Anchoring Effects In Planning Poker Estimation

The Problem of Cognitive Biases in Estimation

Estimation is a crucial activity in software development, allowing teams to gauge the level of effort and resources required to complete work items in a project. However, human judgment in estimation can often be skewed by systematic cognitive biases. One such bias that plays a major role is the anchoring effect.

Anchoring occurs when people rely too heavily on an initial piece of information, the “anchor”, when making decisions or estimates. This anchor value gets fixated in the mind, coloring future judgments and hindering the ability to adjust away from that initial anchor. In software teams, historical data, arbitrary numbers, or even unrelated conversations can serve as anchors that influence the estimates that emerge from planning activities like planning poker.

The anchoring effect manifests in planning poker when the initial anchor value suggests an estimate, and estimators insufficiently adjust away from that starting point even as new information emerges. This anchor exerts a stronger pull than is rationally warranted, leading to estimates that tend to cluster around that initial number. The insidious nature of anchoring lies in estimators remaining unaware of just how much anchors infect sound judgment and accurate estimation.

Table of Contents

Mitigating Anchoring Effects

The good news is that once aware of how anchoring works, teams can proactively mitigate its influence through deliberate strategies. Anchoring tendencies still persist, but their impact can be actively minimized.

The first step is to simply be cognizant of anchoring as a factor that introduces bias. Once estimators know about the mechanism of anchors disproportionately influencing estimates, it allows them to watch for its effects. This meta-cognition supports better assessment of what factors are driving estimates besides just anchoring.

Teams can also use an estimation scale with wide bands, for example estimating in buckets of 5 story points instead of 1 story point increments. This helps counter more precise anchoring on spurious numbers. Wider bands account for inherent uncertainty in estimates while allowing adjustment between buckets.

Another effective technique is to have team members generate independent estimates before sharing and discussing as a group. Independent estimation avoids collective anchoring on the first stated estimate. The subsequent estimation conversation also benefits from already having multiple estimates based on personal assessment rather than reactive responses.

Anchoring remains a challenge though when there are no historical benchmarks or completed precedents to anchor against for a task under consideration. In such cases, teams can leverage empirical data from industry studies on software development effort as the basis for anchoring estimates.

Examples of Anchoring in Planning Poker

While awareness of anchoring can mitigate its effects, certain manifestations still commonly emerge in planning poker activities:

Having a starting estimate that seems reasonable but is actually too high or too low for the work item. This exerts an anchoring effect on all subsequent estimators.
An estimator being influenced by a previous estimate they heard for a separate but related item earlier in the session.
The first estimate stated aloud setting an anchor that pulls others as they inadequately adjust from that starting point.

A basic example is a story point estimate starting either too low or too high due to poor initial anchoring. A task that should have an estimate of 3 story points has the first estimate stated as 8. That implicitly anchors others to a higher estimation scale, and the final estimate settles around 5 story points rather than the more accurate 3.

Anchors can also carry across sessions, with historical estimates and velocity metrics providing anchors that become outdated. Or during a poker planning session, a separate but related user story estimated earlier may provide a faulty anchor for the story currently being discussed.

Estimators often remain unaware of such anchoring issues, falsely assuming their estimate is based on an impartial assessment of the work rather than anchored on situational priming. Careful attention is key to noticing when anchoring prevents appropriate adjustment from inaccurate initial anchors.

Strategies for Reducing Anchoring in Planning Poker

Some potential techniques teams can employ to counter anchoring effects in planning poker include:

Openly discussing anchoring tendencies – Bring awareness of anchoring so estimators actively mitigate its unconscious influence
Allowing changes to initial estimates – First stated estimates are particularly vulnerable to poor anchoring
Using different anchors across the team – Increase variance of possible anchors to diminish bias towards any one anchor
Estimate using ranges vs precise points – Ranges account for inherent uncertainty better while resisting specious precision from anchoring

Having an open discussion about the reality of anchoring allows teams to externalize the issue and consciously watch for it. Calling out situations where anchors might exert undue influence raises overall awareness.

Also, by allowing estimators to change even their initial estimate as new information comes to light, it prevents the first stated estimate from fossilizing due to anchoring effects. This gives estimators freedom to recognize poor anchoring.

Another technique is to purposefully introduce variability into the possible anchors used by having estimators select different starting points for each new estimate. This reduces the chance that the team will anchor on any one number.

Finally, estimating using a range instead of speciously precise story points accounts for the intrinsic uncertainty in estimates while also avoiding anchoring on overly exact numbers.

The Path Forward

Though eliminating anchoring entirely may be impossible due to its rootedness in human cognition, continual improvement towards unbiased estimation remains possible. Progress requires embracing that estimators have biases requiring mitigation, not just through increased rigor alone but also better understanding of psychological dynamics like anchoring. Some future directions include:

Ongoing education on the various cognitive biases that shape human decision-making and estimates
Carefully tracking actuals to detect areas where anchoring or other biases may be affecting estimates
Optimizing planning poker with facilitation techniques and ground rules to specifically counter anchoring

Embedding deeper literacy about cognitive biases into engineering cultures allows teams to promote greater self-awareness. Complementing this perspective with data tracking also gives the means to notice distortions leading to issues like rework due to poor estimates. And purposefully enhancing planning poker dynamics facilitates building team estimation ability over time. Together these measures represent progress towards addressing systematic estimation biases like anchoring.