An Evidence Scorecard for Comparing Indiana Agency Proposals

Give each number a passport before granting it credit

A number needs a definition, population and collection method before it can support a purchase. We would check those features before discussing its apparent size.

These checks can sit alongside an existing scorecard without adding more numerical weights. We would treat a missing definition as an open question first. A candidate may be able to answer it from the source. If the answer stays unavailable, keep that limitation attached to the claim instead of filling the gap with a confident assumption.

Our purpose is to make unlike statements readable on their own terms. A public screenshot, a client-approved summary and direct account access provide different kinds of support. None automatically supplies every fact needed for a purchase. Ask what each shows and what still depends on the person explaining it.

Read the calendar before naming a trend

Period labels can hide overlap, partial days and changed coverage. Our first comparison would inspect the actual dates before treating totals as growth or decline.

1. Write the exact boundaries Use the source dates rather than the export filename alone. We would check whether the intervals share days and whether each has the same type of coverage. A label such as recent can mean different things in different tools.
2. Normalize only a sound comparison Dividing a valid count by covered days can help describe unequal intervals. Our check would still ask about season, page age and changed business activity. A daily rate does not remove those differences or make a small sample stable.
3. Keep averages attached to their weights Do not subtract one average position from a cumulative average to invent an earlier rank. We would ask for the underlying compatible data. If rounded daily means are combined with proper weights, label the reconstructed result as approximate.

What if an agency shows only the date a file was saved?

Ask for the dates of the activity itself. We would keep the save date as a separate fact. It can tell us when someone made the file, but it cannot tell us which days produced the count. A picture can be current while the data inside it describe a much older period.

Should a single busy day be removed from the calculation?

Keep it in the observed record unless there is a sound reason to exclude it. Our extra check would show what the pattern looks like with and without that day. That tests how much the claim depends on it. A burst can be real and still give little help in predicting a normal week.

Leave unseen traffic unassigned

A missing table row does not tell you where unreported activity belongs. We would preserve gaps and ask about collection limits before accepting an allocation.

Can a query missing from an export be called worthless?

No. First determine whether the export is complete for that question. A tool can display a selected or limited set of rows. Our concern would be a proposal turning absence into proof that customers never use a phrase. Ask what the source can omit and whether a targeted view can answer the narrower question.

Can page totals be forced to match a property chart?

Keep each total at its reported level until the reason for a difference is established. We would ask about aggregation, filters and the platform's own counting rules. Do not spread an unexplained remainder across pages or assume a favorite explanation. A neat reconciliation built from guessed allocations is weaker than an honestly marked gap.

Can several page counts be treated as exclusive shares?

Only when the counting method supports that interpretation. Multiple pages may be represented in a view without dividing into mutually exclusive audiences. Our questions would ask what a page impression means in that specific report. Do not infer that a small difference between a site total and one page belongs entirely to its siblings.

How should repeated or damaged rows be handled?

Name them before calculations and keep the original file available. We would ask whether repeated dimensions represent separate groups or an export problem. Adding them together or keeping the last row can change the result. If the distinction cannot be resolved, exclude the affected comparison and say why, rather than repairing it through an undocumented guess.

An export is an object with limits, rather than a complete copy of customer demand. Our questions would cover row limits, hidden categories and grouping choices before drawing a commercial lesson. The proposal can still use the available observations. It should not claim that the missing part behaves like the rows that happened to appear.

When a candidate names a limitation, ask what would reduce it. That might be a filtered export, another date view or an agreed check of the original source. We would distinguish a practical next check from a new promise that every gap can be filled. Some uncertainty may remain part of the purchase.

Keep customer wording separate from landing page ownership

Query and page tables answer different questions unless a filter connects them. We would require that connection before assigning a search phrase to a given page.

1. Read each dimension on its own A query row describes the phrase under the report settings. A page row describes the URL under its settings. We would first note those two facts separately. Similar wording, matching city names or coincidentally equal click totals cannot connect the individual records.
2. Ask for the relevant filtered view For a page-ownership claim, inspect the pages associated with the selected query under a stated period and audience. Our check would retain the query filter and any other settings. A general list of popular pages does not answer that narrower question.
3. Separate overlap from proven harm Seeing several URLs for a phrase establishes an association within that report. It does not, by itself, show that one stole clicks from another. We would ask what supports the proposed consequence and whether the data contain the time detail needed to study it.
4. Match the remedy to the unanswered question A clearer internal link can help readers choose a page without establishing a ranking result. Our first action would remain limited to the problem actually found. A destructive page consolidation needs a stronger basis than similar titles or the presence of several related URLs.

A proposal can make a sensible content recommendation while its ranking effect remains unknown. We would ask the writer to use those two statements separately. Clearer buyer routes can be checked directly on the site. Future search behavior requires later observation under consistent conditions, and a sale requires still more details from the business.

Build a contribution case without claiming sole cause

Marketing can help an outcome while staff, pricing and demand also matter. Our check would trace the proposed contribution without giving one activity all the credit.

• Ask what changed outside the agency task during the measured period. Our questions would include service availability, price changes and how quickly staff handled requests. Those factors do not erase the work. They affect how confidently the result can be carried into a new proposal.
• Keep the customer stages distinct. A click can lead to a message, and that message can be unsuitable or unanswered. We would ask which stage the claim measures and whether later stages were actually recorded. Renaming every event a lead can hide the part the business needs to improve.
• Check the counted population when a project covers several sites or offers. Our interpretation would preserve that group and its observation dates. A result from a collection of properties should not become a forecast for each proposed page or a promise for a new city.
• Ask whose work the example describes. A founder may have worked on a project as an employee, through an agency or with another team. We would keep the role and the named task attached. A genuine contribution can be described accurately without borrowing responsibility for the whole outcome.
• Treat a customer comment as an account of that person's experience. Our check would ask what the quote actually addresses. Praise for communication cannot by itself verify a revenue claim, and a review of one service should not stand for an untested service introduced later.

A contribution argument should explain the path from changed work to a plausible customer effect. That explanation can include uncertainty. We would ask which part is directly visible, which part is supported by records and which part remains a reasoned expectation. Keeping those layers separate makes the proposal easier to question without dismissing every positive observation.

A comparison group can help when it is relevant and measured consistently, but its name alone does not establish fairness. Ask what differs between the groups and why those differences matter. Our reading would consider whether the proposal has shown a plausible contribution, rather than treat the existence of a comparison as proof that all other explanations disappeared.

Check what a technical demonstration actually proves

A technical example is strongest when its conditions are visible. We would name the demonstrated behavior before accepting a broader claim about future customers or search results.

Can an attractive screenshot stand in for a working test?

It can show appearance at the time captured. We would still ask what happens after the visible button is used. A screenshot cannot establish message delivery, account access or the behavior of an error path. Credit the visual work for what it shows and ask for the separate checks needed by the proposed purchase.

What if the technical result cannot be reproduced?

First name the changed conditions. Our questions would cover the tested version, access, provider state and the device used. A difference may have a normal explanation. If the result still cannot be explained, mark the related claim open rather than accepting it because the demonstration was confident.

Google describes normal crawlable links in terms of the anchor element and its destination attribute. That supports checking how a route is built. Our use of that guidance would remain tied to a specific defect. It should not become a promise that changing a link will make a business rank for a competitive phrase.

Compare unlike purchases without blending their units

A repair, a new website and ongoing campaign work solve different purchasing questions. We would keep those units separate before comparing their price or claimed value.

• Compare deliverable units with compatible deliverable units. Our questions would distinguish an edit to supplied copy from collecting new material, checking it and publishing it. The same page label can hide a very different workload and a different demand on the client.
• Separate one-time changes from continued operation. We would ask what keeps working after the purchase and which tasks depend on another arrangement. A recurring fee should be assessed against the work it actually covers, rather than treated as either wasteful or comprehensive by default.
• Keep potential benefit separate from included work. A proposal may aim to improve discovery or reduce confusion, but those goals do not name the purchased tasks. Our check would ask what can be accepted directly and what must be observed after the work is in use.

A proposal with fewer moving parts may fit a narrow problem well. A broader engagement may be justified when several teams or channels must change together. We would judge the connection between the tasks, rather than reward either smallness or breadth by itself. Ask what would break if a proposed activity were removed from the purchase.

What if one offer has more pages but less help after launch?

Ask how the business would use each page and who would keep it accurate. We would count the work the buyer must take on as part of the comparison. More pages can mean more upkeep. The larger package may fit the need, but the page count alone cannot show that it is the sounder purchase.

What if a broad package includes a task already done well?

Ask why that task needs to be repeated. Our concern would be paying to replace a working part without a clear gain. The answer might involve a real dependency, or it might show that the package needs to change. A buyer can ask for a narrower offer without rejecting the rest of the agency's skills.

Keep open claims visible when the scores look close

A close score should lead to a focused question about the purchase. We would carry open claims forward instead of smoothing them into a tidy total.

What if the proposal and the demonstration disagree?

Name the specific difference and ask which version governs the proposed work. We would keep the answer with the claim it affects. A sample may be outdated or drawn from a different arrangement. That can be explained, but the buyer still needs to know what is actually being offered before treating the sample as support.

What if a claim is true but irrelevant?

Give it credit only for the question it answers. Our check would ask whether the demonstrated skill helps with the present constraint. A strong video portfolio can matter to a video purchase while leaving website access or page maintenance unanswered. The scorecard should not let a genuine strength erase an unrelated gap.

What if the candidate cannot share private proof?

Ask what can be shown with permission or through an anonymized method demonstration. We would respect that boundary while retaining the limit on verification. Privacy does not require the buyer to treat an unsupported number as confirmed. The proposed work may still be a good fit for other reasons that can be inspected.

1. Restate the open claim Write the exact statement the buyer is being asked to accept. Our concern would be a precise question turning into a general debate about whether the agency seems trustworthy. Keep the missing fact small enough that someone can answer it.
2. Choose the next check Name the source, permission or explanation needed. We would ask whether that check can reasonably be completed before a purchase or whether the uncertainty must remain. Do not invent a result merely to finish the comparison.
3. Record the effect on the purchase State whether the gap changes the work, the allowed authority or the willingness to proceed. Our conclusion might be to narrow the task, seek another answer or defer it. A comparison is allowed to end without selecting an agency.

The final record should preserve disagreement where it matters. A buyer may accept one uncertainty and reject another because the practical consequences differ. We would make that reasoning visible without pretending the score is a scientific measurement. It is a way to organize questions and the reasons for a purchase.

Apply these tests to our proposals too. Ask us to explain the metric, the dates and the part our work is intended to change. That conversation can establish what has been shown and what remains a future test. It should not turn a forecast, a platform limit or a hoped-for customer outcome into an agency performance claim.