The Map That Was Accurate When It Was Drawn — In 1987.
Why the customer your brand is personalising for — right now, with sophisticated data infrastructure — may not have existed for two years.
The consultant's report arrived in a leather binder. Forty-three pages. Every intersection in the city mapped. Every commuter pattern documented. Every bottleneck identified, prioritised, and assigned a recommended intervention.
The city council had commissioned the study in 1987. Eighteen months of fieldwork. Real data. Real traffic counts. Real commuter surveys. The consultant was qualified, methodical, and thorough. The report was the most comprehensive traffic analysis the city had ever received.
The council approved the routing recommendations and began implementation.
Three years later, the new signage system was directing traffic through a flyover that had been demolished in 1991 — the study had begun before the demolition order was issued. It was routing morning commuters around a park that was now a shopping complex, adding fourteen minutes to journeys that the original study had designed to take eight. It was diverting vehicles away from a ring road that had opened in 1993 and now carried 40% of the city's daily traffic — because in 1987 that road did not exist, and the study had no mechanism for incorporating infrastructure that appeared after its data collection phase ended.
The consultant had done nothing wrong.
The data was accurate.
In 1987.
The council had done nothing wrong.
The implementation was faithful to the report.
The report described a city that no longer existed.
The Mechanism That Explains Why This Keeps Happening
In 1974, two psychologists named Amos Tversky and Daniel Kahneman published a paper in Science magazine that identified one of the most consequential errors in human judgment. They called it anchoring bias.
Anchoring bias refers to the tendency to rely too heavily on the first piece of information encountered when making decisions. The mechanism works in two stages. First, the mind establishes an anchor — an initial reference point, a starting value, a first impression. Second, it adjusts from that anchor as new information arrives. The adjustment is almost always insufficient. People consider the anchor as a starting point and adjust away from it until they reach a zone of uncertainty — resulting in insufficient adjustment because adjusting away from the anchor is a conscious and effortful process.
The anchor is sticky. It resists replacement. Even when new information is available, the mind does not start fresh — it modifies the anchor. The modification is smaller than reality requires.
The city council's anchor was the 1987 report. Every subsequent piece of information about the city — the demolition of the flyover, the conversion of the park, the construction of the ring road — was processed relative to that anchor. The adjustments were made. They were insufficient. The map was updated in small ways while the city changed in large ones.
Now look at the anchor your marketing system set for every customer at the moment they first appeared in your database.
Their email address. Their job title. Their company size. Their location. Their stated preferences. Their first purchase. Their demographic profile.
That is the anchor. Every communication decision your brand makes about that customer adjusts from it. The adjustment is insufficient. The customer has changed. The anchor has not.
The council was routing traffic through a demolished flyover.
Your personalisation engine is routing campaigns to a person who changed their job two years ago.
The Rate at Which Your Map Becomes Wrong
B2B contact data decays at an average rate of 22.5% per year. Nearly one quarter of every customer record becomes inaccurate within twelve months of collection.
70.8% of B2B contacts change their job title, company, or professional responsibilities within a single year. The email address may still be valid. The person who filled in the form — with that title, at that company, with that budget authority — may not work there anymore.
Email addresses themselves change at 30% per year in high-churn professional categories. Phone numbers at 25% annually in mobile-first markets.
But these are the visible forms of decay. And visible decay is not the problem.
The Data That Is Wrong and Working
This is the insight that the standard data quality conversation misses entirely.
A hard bounce is visible decay. The email fails to deliver. The system generates an error. The record is flagged or removed. The brand knows the data is wrong. The map shows a demolished flyover and the system cannot route the car there — the error is immediate and obvious.
The dangerous decay is invisible.
The customer whose email address still delivers. Who still opens your emails occasionally. Who appears in your engagement metrics as a marginal but present subscriber. Whose CRM record shows them as a Senior Marketing Manager at a mid-size enterprise in Bangalore — because that is what they were when they filled in the form in 2022.
They are now a VP of Digital at a different company in a different city, with a completely different budget authority, a completely different relationship to your product category, and a completely different set of communication preferences.
Your email delivers to them every Tuesday. They open it approximately once a month. Your system registers them as an active subscriber. Your personalisation engine addresses them by their 2022 title at their 2022 company with content calibrated to their 2022 needs.
The data is wrong. The data is working. Nobody knows.
This is the map that shows a park where there is now a shopping complex. The car does not crash. The navigation completes. The driver arrives at a destination that is not what they came for. They leave. Nobody flags the error because the journey technically succeeded.
The invisible decay is the decay that compounds silently inside your database for eighteen months before it appears in your conversion rate as a decline nobody can explain.
Because the anchor set in 2022 is still governing decisions in 2024. And the adjustment from that anchor — in the way Tversky and Kahneman predicted — has been insufficient.
Four Dimensions of Customer Data— Ordered by How Visibly They Decay
Dimension 1 — Identity Data (Most visible decay)
Name, email address, phone number, company, job title. This is the data that decays fastest — 22.5% per year — and most visibly. Hard bounces signal email address decay. Returned calls signal phone number decay. LinkedIn profile changes signal title and company decay.
The visibility of this decay creates a false confidence. Because the identity data that fails visibly gets cleaned, the identity data that does not fail visibly is assumed to be current. The email address that still delivers is assumed to still reach the same person in the same context. It may not.
The anchor: who this person was when they gave you their details. The reality: who they are now.
Dimension 2 — Behavioural Data (Moderately visible decay)
Purchase history, browsing patterns, content engagement, channel preferences. This data decays more slowly but the signals of decay are easy to misread.
A customer who purchased annually for three years and then stopped is not the same as a customer who never purchased. But your behavioural model may treat them identically — classifying both as non-purchasers, applying the same suppression or re-engagement logic, missing that one represents a relationship that ended while the other represents one that never began.
The anchor: the behavioural pattern at peak engagement. The adjustment — insufficient — treats a relationship that has changed as one that simply paused.
Dimension 3 — Contextual Data (Least visible, most consequential)
Life stage, professional circumstances, financial situation, family structure, organisational context. This is the data that governs why someone buys — and it changes continuously in ways that are almost entirely invisible to your database.
The customer who bought baby products three years ago has a three-year-old now. Their household has a different income allocation, different time constraints, different priorities. Your anchor shows a new parent. The reality is a parent navigating the questions that arise at age three. The content aimed at the anchor is irrelevant to the reality.
The customer whose organisational champion — the person who originally approved your product — has moved to a different role. The budget authority has shifted. The renewal conversation is now with someone who has no context for the original purchase decision. Your CRM anchor shows a warm account. The reality is a cold one waiting to churn.
Dimension 4 — Attitudinal Data (Invisible decay, most dangerous)
How the customer feels about your brand. Their level of trust. Their perception of your category. Their openness to communication from you.
This data was never captured at scale. It exists nowhere in your system. But it changes — continuously, driven by every experience they have with your brand and your category and the channels you use to reach them.
The customer who opted in enthusiastically in 2022 may have developed, through thirty months of Tuesday emails, a specific conditioned response to your sender name. Not hostility. Something quieter. The habituated non-response that Post 01 in this series described. Their attitudinal anchor from 2022 said: this brand is worth hearing from. Their 2024 attitudinal reality says: this notification does not require attention.
Your system has no record of the change. The anchor from 2022 still governs the frequency, the channel, the content. The adjustment — as Tversky and Kahneman predicted — has not happened.
The Personalisation Paradox
There is a specific failure mode that only occurs when personalisation technology meets decayed data. And it is worse than no personalisation at all.
When a brand with no personalisation sends a generic email — the customer receives a message that makes no claim to know them. It is impersonal. It is slightly less relevant than it could be. But it does not signal anything about the brand's understanding of who the customer is.
When a brand with sophisticated personalisation sends a highly targeted email — built on a customer profile that is two years out of date — the customer receives a message that claims to know them and is wrong. It addresses them by a title they no longer hold. It references products relevant to a company they no longer work for. It offers content calibrated to a life stage they have moved past.
The message does not just fail to land. It signals actively that the brand has been paying attention to a version of this customer that no longer exists. That is more unsettling than being unknown.
Being confidently wrong about who someone is produces a deeper trust failure than not knowing them at all.
This is the personalisation paradox that anchoring bias creates. The more sophisticated the personalisation system — the more precisely it uses the acquisition anchor to construct targeted content — the more damage it does when the anchor has diverged from reality. The scalpel is more dangerous than the butter knife when pointed in the wrong direction.
And the anchor diverges from reality at 22.5% per year.
What Updating the Map Actually Requires — Three Practical Actions
The antidote to anchoring bias is not better data collection at acquisition. Tversky and Kahneman were clear on this — the anchor established first is the most persistent, regardless of its quality. The antidote is a systematic process for replacing the anchor at regular intervals using current signals.
Action 1 — Score every record by field-level age, not record-level age
A record created in 2020 may have had its email address validated in 2024. It may have had its job title never updated. These are two different records for the purposes of personalisation — one field is current, one is a four-year-old anchor.
Build field-level freshness scoring into your database. Every field should carry a last-verified timestamp. Every personalisation decision should query field freshness before applying field value. A job title that has not been verified in 18 months should be treated as unknown — not used as the basis for targeting decisions.
This is the equivalent of adding the demolition date to the city map's legend — not just recording what was there, but recording when it was last confirmed to still be there.
Action 2 — Add a re-permission trigger at 18 months
Any subscriber who has not actively updated their preferences, changed their engagement behaviour meaningfully, or responded to a re-permission request within 18 months is operating on an anchor that is approaching two years old.
A re-permission communication — one that acknowledges the passage of time, asks three specific questions about current context, and makes it easy to update or exit — serves two functions simultaneously. It refreshes the anchor with current data for the subscribers who respond. It surfaces the subscribers whose silence indicates the anchor has been wrong for longer than 18 months and they simply have not had a reason to signal it.
This is the new city survey. Commissioned at regular intervals. Not because the previous one was wrong at the time — but because the city has kept moving since it was drawn.
Action 3 — Watch for behavioural discontinuity as a decay signal
A customer who engaged consistently and then went silent is not the same as a customer who never engaged. The discontinuity — the point at which engagement changed — is the signal.
Engagement discontinuity should trigger a contextual review rather than a re-engagement campaign. The question is not "how do we bring them back?" — it is "what changed in their context at the point of discontinuity?" The answer often lies in life stage, organisational change, or category shift that your anchor did not capture.
The customer whose engagement dropped in the same month they changed companies is not disengaged. They are different. The campaign targeting the old anchor will not re-engage them. The campaign that acknowledges the gap and asks what has changed — might.
When the City Commissioned a New Survey
The city did eventually commission a new traffic study.
Not because anyone admitted the 1987 report was wrong. Because enough routing failures had accumulated — enough commuters arriving at demolition sites, enough journey times diverging from projections — that the evidence of anchor divergence was impossible to ignore.
The new survey took the 1987 report as its starting point. Not because the starting point was reliable — but because the field teams needed an anchor from which to note every change. The demolition. The conversion. The new ring road. Every change was documented relative to what the 1987 map had shown.
The final map was not the 1987 map corrected. It was a new map. Built on a new survey of the current city. Using the old map only as a reference for what had changed and by how much.
That is the update your customer database requires.
Not a correction of the existing records. A new survey of the current reality — using the acquisition data as a reference for change, not as the governing truth.
The customer you captured in 2022 is the starting point for understanding who they are in 2024.
They are not the answer.
The map was accurate when it was drawn. The question is whether you are still navigating from it — or whether you commissioned a new survey when the city kept moving.
If your brain is already triaging this page for a 5-second window, skip the reading—the complete narrative is perfectly laid out in the infographic below.
Published by Hetvabhas — independent analysis of brand communication.
No vendor agenda. No sponsored content. No false reasoning.
