B2B attribution: Build vs Buy?
Right, you’re on the verge of deciding how you’re going to solve your attribution needs but don’t know if to go-it-alone or rely on an off-the-shelf solution.
At Dreamdata we’ve encountered a good few customers facing this same dilemma.
And whereas sometimes leading the charge as a lone-wolf has worked well, other times - I dare say most times - a team effort with a leading vendor has offered the best strategy.
But how do you know which is the best outcome for you?!
The way we’ve helped resolve the dilemma for our customers has been by unpacking what is actually needed to arrive at a powerful (and sustainable) B2B revenue attribution solution.
Revenue attribution? Yes. As a B2B company you need to be looking at revenue attribution to make sense of your complete customer journey and all your revenue activities. When you are looking at attribution should you look to build or buy? What is the best revenue attribution model?
A B2B revenue attribution solution does this in three important ways:
Offers B2B marketers unprecedented insights into every touch of every customer journey.
Reveals the revenue-generating performance of activities across the pipeline, channels and campaigns.
Deep dives into every aspect of every channel, campaign and experiment to help rapidly drive up ROI.
So, in this post we’ll be covering the process of building a B2B revenue attribution solution in order to help you down the road of building or buying.
Contents:
*No time to read? Jump to the Evaluation (TL;DR) at end of the post👇*
On assumptions in build vs buy scenarios
Our experience has taught us that there are a number of assumptions B2B companies make when considering whether to build or buy their revenue attribution solution.
You have likely given most thought to the business-value aspect of attribution, such as what you expect to be attributing and with what model.
In fact, your motivation behind even considering an in-house solution might be a combination (or all) of these:
You don’t need some of the features an off-the-shelf platform offers;
You have the resources and specialist know-how to build a solution in-house;
An off-the-shelf platform doesn’t give you the bespoke tailored solution you want.
Our experience has proved that these assumptions typically underestimate the process of setting up revenue attribution, miscalculate what features you actually need, and misjudge what an off-the-shelf solution actually provides. Possibly ending up as a nightmare data-project draining resources without ever really being completed (More on this in the Evaluation)
Companies tend to scope their projects from a top-down perspective, which results in missing vital components of the attribution process - and most importantly, the benefits they derive from it!
As the following sections demonstrate, a successful revenue attribution solution is built bottom-up, and depends above all on:
Digitalising as many company-wide revenue activities as possible and storing them in a database;
How you clean and merge the data you’re collecting;
How much automation is built into the process;
The intelligibility of the reporting and analytics;
Let’s see what it takes to achieve this…
Collect and store all your revenue-related data
The first step is to collect all the data you need to get the most out of your B2B revenue attribution.
The collection and storing process needs to be guided towards the B2B customer journey. This means it has to accommodate the (account based) multi-touch nature of the journey from end-to-end.
Determine what data you want to collect
“End-to-end” is actually a great place to start. You firstly need to decide on what data you will be collecting. That is, decide on the cut-off point.
In theory, you can cast the net as narrowly or as widely as you want. For instance, you can look to start collecting data from the moment a user converts as an MQL or SQL, or from the very first moment an anonymous user comes onto your site.
In practice, as a B2B company, you want to be casting your net as widely as possible so you can really get a full picture of the buyer journey. That is, to have a complete holistic view of your customer journey, which includes acquisition through ads, etc. you want to be tracking from as early as possible on that journey.
I mean, how can you decide on what to scale (and drive growth) without this?! You need the complete picture of what’s impacting your customer journeys.
From our own customer data, Dreamdata has found that customers with more than one session prior to converting were 5x more likely to be closed-deal-as-won customers. Meaning that tracking before conversion matters, a lot.
In fact, there are two reasons for this:
Marketing. You will be able to accurately see what campaigns are driving the conversions, as well as have a much more accurate reading of your ROAS and ROI. It also means that marketing departments will be able to get credit for their activities.
Sales. Your ability to accurately qualify leads will increase significantly. After all, you probably want to dedicate more time with those that have spent more time exploring your product showing interest.
So, how do you go about casting this wide net and collecting all the relevant data?
Well, you need to introduce smart on-site behavioural tracking and integrate all relevant commercial data from CRM, Ads, Automations, and Customer Success tools.
On-site tracking and identification
For this you need to set up some form of tracking and identification process, for which you can rely on an off-the-shelf CDP solution like Segment, or your own tracking javascript.
Either way, the objective is twofold: (1) gather as much user behaviour as possible. (2) link this behaviour to an identified user.
To achieve this, your behavioural data tracking needs to start from the moment a user comes onto your site (and accepts ‘statistics cookies’) - whether through paid ads, direct, or organically.
Of course, making sure to have UTMs on all links and ads to your site, as this will allow you to complete the tracking and learn more about your users’ journey. More on this below.
Your tracking should then record users’ interactions with your site, e.g. reading a blog post or subscribing to a newsletter, as well as any properties that describe that action. From which you are able to assign ‘traits’, such as first name and email, to these users’ actions.
The stored data will eventually be converted from an anonymous user to an identified user - complete with all relevant details (email, company, etc.).Integrations
The rest of your data comes from the tools you’re already working with, e.g. CRM, Ads, Automations, and Customer Success tools, as these already hold critical details about customers across your pipeline.
The integrations in themselves are not technically challenging, and so easily set up.
What does present a challenge is how these tools are defining their data. That is, how they’re labelling their fields, etc. But this relates to the cleaning process outlined in the cleaning and merging data section below.Storing the data to help make a build vs buy decision
You want to be gathering the data in a cloud data warehouse such as Google Bigquery, Snowflake, AWS Redshift, and Microsoft Azure. You can find the benefits of cloud data warehousing here.
This can be done relatively simply by syncing data to the data warehouse using standard connectors, like Segment, Stitchdata.
Alternatively, you need to set this up yourself.
How does an off-the-shelf platform do this?
A B2B Revenue Attribution platform like Dreamdata, will help you collect data from all your sources. For site tracking, Dreamdata can either integrate with a CDP tool you’re working with, or introduce our own class-leading tracking. See more here. Dreamdata also integrates with virtually every commercial tool out there. See our current integrations here.
Cleaning and Merging data
This is where the success - or failure - of a revenue attribution solution lies.
How you treat all the valuable data coming into your model will determine what attribution and analytics you’ll run (and how accurately!). Which means, if this step isn’t done correctly all the subsequent exercises will be, to put it mildly, pointless. After all, it is difficult to become truly data driven without accurate and understandable data.
Determine how you want to analyse your data
As with data collection, here you also need to reflect on what the scope of your future analyses is going to be.
This cannot be emphasised enough, not accounting for data here will mean having to re-do your algorithms in the future - and that means more than the cost of spending resources on it, it means potentially losing data whilst you’re redoing this (more on this in the maintenance section below).
Let’s look at a quick example.
You’re thinking about how you want to merge data coming in from your ads platforms - which give you clicks and impressions and the cost of these - with user data from your site, and revenue details from your CRM.
Are you going to be merging ‘high level data’, such as channel, only? Or are you going to want to be taking into account campaigns too? Are you considering looking at cross-channel ad groups?
The only way to answer this is which of these will best help resolve your business problems?
If you’re looking to fire your growth through scaling your best performing ads, you’re necessarily going to go granular.
But of course, the more granular, the more complex the setup.
Cleaning and Merging
Once you’re clear on what the scope of your data collection is, the next step is to structure data in a uniform way. In particular you want the data coming from all your tools and site to have the same fields matched and merged under one field with a single label.
For example, your CRM might use “customers”, your ads platform “users”, and your success tool “clients”. These need to be matched and brought under one label, e.g. “users”.
As so for every single item of data.
Here’s a list of factors to consider:
UTM mapping
After utm parameters values are extracted (e.g. utm_source, utm_campaign, utm_content) from the url, you need to map certain combinations of utm parameter values to a customized source and channel name. For example: map (utm_source=hs_automation) to (Source: Hubspot, Channel: Email).
Deduplication
The data coming in from your tools and the site will often overlap with each other. This needs to be sorted and streamlined.
Removing unwanted data
The algorithm needs to also take into account any null values or not set values within fields, and exclude irrelevant tracking like the activities from your own employees.
Merging to account
A critical component in the data set for B2B revenue attribution is the company (account). As this will be the information that places the various users from each company into one single customer account.
The most important element here is to find the company from amongst the data. Typically you want to set the following hierarchy:
Set a default source for this, e.g. Salesforce (CRM);
Add an alternative tool if it is missing, e.g. when it’s a lead;
If alternative tools have not got that information, your merging algorithm could pick up the email account and assess whether that is a company email address.
Or if not, reverse look-up IP address.
How does an off-the-shelf platform do this?
With a B2B Revenue Attribution platform like Dreamdata, all your data will be cleaned and merged. This is irrespective of whether you end up using that data or not. Which means, should you decide to start applying a different attribution model in the future, you will immediately be able to run these without re-doing algorithms.
Attribution modelling for build vs buy
Once you have all your data cleaned and sorted it’s over to the attribution algorithms to make the magic happen and fuel your growth.
There are two main elements here:
The configuration of your attribution - what your algorithms are going to be doing.
Doing the attribution and analytics - how you’re going to be modelling your attribution presenting the outcome of the models.
Configuring your attribution
When configuring your algorithms you’re mostly going to be defining parameters for your different activities and metrics.
Once again here, it’s necessary to reflect on how these parameters can be best placed to meet your particular needs. For example, how you define a conversion needs to match your unique sales process.
If you don’t know what the scope of here is, you run the risk of spending time on algorithms you might not need at present.
Although, ideally, you do want to ensure you have the maximum possible control over your activities; from marketing campaigns to sales meetings, which you can select and deselect depending on your needs.
How does an off-the-shelf platform do this?
A B2B revenue attribution platform will typically have all these algorithms defining and setting parameters already in place. With Dreamdata you’re also able to apply a high degree of customisation on what you want to be working with and when.
Here are just some of the parameters you’ll be looking to define:
Conversions
You need to set your conversion goals, i.e. what is a conversion, and crucially, be able to identify these from your data.
Sessions & Events
You’ll need to set up the parameters for your sessions and events on all your different data sources. Most companies rely on Google Analytics to define their sessions for web traffic (which carries some risks in its own right, more here) but when you have your own tracking and combine more sources, you’re going to have to define these yourself.
You’ll also want to consider whether some sessions or events should not be excluded, such as we might track sent emails, but we do not want to attribute value to a sent email, but rather to emails that was clicked.
Attribution Goal/Benchmark
You need to define what you’re attributing to? For example, your MQLs, SQLs, or New Business.
Ideally, you want to be able to switch between these models in order to compare what is working against each goal. For instance, MQLs and not for New Bizz, etc. This will help see what occurred at different points in these deals with different marketing/sales models.
Currency
If you’re working on campaigns or channels with different currencies you’re also going to have to solve for which currency you want to be reporting in.
Doing your Attribution and Analytics
Now we get to the final piece of the puzzle: actually doing your attribution and analytics.
Here you need to decide on:
How you’re going to be separating and breaking down your channels and activities,
What metrics you want to be measuring;
What attribution models you plan to be using;
Whether you want to visualise your customer journey;
How you’re going to be presenting these reports.
Note that this business intelligence side of things will require a different set of skills from your data team, and will need input from any marketing analysts or marketing ops folks.
Analytics
You want to be able to run reports on all aspects of your activities across the entire pipeline (Paid, Organic, Sales, etc). Such as, finding your performance, identifying source/medium, segment the data into cohorts, ROAS, ROI, LTV, etc.
What you choose to run should be aligned with your business objectives at this stage. If you’ve addressed the preceding (data cleaning and merging) steps properly, you can always model new analytics to handle any new reporting, without having to touch the data, in the future.
Attribution
As a B2B company, the basis of your attribution modelling is account based multi-touch attribution.
That is, the modelling should accommodate not only individual multi-touch but also stakeholder multi-touch to create the full account-based multi-touch attribution, so that the complete B2B customer journey is mapped and attributable.
Remember the ‘merging to account’ section above? Well, this is what it was for. If you’ve done this step correctly, you’ll easily be able to pull and model the data.
In terms of specific attribution models, you need to have an idea of which model(s) you want to be operating. Standard models to consider are Linear, First-touch and Last-touch. Alternatively you can go for fully algorithm-customised models from the get-go.
With that said, ideally, you want to be able to apply more than one model to the pipeline, as this will allow you to compare how certain activities are performing at different stages.
For example, First-touch will reveal how your paid campaigns are performing in generating leads which convert to closed-as-won deals. Linear, will reveal how activities across the pipeline have performed.
A more advanced setup is being able to break down your pipeline and applying different attribution models to each.
For instance, you might want to use first-touch for marketing activities to understand how effective your campaigns have been in acquiring a customer. And have last-touch, for your sales activities, to see how effective your deal-closing sales meet has been - although here you want to exclude some of the final sessions, such as a sign-off on documents, etc.
You could also go for a completely custom solution, where you build an attribution model that best suits the customer journey specific to your buying process and analyzes your existing customer data.
This however, is probably the most difficult and time-consuming model to build, maintain, and use. It also has inherent bias towards what you attribute credit to, even though those activities/events/channels may not actually be the best for your business.
Get the complete lowdown on all revenue attribution models here.
Presenting your reporting
You also have to give thought to how your end user (demand gen, marketing ops, analyst, content manager) is going to interact with the reporting.
In other words, how are you to move from a table in a data warehouse to providing value to the end-user in your company.
This will require assembling dashboards, which once again requires some experience and expertise in creating.
How does an off-the-shelf platform do this?
With off-the-shelf you get all the analytics and attribution modelling and reporting from the get-go. This includes a wide variety of attribution models, from First-touch, to Linear, to W-shaped and U-Shaped. With leading platfroms like Dreamdata you can also have unique attribution customisation. As a standard with Dreamdata, you are also able to attribute between and across each part of the pipeline.
More on Dreamdata’s Data Platform
Maintenance
So, you’ve done it. Now you’ve set up your attribution modelling and are ready to crack on with scaling your growth!
Well, unfortunately, that’s not quite the end of it. Setting up your revenue attribution is one thing, keeping it running is another.
You now need to turn your focus towards the future, with the likes of support, monitoring, fixes, updates, and innovation.
In fact, there are a number of likely scenarios you will face in the (very) short-term, including:
Ensuring the dashboards are presenting the correct data.
You’re going to need systematic monitoring of the data that’s being pulled, i.e. that your algorithms from top to bottom are all doing exactly what they should be. Otherwise, any analyses and attribution your running will be as good as useless.
Changes in requirements
What if your CMO or growth leader changes and sets out new performance targets and metrics? Here is where the extent of automation built into your modelling and reporting becomes very useful. Unless automatic, not a question of pressing a button. More redoing.
Changes in commercial techstack
Say you buy another tool for your commercial techstack, or even replace your CRM, how is your system going to manage the changes necessary?
Troubleshooting
There’s the obvious problem of troubleshooting any errors that prop up across the system. A particularly important point to bear in mind is if there’s a break taking three weeks to solve, what happens with the tracking data then?
Innovation or improvements
Are you planning to run with the pack (or peloton if you’re into cycling) or are you striving for a leadership position where you’re creating enduring competitive advantage?
Sure, keeping up with trends and changes in the techstack might suffice, but to really draw continued maximum benefit, you will need to have some innovation.
How does an off-the-shelf platform do this?
A platform like Dreamdata will have all the support help you need for as long as you’re using the platform. This means, monitoring of data quality, troubleshooting and updates will be part of the package. An off-the-shelf platform like Dreamdata will also be constantly using all its resources to innovate and find better and faster ways of doing things. Guaranteeing an enduring competitive advantage in your scaling and growth activities.
Resources
Now that you have a clearer picture of which elements of data you’ll need to be collecting and what you want your algorithms to be doing, you need to assess whether you have (or want to have) the resources to go it alone.
Here we’re talking about how long it’s going to take to set up a revenue attribution solution, how many people you’re going to need to achieve this, and how much that is likely to cost.
Team?
So what should your data and engineering team look like?
As a bare minimum you’re going to need a full-time data scientist/analyst/engineer for the duration of the set up. Ideally, you’ll have a small team of two or three data engineers.
These need to have the following skillset:
Be competent in at least one data engineering language (generally either Python or Java)
Have ETL experience (typically with SQL)
Have strong data modeling skills
Have experience in monitoring data-intensive applications for errors
Business Intelligence acumen/ experience
You’ll also need someone, likely an Ops professional, with exposure to your present commercial techstack, (CRM, Customer Success, ads platform, etc.) to help steer the algorithm parameters of the project.
And finally, you’ll need access to someone who’s well acquainted with the revenue you’re generating.
If you haven’t yet got these bare minimum resources, you should think twice about initiating the process on your own.
Time-to-value should be the predetermining metric as you consider any product build vs. buy. How much you have to invest and for how long before you start deriving value, is critical to any decision-making.
And at this point, if you’re adding a possible recruiting process into the mix you are already tipping the scale overwhelmingly to buy.
Time?
If you do have in-house talent ready on the sidelines, a small team of two or three engineers dedicated to the attribution project will have it set up between 6 to 12 months.
Of course, your team resources and time are mutually interlinked; the more hands you have on the project the quicker it gets done. There is an important cost factor to consider if this is your course of action too.
Cost?
Speaking of cost, let’s look at some specifics.
An in-house team working on this, is, without doubt, the most expensive option. The average salary of a data engineer in the EU is around €62,000 per annum. Multiply that by the number of team members you have.
You also have the option of outsourcing the project. The cost of this can vary wildly, but will not fall much below the €50,000. Note also, any maintenance needing to be done after setting up.
By comparison, an off-the-shelf solution will cost you somewhere between €10,000 and €50,000 p/a depending on how much you rely on customisation.
You want to be keeping an eye out for any extra charges, such as access to data, consulting fees, etc.
Dreamdata, for instance, uses a fixed-pricing model where you get everything included.
Evaluation (TL;DR)
Ok, so what’s it going to be, Build or Buy?
From the article five things about building a B2B revenue attribution solution should be clear:
You need to collect as much data as possible from your tool and site;
Cleaning and merging the data is vital to successful revenue attribution;
A lot of attention needs to be paid to how you’re defining the parameters of your metrics;
You need to take into account how you’re going to visualise and use the reports and analyses;
Maintenance is essential to the success of your solution.
But what are the benefits of doing this on your own versus buying it from someone who’s already done it?
To make the assessment I’m going to use those assumptions I listed way back at the very beginning of the post.
Assumption 1 - You don’t need some of the features an off-the-shelf platform offers
Yes, there are some features from an off-the-shelf platform which you might not make use of - at least as far as you’re presently aware of.
Building your own means you would only have those features which you know that you need (at this stage).
However, as we learned in the post, these typically sit at the very top of the platform. That is, they are part of the final set of algorithms and reporting models.
Focusing only on these when building your solution in-house risks not handling the data properly at an earlier stage, e.g. when cleaning and merging your data.
This means that should you find out later that you do in fact want to run an attribution model you excluded initially, you will have to re-do almost the entire data processing and all subsequent steps again.
With an off-the-shelf solution like Dreamdata, you can simply choose not to use a feature without affecting the data collection and cleaning process. With Dreamdata all your data will be cleaned and merged irrespective of whether you end up using that data or not. Which means, should you decide to start applying a different attribution model in the future, you will immediately be able to run these without re-doing algorithms.
Assumption 2 - You have the resources to build a solution in-house
If you’re making this assumption it is likely that you have given some thought to what it will take to build the platform in-house.
However, from the post we can see that there are many details which can easily be overlooked. In fact, at Dreamdata we encounter an oversimplified understanding of what happens in an attribution tool quite often.
Many B2B CMOs, ops professionals, and growth leaders (those typically wanting the solution) do not realise just how complex revenue attribution is until they start the process of setting up a solution themselves - this includes setting up both bought off-the-shelf and built in-house.
As laid out in some detail in the post, it really isn’t just pulling data out of the tools and putting them into a dataset. The complexity lies in the sum of all the intricate individual steps in the process.
The devil really is in the detail. If you haven’t set about collecting the right data, or things aren’t being mapped in a correct fashion, or you are not carrying out the necessary maintenance of the platform, you lose the all important overview and the ability to derive maximum value from doing revenue attribution.
Assumption 3 - An off-the-shelf platform doesn’t give you the bespoke tailored solution you want
It is true, almost by definition, that an off-the-shelf platform cannot offer the same level of bespokeness that an in-house solution can.
But, when considering this factor you really need to understand precisely what bespoke solution you’re considering.
Some B2B revenue attribution platforms like Dreamdata, offer a very high degree of customisation and support. The best of off-the-shelf options offer a partnership solution with continual support to ensure the platform is adapted to your needs to the greatest degree possible.
You can limit the scope (and thereby the cost and risk) by utilising an attribution platform like Dreamdata to do the time consuming data crunching and use the resulting cleaned data to power your own BI or dashboard solution. In this way, you can build bespoke solutions tailored to your business in-house, without having to build up all the data infrastructure.
It is also worth noting the point made in the ‘maintenance’ section above. You need to weigh in the fact that an off-the-shelf platform like Dreamdata will be constantly using all its resources to innovate and find better and faster ways of doing things.
Creating this enduring competitive advantage (which is very unlikely to be done in-house) carries considerable value in its own right, especially if it factors in customisation.
In conclusion
If you’re considering building an in-house B2B revenue attribution solution, you need to give very serious thought to the business value of doing so.
When you build yourself you take a risk (time and cost) for something you do not yet know will bring value (or how much value).
An off-the-shelf platform like Dreamdata, can at the very least allow you to experiment with features to understand the value you’ll get before making the costly decision of building your own.
If you’re still unsure of whether to go it alone or buy an off-the-shelf solution, book yourself into a Demo call with our experts, who’ll be more than happy to help you make the best decision for you.
