Why CRM and Marketing Automation Systems Drift Into Disorder

Understanding CRM data entropy in complex marketing and revenue systems.

“These systems were working fine a few months ago. Why are they going wrong?”

This is a classic data systems question, and one that’s perfectly understandable for a senior leader to ask of their organisation.

It’s one I’ve grappled with from the first time I worked with big data sets back in 2009. It’s reasonable to expect your systems to function effectively and efficiently, right? 

A marketing automation platform, a CRM/Sales Operations platform, and the analytics/reporting platform should all function smoothly — like a car, right? Make sure you service it regularly, add oil, pump the tyres, and Bob’s your uncle.

But no.

One big difference with CRM systems is all the gremlins; They screw things up for you. I’m joking, of course. Gremlins were invented by engineers in WWII to explain strange behaviour in Air Force planes.

Rather like planes, with data you are not just dealing with systems; you are dealing with a lot of human beings. And humans do weird and wonderful, and often irrational things with systems.

In World War two, Gremlins were invented as a creative way for the Royal Airforce to maintain good morale during the Battle of Britain, and avoid playing 'the blame game' when beset with technical problems.

In addition, specifically when using current platforms, you are dealing with code, and coders. Each coder has their own way of coding. Not only that, but they often save the same chunks of code that they insert time and time again into various systems to fix issues. 

You also have coders increasingly using AI to write code for them. That can be difficult to monitor and control.

Your developers may be implementing numerous “workarounds” that are only temporary patches for long-standing issues.

Which brings us to the next response to that question:

“But was the system really working fine a few months ago?”

But how do you know for sure it was 'working fine' before? Could it have been patched up by developers in a way that worked fine at that time, only? 

Could large amounts of additional data, and additional systems (a new software integration, for example), have exposed deeper issues, cracks in the system?

In short, was it fully functioning, or was it functioning at that point with these workarounds? 

A good example of this is building a bridge. The bridge may appear perfectly sound under light traffic. 

But as traffic increases, small structural weaknesses begin to show — bolts loosen, joints flex, stress fractures appear. 

We have this exact same issue around the corner from my house, at Hammersmith Bridge - closed to all traffic except pedestrians and cyclists for the last seven years for these very reasons.

Secondly, the second law of thermodynamics states that all systems gravitate toward entropy. Meaning that even if a system is truly working 100% effectively (highly unusual with large integrated systems), it will naturally gravitate toward disorder and chaos.

If you think of several complex data systems all working together, they all rely on human input. It is partly a machine, yes, but it’s partly an organism.

Your systems most likely function more like a beehive or an ant colony than a high-performance automobile.

When the Hive Works

A beehive works great when all the bees know what to do, and everything is in order.

When the hive is healthy:

• Nectar flows in
• Honeycomb is built efficiently
• The colony stores food
• The queen lays eggs
• The hive grows
• The hive produces honey.

Information flows cleanly through the colony.

In CRM terms:

• leads are captured correctly
• Fields are filled consistently.
• Deals are updated accurately.
• Dashboards reflect reality
• The organisation produces revenue insight.
• The company grows revenue, and profits.

When the Hive Starts to Break Down

Now imagine something begins to go wrong in the hive.

• Some bees start bringing back the wrong pollen.
• Others forget where the nectar fields are.
• Some bees begin storing nectar in the wrong cells.
• A few bees stop communicating their waggle dances correctly.

Individually, these errors seem trivial.

But collectively they begin to destabilise the colony.

The CRM Equivalent

In CRM systems, the same thing happens.

Small behaviours introduce tiny distortions:

• A salesperson skips required fields
• A campaign uses inconsistent tracking parameters
• A deal is created manually rather than through the lead flow
• A Marketing Manager forgets to save lead scoring on a form
• A Data engineer makes an error in the dashboard logic
• A contact is duplicated
• Attribution fields are overwritten

Each individual action feels harmless.

But the system is cumulative.

Just as with the hive, thousands of small inconsistencies compound.

Data Entropy in CRM Systems

Going back to my entropy point, to ensure that your “beehive” is functioning well and driving all the “honey” you possibly can, you need to expend a lot of energy to keep everything in order.

1. Human Behaviour

Sales teams optimise for closing deals, not data quality.

Examples:

• Creating manual deals
• Editing deal values after closing
• Skipping mandatory fields
• Overwriting campaign sources
• Each action introduces micro-disorder.

Anecdotally, our top UK salesperson in the UK, at cybersecurity company Zscaler, rarely put information into our CRM platform, Salesforce.

When he left the company, we struggled to find the information we needed on some of our biggest clients, for which he had originally closed million-pound+ ARR deals. 

At Nielsen, and at Hansen, we often struggled with deal attribution between Sales and Marketing, since some of the relationships with sales teams went so far back, that it would perhaps be impossible to get a perfect read on this.

My point is that behind the 'clean' visuals on a dashboard, you often have very messy human interactions. 

2. System Complexity

Modern stacks involve many systems:

• Customer Relationship Management (CRM)
• Marketing automation
• Intent data platforms like Demandbase or 6sense
• BI tools
• Sales engagement tools, like Clay, Apollo, or Zoominfo.

Every integration increases the chance of:

• Mismatched IDs
• Sync failures
• Schema drift
• Attribution conflicts 

Complex systems amplify entropy.

3. Time

Even if nothing changes:

• Definitions evolve - How many times have the definitions of leads changed at your company (even subtle changes in lead scoring)?
• Teams change
• New fields get added.
• Legacy data persists
• Over time, the data model drifts from its original design.
• New, increasingly complex tech integrations are added
• CRM systems are not simply databases. They are living organisms.

They are socio-technical systems, a mixture of humans, incentives, processes, software, integrations, and behaviours unique to your organisation; Why working with two companies, using the exact same platforms, can be completely different.

People who constantly ask: “Why do these numbers not match?' are perfectly reasonable to do so. But behind the clean dashboard they see, any number of 'gremlins' can lurk.

Going back to the air force analogy - using a dashboard should be like flying a plane; It is one indicator you use, but it is not the only one (except in a time of dense fog, for example). 

Dashboards Are Instruments, Not Reality

Using dashboards to manage a business is a bit like flying a plane using cockpit instruments.

The dashboard provides important signals such as altitude, speed, heading, but it is not the sky, the weather, or the terrain itself. It is a representation of reality, not reality.

The same principle applies in organisations.

Dashboards can provide powerful insights, but they should never be treated as the entire picture of what is happening in the business. 

And just as in aviation, the most effective organisations combine instrument readings with context, culture, and human insight to understand what is really happening.*

The Reality for Leaders

So to sum up, it’s perfectly natural as a CEO, President of a division, a CMO, or a CRO to expect these systems to work efficiently, providing the functionality teams need, the actionable insights, and the ability to execute.

Senior leaders face so many demands that frankly blow my mind, and would likely overwhelm someone like me. Why shouldn’t they expect these systems to be the firm bedrock on which to build their business?

But the reality is that managing these systems requires constant vigilance, thought, testing, imagination, planning, and long-term strategy, just as the rest of the business does.

Below: Solving Data Disorder, Managing Organisational Culture

Because left unattended, complex systems drift toward disorder. 

How you manage this complex system will depend a lot on your 'problem-solving culture'. if you look at the matrix above, and according to Jim Collins, author of 'Good to Great' only about 5% of organisations sit in the ideal top right hand quadrant. 

But essentially, CRM/Marketing Automation/Analytics systems are no exception to the second law of thermodynamics - they drift into disorder, naturally. 

Like any complex system built on human inputs, software integrations, and evolving processes, they naturally accumulate entropy over time.

Which means the question is never whether disorder will appear.

The real question is who is paying attention when it does.

*I'd like to thank my friend, who is a leader in the financial services sector, as well as an Auxiliary Group Captain in the Royal Air Force, for providing both this outstanding insight, and with the aviation parallel to reading business dashboards.

The Real Edge of Private Equity: Active Ownership

I’m a big fan of Scandinavian thrillers, especially the original The Girl with the Dragon Tattoo. So when I walked into the auditorium at the London School of Economics, I had the strange feeling I was looking down at three lead actors from a Nordic noir drama.

The speakers were Ulf Axelson, Professor of Finance and Private Equity at LSE; Per Strömberg, Professor of Finance at Stockholm School of Economics and LSE; and Kurt Björklund, Founder and Executive Chairman of Permira, with roughly $50bn under management.

What followed was one of the clearest, data-driven explanations I’ve heard of why private equity (PE) ownership so often outperforms public equity, and where its limits lie.

Why Private Equity Outperforms: Start with the Data

The first half of the lecture was led by Per Strömberg and focused squarely on the evidence. Rather than starting with anecdotes or ideology, he began with productivity data across countries and firms.

In Germany, for example, fewer than 1% of firms accounted for roughly 65% of positive productivity growth over the period studied. Most firms contribute little. Some actively destroy value.

This matters because private equity does not rely on averages. Its entire model is built around identifying, creating, and scaling outliers.

       

The Mechanism: How PE Actually Creates Value

Strömberg argued that the performance gap between PE-owned and publicly listed companies is not primarily due to regulatory arbitrage or tax advantages, though those exist at the margin.

The core driver is active ownership.

Drawing on both academic literature and operating evidence, PE value creation can be grouped into three broad mechanisms:

1. Governance engineering

PE owners are not distant shareholders. They:

• Sit on boards
• Hire and fire management
• Set incentives tightly linked to value creation
• Intervene early when performance slips

This sharply reduces classic agency problems between owners and executives.

During my MBA at Northeastern, one of my finance professors specialised in corporate governance, and I conducted research on shareholder activism. One theme emerged repeatedly: public-company executives often optimise for personal incentives that diverge from shareholder value.

Below: PE-owned companies are rigorous in selecting customers that add value

PE ownership compresses that gap. In the same way that active shareholders hold senior leadership to account, Private Equity owners can step in to ensure the company is run as efficiently as possible. 

Per explained that the productivity and efficiency gains of Private Equity ownership (according to him, 2-3% higher than Public Equity, according to Kurt, head of a PE firm, it is closer to 6-7% higher), can be divided into three key categories:

Three types of engineering/tools

1. Governance engineering – being an active owner of the company

2 . Financial engineering – reduce cost of capital 

3. Become sector experts – can leverage networks to assist management

Well, that begs the question – why don’t other companies copy the behaviour of PE companies, to improve their performance?

According to Strömberg, this opens a “can of worms”.

First, PE performance may not be indefinitely sustainable. Funds have finite holding periods, typically six to seven years. Active ownership delivers diminishing returns once the biggest inefficiencies are removed.

However, within that limited time frame, PE seems to be doing better than ever. Exit value experienced a rebound in 2025, increasing 41 per cent to $1.3 trillion, the second-highest year on record. 

Second, PE capital is more expensive. While leverage can be cheaper than equity, the cost of financial distress rises sharply as leverage increases.

PE is not a universal solvent. It is a precision tool, effective under specific conditions.

An Operator’s Perspective: Kurt Björklund of Permira

The second half of the session (unrecorded) shifted from data to practice. Kurt Björklund described himself not as a financier, but as a “financial entrepreneur” and "Sector disrupter".

His framing was revealing.

Public equity investors, he argued, suffer from information asymmetry. Even large shareholders rely on periodic disclosures and carefully curated narratives.

PE ownership, by contrast, is built on information abundance:

• Proprietary KPIs
• Weekly operational interaction
• Direct access to management and systems

Björklund was blunt: unlike asset managers such as BlackRock, he cannot afford to be wrong. Every investment must succeed. That forces extraordinary diligence and relentless focus post-acquisition.

He also warned about classic PE pitfalls:

• Buyer’s curse in auction processes
• Cyclicality of capital markets
• The temptation to “take your eye off the ball” during exit processes

“In my business,” he said, “only the paranoid survive.”

Disruption, People, and the Role of AI

One of the most charged parts of the discussion came during the Q&A, where students (from the LSE, Imperial, Oxford, and Berkeley, USA) repeatedly asked about AI and job security. There were also several questions from analysts at various Private Equity firms.

Björklund acknowledged the anxiety, but did little to soothe it.

He described investments in complex B2B businesses where agentic AI, and improved automation have reduced headcount by orders of magnitude, particularly in areas such as KYC and compliance.

In one example, automation reduced a team from 5,000 people to 500, while increasing profitability. Many in the organisation were conducting relatively complex tasks, which could nevertheless be performed more effectively with AI and Automation.

Above - Top Target Universities (non-US) for Goldman Sachs. Source: Krugman Insights

His view was unsentimental: there will always be jobs for the very best, and the traditional path: An elite education, a top investment bank such as Goldman Sachs, and then a good Private Equity firm, remains viable. But the middle is being hollowed out.

Interestingly, he noted that older employees often adopt AI more effectively than younger ones, attributing this to psychological barriers to AI in younger workers. 

Perhaps it's also because you need deep experience in solving the problems, to ask AI the right questions? It's very easy to generate 'AI workslop' that drives no insight, and diminishes your credibility in the organisation. And that is no doubt from whence that fear emanates.

The recording was switched off halfway through the lecture, at which point the atmosphere in the room changed perceptibly. Kurt (The Chairman of Permira) smiled and said he would assume there were no journalists present, which meant he could now speak a little more freely than usual.

The professors, clearly enjoying the moment, joked that in Sweden, Kurt is known as “Superkurt”: the complete package: handsome, physically fit, wildly successful, and extremely wealthy.

Kurt laughed, didn’t deny it, and carried on.

Which confirmed something I’ve learned from working with private equity firms in the past: there is remarkably little self-deprecation in the room, even when the person in question is a typically reserved Swede.

Joking aside, this was one of the best lectures I've seen, unique in that it presented top-level insights from both the academic and 'real-world' perspectives.