Broad Match Modified do very similar things. | They don’t attempt to find similar words. They just search for phrases / words within longer search terms, like “I want to buy some hats for women today”. Phrase is not that useful anymore because BMM allows you to catch people who search for “buy hats for women” and “buy womens hats” within one keyword. Building out Phrase match campaigns for all combinations is extremely cumbersome and provides no benefit.
  3. A combination of Exact and BMM match is the best system to use. | Not only were Currency Fair advertising on Broad Match, but also the account was extremely reliant on this match type.
    • Match Type
      Broad
      Exact
      Phrase
      BMM
      % of Spend
      62%
      30%
      2%
      6%
  4. Billing for Performance Aligns our Incentives with Yours | Most agencies bill based on a percentage of media spend or hourly fees. This means that their incentives are either to increase media spend or to increase the number of hours for which they can bill you. But that’s not what Currency Fair wanted, they wanted sustainable ROI-positive growth.

Segmatic’s approach was different to other agencies. We wanted to align our incentives with Currency Fair, so we developed a billing model based on sustainable ROI-positive growth.

We split the campaigns into two: Branded and Non Branded.

Campaign Type
Branded
Non Branded
Currency Fair Incentives
Drive down CPA
Drive up growth within CPA
Segmatic Billing Model
Decrease in CPA
Incremental acquisition within CPA targets

Branded billing model | We established a baseline CPA. We have no influence over the number of people searching, only on the CPA on those impressions. As such, Segmatic agreed to only bill Curreny Fair for the decrease in CPA that we achieved.

Non-branded billing model | We established a baseline, the average number of customers Currency Fair had gotten per month at an average CPA. We would then only bill them for any additional customers above that baseline. This meant that we had identical objectives to Currency Fair. If their average number of customers did not increase, Segmatic would not bill them.

Building Currency Fair’s account from scratch

Control is paramount.

In order to get the level of control that Segmatic wanted over Currency Fair’s ads we knew that we needed to rebuild their entire account. The focus of the rebuild would be granularity. Campaigns need to be structured in very specific ways to ensure effective keyword management. In order to achieve this goal we use a layered approach to campaign building. How does this work?

Step 1 | Identify geographic areas and languages (English UK and English US)

Step 2 | Separate Match Types (Exact and Broad Match Modified)

Step 3 | Identify key product terms (Currency Transfer, Currency Exchange, etc.)

Step 4 | Identify modifiers that will affect conversion rate (Online, International, etc.)

An example of this approach is shown below:

Tier
Tier 1
Tier 2
Tier 3
Tier 4
Total
Type
Geographic & Location
Match Type
Product Term
Modifiers
Example
English UK, English US, Spanish US
Exact, Broad Match Modified
Currency Transfer/Exchange
Online, Web, International
Number
3
2
2
3
36

In this example there would be a total of 36 separate campaigns to be created – though naturally a real campaign build would require many more than this. The first two campaigns in this example would be:

[UK & EN | Exact | Currency Transfer | Online], [US & EN | Exact | Currency Transfer | Web]

[US & EN | Exact | Currency Transfer | International] would be the first generated.

While this approach is essential to allow for full control over campaigns – and visibility into their results – it poses obvious challenges. Each additional location, core term or modifier that is added increases the number of campaigns that need to be created, and reported on, exponentially. However, this type of attention to detail is essential for maintaining control. Currency Fair has over 9,800 different online campaigns running simultaneously.

In a world of big and small data, effective reporting requires advanced analysis techniques

This approach to campaign building also allows for full visibility into the performance of Segmatic’s advertising. In the world of big and small data, however, effective reporting requires advanced analysis techniques. Our layered approach to campaign building allows us to perform two necessary types of analysis: vertical & horizontal:

Vertical analysis allows you to identify the overall performance of an individual campaign relative to another one. In the example below, you can see the various performance of campaigns, noting that Campaign 3 is performing particularly poorly.

Targeting
"Geographic & Language"
Match Type
Product Term
Modifier
Campaign 1
EN & UK
Exact
Product Term 1
Modifier 1
Campaign 2
EN & UK
Exact
Product Term 1
Modifier 2
Campaign 3
EN & UK
Exact
Product Term 1
Modifier 3
Campaign 4
EN & UK
Exact
Product Term 2
Modifier 1
Campaign 5
EN & UK
Exact
Product Term 2
Modifier 2
Campaign 6
EN & UK
Exact
Product Term 2
Modifier 3
Campaign
Campaign 1
Campaign 2
Campaign 3
Campaign 4
Campaign 5
Campaign 6
Clicks
6204
8624
506
45600
21000
2603
Cost
£3,102
£4,312
£1,214
£9,120
£3,761
£3,124
Conv. Rate
0.03
0.012
0.04
0.01
0.01
0.005
Registrations
186
103
4
456
197
17
CPA
£17
£42
£304
£20
£19
£184
vs. Target
42%
104%
759%
50%
48%
459%

The real insight into marketing performance, however, is provided by layering a horizontal analysis of the layered campaigns on top of this vertical analysis.

Horizontal analysis provides greater insight than a simple vertical analysis ever could. By grouping common characteristics of the various campaigns together, we can achieve statistical significance in order to generate insights that would otherwise be delayed, all while running campaigns that target niche groups of potential customers.

Product
Product Term 1
Product Term 1
Clicks
15,334
69,203
Cost
£8,628
£16,005
Conv. Rate
1.90%
1.00%
Registrations
294
670
CPA
£29
£24
vs. Target
73%
60%
Modifier
Modifier 1
Modifier 2
Modifier 3
Clicks
51,804
29,624
3,109
Cost
£12,222
£8,073
£4,338
Conv. Rate
3.00%
1.20%
1.20%
Registrations
642
300
21
CPA
642
300
21
vs. Target
48%
67%
516%

Horizontal analysis shows that the modifier “international” performs significantly worse than other modifiers, allowing us to make improvements to both campaign 3 and 6. We have also identified that Currency Exchange outperforms Currency Transfer in both volume and CPA, so there is an opportunity to push the bids up. This opportunity would have been hidden if you looked at each campaign individually.

Quality Score is the Most Important Factor in How Much Your Advertising Costs, and it is Fundamentally Based on Account Structure

Google rewards good advertising through a system known as Quality Score. At its most simple. Quality Score is designed to reward advertisers for the relevance of their ad to the search term entered by the individual user. From Google’s point of view, this improves user experience - members of the general public are more likely to continue using the platform if their searches are more effective. From an advertiser’s point of view, Quality Score provides a great opportunity to gain a competitive advantage simply by implementing good practice.

Quality Score management is about the click-through-rate (CTR) of your ads. The better your click-through-rate, the cheaper your ads will be. If your account has a strong CTR, everything will be cheaper, so every decision you make will affect your overall cost per click.

alt
Impact of Quality Score on CPC
alt
Impact CPC of 1-Pt QS Change

The idea behind Quality Score management is a simple one – deliver ads relevant to the customer. Once again however, the devil is in the detail. If two potential customers search for “online currency exchange” and “international currency exchange” they should get different ad copy with online and international in the headlines. Being able to replicate this approach for many keywords across several thousand campaigns requires real skill – this is where Segmatic comes in.

Our layered approach to account structure ensures we can deliver the most relevant ad to each audience, thereby improving CTRs and driving down advertising costs. You can see how effective this approach was for Currency Fair – with Quality Score performance doubling after Segmatic took over the account.

Paid search gives the advertiser a lot of data. What you do with that data is what counts.

One of the most common mistakes that people make when working on data driven marketing is to use numbers to justify their decisions, even if the numbers don’t support the decision. Let’s explain with an example:

If you tossed a coin 10 times, and you got 7 heads and 3 tails, which side would you pick on the 11th toss? As almost everyone knows, both sides have an equal chance of occurring, so you would simply pick at random.

If you don’t achieve statistical significance then your approach is faux quantitative

One of the most common mistakes that people make when working on data driven marketing is to use numbers to justify their decisions, even if the numbers don’t support the decision. Let’s explain with an example:

If you tossed a coin 10 times, and you got 7 heads and 3 tails, which side would you pick on the 11th toss? As almost everyone knows, both sides have an equal chance of occurring, so you would simply pick at random.

If you set up two new campaigns, with the same bids, and over the first few days, one has 7 conversions for £100 spend and one has 3 conversion for £100 spend, which campaign do you increase the bids on? Just as in the above example, we still do not have enough data to make any decisions, so any changes are being decided at random.

How do you determine when there is enough data to make a decision? Using statistical methods, you can determine what is known as “statistical confidence”. This allows you to determine when there is enough of a difference between campaigns to make decisions.

Lots of advertisers will simply change their strategy every day, following whatever the most recent “data” tells them. While this approach might get lucky and increase performance for a small number of campaigns, it is impossible to apply the ‘lessons’ of a faux quantitative test to any account that has ambitions to scale.

Ad copy testing

Ad copy testing is the most common example where people are ‘”faux quantitative”. Testing will be run on lots of individual campaigns, with the “results” showing that X ad or Y ad is better. Most ad copy tests if run through statistical significance calculators do not reach enough data. That’s why one ad that works great for one Keyword seems to work so badly in another.

The Segmatic approach allows us to perform large scale ad copy tests across all of the campaigns. The tests can answer key questions like “Should you use your brand term in Non Brand ads”, “What is the best way to incorporate the relevant product and modifier into your ad?” and “What is the best promotional offer to run?” Above all, these grouped tests allow us to deliver results significantly quicker.

Bid optimisation

Bidding often falls into the same trap of faux quantitative. Each individual Keywords is analysed and then the appropriate bid is determined. However, it is likely that only 5-10% of your Keywords will ever have enough data to determine their own conversion rate. Most of the decisions being made by automated bidding platforms are based on this flawed understanding of statistics.

Leveraging our vertical and horizontal analysis, Segmatic is able to work out the different conversion rate uplifts for each product term and modifier. These percentages are then applied across all campaigns, resulting in thousands of different bids, all based on strong statistical evidence.

New campaign creation

In Paid Search advertising, there is a distinction between Search Queries and Keywords. Keywords are used to trigger ads on search terms. The keyword “+Currency +Transfer,” for example, will trigger ads on the search query “I want to make a currency transfer” and “I want to make a currency transfer online immediately”. Search Engines offer you the ability to see the search queries that your keywords are generating.

Segmatic performs continuous search query reports on our layered campaign structure to identify new product terms or modifiers that can be added to the account. This approach allows us to develop out long-tail keywords, which are significantly cheaper as they are less competitive. Combined with our layered approach, which allows for highly relevant ads to each search term, this gives us a further advantage on long-tail. The extent of this shift can be seen in the graph below.

Percentage of conversions from top ten keywords

The table below shows the process by which this shift towards long-tail took place for Currency Fair. Each search query report gave us additional insight into the campaigns, allowing us to steadily improve the structure, making it more refined and targeted.

Build Version
Version 1
Version 2
Version 3
Version 4
Time
-
30 Days
B70 Days
180 Days
Number of Keywords
200
1200
3000
8000

What this all meant for Currency Fair

Segmatic were able to decrease CPA for Currency Fair – while dramatically increasing the number of conversions happening every month.

If you would like to find out how Segmatic could help your company to do the same, get in touch!

Many thanks to Currency Fair for allowing us to use their account as a case study.