DIMOND, Justice.

Appellant's merchandise, consisting of groceries, houseware items, clothing, drugs and miscellaneous stock, was damaged by soot and smoke in a warehouse fire in Fairbanks. The total net loss to appellant was in excess of $38,000. In this suit against insurance companies to recover for the loss, the trial judge sustained appellees' position that they were responsible for insurance coverage only to the extent of $2,500. Appellant maintains on this appeal that the judge's determination was clearly erroneous. The first question raised by appellant relates to the scope of our review of the trial judge's findings.

SCOPE OF REVIEW

Civil Rule 52(a) provides in part:

Findings of fact shall not be set aside unless clearly erroneous, and due regard shall be given to the opportunity of the trial court to judge of the credibility of the witnesses.

We have applied the clearly erroneous standard of review many times. In doing so we have frequently pointed to the trial judge's opportunity to observe the demeanor of witnesses and judge their credibility, as affecting the scope of our review of his findings.

In this case no witnesses appeared at the trial. The case was tried entirely on the basis of written depositions, documentary evidence, and a statement of undisputed facts. Since demeanor evidence was not involved, appellant contends that we are in as good a position as the trial judge to evaluate the evidence, and therefore are not bound by the clearly erroneous standard of review.

Earlier decisions of this court suggest a basis for the practice urged by appellant. In Paskvan v. Mesich and Fairbanks Publishing Co. v. Pitka it was held that where nondemeanor evidence was involved, this court was in as good a position as the trial judge to determine issues of fact. In the more recent case of State v. Phillips, we held that "where the trial judge's findings are based on nondemeanor sources, such as documentary evidence, deposition testimony, or transcribed testimony, our scope of review is broader than under the clearly erroneous standard."

In these few decisions we have certainly indicated that where findings are based on nondemeanor evidence, our scope of review is different from the situation where findings may be based on the trial judge's assessment of witnesses' credibility, which he makes from hearing and seeing them in person. But we have not yet articulated precisely what the standard of review is where findings are based on documentary or other nondemeanor evidence. We do so now.

The portion of Civil Rule 52(a) involved here is identical with a procedural rule in the federal judiciary relating to findings of fact made by the United States District Courts. In interpreting that rule, where findings are based upon documentary evidence or undisputed facts, it is said that the federal courts of appeals are "indescribably confused", and that "[e]ven within a single circuit, decisions vacillate inexplicably from one test to another."

Some federal courts have taken the position that where the trial judge did not see the witnesses, the appellate court is in as good a position as the trial court to interpret the evidence, and will "more readily" declare the trial judge's findings to be clearly erroneous. This is what has been referred to as placing a "gloss" on the rule.

Other federal appellate courts have gone further and held flatly that they are not bound at all by the clearly erroneous standard where the evidence was not oral, and that they are free to completely disregard the trial judge's findings and make for themselves a de novo review of the evidence. The decision most frequently cited in support of this approach is that of Judge Jerome Frank in Orvis v. Higgins. Judge Frank stated that if the trial judge decides a fact issue on written evidence alone, the appellate court is as able as he to determine credibility, and could disregard the trial judge's finding. Judge Frank then went on to formulate different degrees of freedom of review depending upon the proportion of the evidence that was oral and the proportion that was written, including the effect that the written testimony may have had upon the credibility of the oral testimony. This so-called "Frank" position on standard of review has been recommended as proper by Professor Moore in his work on federal practice and procedure.

Then there are the courts that take the position that the words "findings of fact shall not be set aside unless clearly erroneous" mean what they say, and that the clearly erroneous test applies to all nonjury cases, regardless of the nature of the evidence involved, whether oral, written, or both. The writers principally espousing this view are Judge Clark, one of the draftsmen of the Federal Rules of Civil Procedure, and Professor Charles Alan Wright in his revision of Barron and Holtzoff's work on federal practice and procedure. It is pointed out by Professor Wright in 2B W. Barron and A. Holtzoff, Federal Practice and Procedure, section 1132, at 522-23, that the United States Supreme Court took this view of Rule 52(a) in its 1948 decision in United States v. United States Gypsum Co., and has reiterated that position in later decisions.

The pertinent language of Civil Rule 52(a) states that "[f]indings of fact shall not be set aside unless clearly erroneous * * *." A plain English reading of this language leads to only one conclusion — that the clearly erroneous standard of review applies to all findings of fact, regardless of the nature of the evidence they are based upon. The additional words, "and due regard shall be given to the opportunity of the trial court to judge of the credibility of the witnesses", form a conjunctive, and not a hypothetical proposition, as Professor Wright has pointed out. The rule does not establish the clearly erroneous standard of review only in those cases where the trial judge has had the opportunity to judge of the credibility of witnesses. This standard of review applies to all findings. The provision regarding the credibility of witnesses means only that when there has been oral testimony, and the trial judge has observed the witnesses in person, we must pay some deference to his judgment as to credibility to the extent that his findings are based on such oral testimony. This makes sense, because until such time as we review cases, perhaps on the basis of videotaped proceedings, we cannot have the advantage that the trial judge has had of basing a judgment as to credibility on the demeanor of the witnesses that appear before him.

In a case where there has been no oral testimony, the rule still provides that "[f]indings of fact shall not be set aside unless clearly erroneous." We cannot agree with the proposition that in such a case the clearly erroneous standard has no application, and that the trial judge's findings may be simply disregarded. We must "regard" the judge's findings, in the context of the evidence presented, in order to form a considered judgment as to whether such findings were not just erroneous, but clearly so. The question is not whether the clearly erroneous standard applies, which it obviously does, but what it means so far as our appellate review jurisdiction is concerned.

In determining the meaning of "clearly erroneous", it might be helpful to start with what it does not mean. We do not agree with the concept that we may not substitute our judgment for that of the trial judge where conflicting inferences may be drawn from the evidence by reasonable men, and the trial judge's findings are based on inferences which a reasonable man might have drawn. This is basically the rule applied in jury cases to determine whether a particular issue must be submitted to the jury for determination, or whether it should be decided as a matter of law by the judge. It is applied in such cases because of the historic right to trial by jury as it existed at common law, which is preserved by our constitution.

We believe such a rule should not be applied to our review of a trial judge's findings. There might be several competing inferences upon which reasonable minds could differ, which could be drawn from documentary evidence or undisputed facts. The fact that the trial judge chooses one of these inferences does not necessarily mean that we could not consider his choice clearly erroneous.

Nor do we agree with the suggestion that a finding which is supported by substantial evidence may not be clearly erroneous. This is the rule we apply on judicial review of administrative findings. But this is so because of our recognition of the respective functions of administrative agencies and the superior court, and of the deference the courts feel constrained to show to findings made by such an agency charged by law with the making of factual determinations in a particular area within the scope of executive power.

To limit our setting aside of a trial judge's findings to instances where there is no room for difference of opinion among reasonable men, or where there is no substantial evidence to support such findings, would too narrowly limit the scope of our review. A broader scope of review is, to use the words of Professor Moore, "a natural and proper concomitant of appellate power." Our constitution provides that ."[t]he supreme court shall be the highest court of the State, with final appellate jurisdiction." We shall not adopt a construction of Civil Rule 52(a) which could result in an undue narrowing or abdication of the "final appellate jurisdiction" vested in this court by the constitution.

At the same time we must be aware of the need to exercise what one might call judicial self-restraint. Our function and our desire is, of course, to do justice. We could extend this function and desire to the point of setting aside any findings of a trial judge which we considered wrong because we might have made different findings in reviewing the same evidence he did.

But if we view our scope of appellate power too broadly, there is the danger that in our desire to do justice in a particular case, actual overall injustice may result. As Professor Wright has pointed out in a thoughtful study on this subject, as the scope of appellate review becomes broader, less respect and confidence is had for the trial courts, and the number of appeals increases. The cost of increased appeals — in the way of delays in finally ending litigation, and in the way of further congesting already crowded trial courts with new trials — cannot be discounted.

Appellate judges are not the only trustworthy ministers of justice. As Judge Chase observed in his dissent in Orvis v. Higgins:

Though trial judges may at times be mistaken as to facts, appellate judges are not always omniscient.

We must be careful not to broaden our scope of review under the illusion that we are infallible, or under the misconception that trial judges in general are not as capable of doing justice as appellate judges are. There is hardly any limitation on the scope of our "final appellate jurisdiction", except as we exercise self-restraint and place our own limit upon such jurisdiction. We must be always careful, because of the corrupting influence of great power upon human nature, not to confuse our desire to "do justice" with an inordinate pride in our position and function as the "highest court of the State." We must always maintain, as it has been said, "a lively sense of our own defects".

As is the case in so many instances of formulating rules of law, our task is to achieve a proper balance — in this case, between our desire to do what is right and just according to the facts in a particular case, and the need to give recognition to the function of and to maintain confidence in the superior court, the trial court of general jurisdiction in this state. We know that we cannot achieve a perfect balance. As we said recently in Whitton v. State, the problem involved cannot be solved by the easy application of a mechanical formula. An approximate solution, with the hope that the needs of justice in a broad, overall sense will be best served, is the most that we are able to accomplish.

The rule we adopt in reviewing findings of fact by a trial judge is substantially the same one we have utilized in the past, which we first adopted from the United States Supreme Court decision in United States v. United States Gypsum Co. The rule is this: Under Civil Rule 52(a) we shall not set aside the finding of fact of a trial judge unless it is clearly erroneous. A finding is clearly erroneous when, although there may be evidence to support it, we are left with the definite and firm conviction on the entire record that a mistake has been committed.

This rule applies to any finding, regardless of the nature of the evidence upon which it is based. The only difference between our review of findings based on oral testimony, and those based on documentary evidence or undisputed facts, is that in the former case we must pay some deference to the trial judge's assessment of the credibility of witnesses, whereas in the latter case we need not. It is because of the deference we pay to the trial judge's assessment of credibility of witnesses where there is oral testimony that we have characterized our scope of review in cases where there is no oral testimony as being a "broader" type of review. Even in the latter situation, clear error must appear under the rule we apply.

The clearly erroneous standard, as we apply it, means something more than merely showing it is more probable than not that the trial judge was mistaken. We must be convinced, in a definite and firm way, that a mistake has been committed. We must be well persuaded by the party seeking to set aside the trial judge's findings before we will hold he was wrong.

THE FINDINGS OF THE TRIAL JUDGE

At the time of the damage, appellant's merchandise was covered by policies from four insurance companies. Seventy-five percent of the coverage was with Birmingham Fire Insurance Company and National Surety Corporation, and was handled through the J.C. Morris Agency, at Fairbanks. The remaining 25 percent of the coverage was with American Manufacturer's Mutual and United States Fidelity and Guaranty Company, and was handled by the Fairbanks Insurance Company, also at Fairbanks.

The insurance policies were of the "provisional" or "monthly reporting" type. Under this type of policy an annual premium is paid. The amount of the premium is flexible, depending on the value of the inventory reported each month by the insured as being on hand at the insured location.

The mechanics of the arrangement are that initially 75 percent of the premium is paid based on the full value of the policy. This payment is subject to increase or reduction at the end of each year based on the monthly reports of the value of merchandise located at the insured location. In the event of fire damage, the amount that the insurers agree to pay is based on the value of merchandise stated on the last monthly value report submitted by the insured prior to the loss.

The main area of dispute between the parties is over what monthly value report was the last submitted by appellant prior to the loss of February 23, 1963. A report for September and October 1962, stating the value of merchandise to be $2,500, was received by each of the insurance companies involved well in advance of the loss. Another report for November and December 1962, stating the value of merchandise at the insured location to be $25,000, was also submitted by appellant. But this report was not received by the insurance companies until after the loss.

The superior court found that the November and December 1962 report was not effectively submitted by appellant until after the loss, and therefore that the coverage at the insured location at the time of the loss was controlled by the $2,500 report of September and October 1962. Appellant asks us to set aside this finding.

The pertinent evidence relating to this finding was as follows:

(1) Alaska Foods kept no records of monthly reports of value.

(2) Bonnie Goes was hired as office manager of appellant on January 14, 1963. She testified that she prepared the $25,000 November and December 1962 value report under the supervision of appellant's accountant, Mr. Doremus. She said that in preparing this report she used value figures supplied by Mr. Carr, owner of Alaska Foods. Bonnie Goes also testified that she assumed she had mailed the November and December 1962 report to the J.C. Morris Agency in January 1963, because she had no reason to hold it.

(3) Mr. Carr's testimony indicated that he would not have had detailed knowledge of the contents of the records since he had had them prepared by office personnel.

(4) Appellant's accountant, Mr. Doremus, testified that Bonnie Goes had been instructed generally to make out insurance reports, and that he had assumed that monthly reporting was continuing. He testified that appellant's records were in such poor condition that reports to management, including information relating to inventory, could not be prepared. According to Mr. Doremus, this situation had not been remedied at the time Bonnie Goes was employed on January 14, 1963.

(5) Mr. Baxter, manager of the J.C. Morris Agency, and Mrs. Morris of that agency, testified concerning the practice of the agency with respect to forwarding the value reports to the insurance companies after the agency had received them. According to their testimony, all of the monthly value reports were handled the same way. The agency would mail the report to the insurance company on the same day it was received from the insured. Usually the report would be taken from the insurance envelope and immediately placed in a preaddressed envelope which had been provided by the insurance company involved.

From these basic facts the judge drew the following inferences:

(1) It was doubtful that Bonnie Goes actually obtained direct supervision from Carr and Doremus in making the November and December 1962 value report.

(2) It was doubtful that Bonnie Goes would have been able herself to prepare the necessary report.

(3) The November and December 1962 report was not mailed by appellant until after the loss for the following reasons:

(a) Bonnie Goes' testimony that she prepared and submitted the report to the J.C. Morris Agency in January 1963 was probably incorrect.

(b) The insurance companies did not receive the November and December 1962 report until at least about two weeks after the loss.

(c) It was established that the J.C. Morris Agency's business practice was to forward to the appropriate insurance companies value reports received from an insured immediately after receipt.

Appellant contends that the inferences and findings of the trial judge are erroneous because the business practice of the J.C. Morris Agency was not adequately shown by the evidence, and that even if it was, the practice followed in this case did not correspond to that business practice.

The authorities cited by appellant, such as Lieb v. Webster, 30 Wn.2d 43, 190 P.2d 701 (1948), concerning what must be shown by way of business custom and compliance therewith to establish mailing are not pertinent here. The superior court did not rely on the J.C. Morris Agency's business practice to establish the fact of mailing. This was sufficiently established by the insurance company's receipt by mail of the value reports. The business practice was relied upon to show that the J.C. Morris Agency promptly forwarded the value reports to the insurance companies, and that the Agency did not let the November and December 1962 reports lie around its office for several weeks which might have accounted for the late receipt of the reports by the insurance companies.

Appellant argues that because certain typing appeared on the November and December 1962 report between the time it left the insured's hands and the time it arrived at the insurance companies, someone at the J.C. Morris Agency must have placed the typing on the report. Although this may indicate a slight deviation from J.C. Morris' usual practice of transferring the value report immediately from the insured's envelope into the insurance company's envelope, this does not necessarily indicate any departure from the business custom of mailing the report to the insurance company on the same day it was received from the insured.

The testimony of Baxter and Mrs. Morris established a sufficiently regular and uniform business custom. The circumstances appearing in the present case are sufficiently similar to that custom to make the business practice relevant evidence. The trial judge could properly have considered this evidence in making his findings.

Appellant also argues that the trial judge erred in finding that the $25,000 November and December 1962 report was not filed prior to the loss, because there was evidence other than that relied upon by the judge which points to the conclusion that this value report was filed before the loss. We have examined appellant's contention, but we need not go into a detailed discussion here of the evidence relied upon by appellant. The inferences which appellant wishes us to draw from this evidence may be legitimate and not unreasonable. But at the same time, the trial judge's inferences which he drew from the evidence in this case, and which led to his ultimate finding, are also legitimate and not unreasonable.

In viewing the entire record, there may be some evidence to support appellant's point of view. But that is not enough to persuade us to upset the finding of the trial judge. The inferences he drew from the evidence are legitimate and his ultimate finding is supported by such inferences. Even though the inferences and finding of the judge are based on nondemeanor evidence, and even though there is some evidence to support appellant's contentions, we are not left with the definite and firm conviction on the entire record that the trial judge made a mistake in finding that the November and December 1962 value report in the amount of $25,000 was not filed prior to the loss. Therefore, we cannot hold that finding to be clearly erroneous.

OTHER ISSUES

There are two other issues to be decided. One concerns the doctrine of waiver, and the other, the question of whether the September and October 1962 value report in the amount of $2,500 was a result of clerical error.

Waiver

The facts show that appellant had on some occasions been delinquent in submitting its monthly value reports. On these occasions the insurance companies would initiate action through the local agencies to obtain the delinquent report from appellant. The insurance companies never threatened cancellation of the policies because of delinquent reporting. The question raised by appellant is whether the insurance companies, by accepting the late value reports without objection, waived their right to rely on the monthly reporting and related provisions of the insurance policies.

Some cases hold that there can never be a waiver of monthly reporting and related provisions of a provisional insurance policy. These cases proceed upon the theory that the monthly reporting provision is one going to the coverage of the policy, and that conditions going to coverage or scope of a policy of insurance, as distinguished from those furnishing a ground for forfeiture, may not be waived.

We agree that the monthly reporting provision of a provisional insurance policy like the one here being considered is a condition going to the coverage of the policy. Failure to comply with the monthly reporting condition through tardy filing results not in a forfeiture of the policy, but in a limitation on the coverage of the policy determined by the last report filed. However, we need not decide the broad question of whether there can ever be a waiver of an insurance provision going to coverage because, in our opinion, mere request for an acceptance of late monthly value reports would not constitute such a waiver under the terms of the provisional insurance policies in this case.

Provisional insurance, like that involved here, is designed to provide insurance coverage for fluctuating stocks of merchandise at a minimum of expense. The insurance policies require the filing of reports for each month within a certain time limit, and specify that if a report is not filed within the time limit, the coverage of the policies is determined by the last report filed.

Regardless of when the reports are filed, the policies give both parties the right to a value report for each month. The insured is entitled to file a value report each month to determine his insurance coverage. Although the policies state that the intent of the insurance is to provide full insurance coverage, there are clauses in the policy which make it clear that the insured can elect to insure less than his full inventory at a consequently lower premium computed in accordance with provisions of the policy. The insurance company has the right to rely on the monthly report to determine its liability in the event of loss and the premium owed it under a provision of the policy for premium adjustment.

To constitute a waiver, the acts of a person must evidence "an intentional relinquishment of a known right or privilege." An insurance company's act of requesting and accepting delinquent monthly reports alone does not evidence such an intentional reliquishment of its right to determine its liability with reference to the last report filed. The policies require value reports for each month. The company has the right to request and accept value reports for each month, even if such reports are delinquent, in order to make more meaningful the company's right to determine the premiums with reference to such reports under the provision of the policy regarding premium adjustment. An insurance company may want to request delinquent reports as a way to remind the insured to file its reports on time. Whatever the reason, merely requesting and accepting delinquent monthly value reports does not show that an insurance company intends to waive the value reporting and related provisions of the provisional insurance policy.

Furthermore, the provisions of the policy indicate that it is designed to be flexible, enabling the insured to elect how much insurance he will carry from month to month, and providing for adjustment of the premium accordingly. If the insured files a timely report stating the value of his inventory as some fraction of its true value, the policy dictates that his insurance will cover the same fraction of any loss which occurs. Under the provision requiring the filing of value reports, the insured can elect a fixed dollar amount of coverage by filing his value report late.

If we held that acceptance of late value reports waives the value reporting provisions of the policy and reinstates full coverage up to the policy limits, we would be creating an entirely new contract for the parties. The flexible nature of the policy indicates that the parties intended that late value reports could be filed as often as desired by the insured, and that the result of this would not be full coverage up to the policy limits, but coverage limited to the amount stated on the last, although tardy, value report filed.

The cases cited by appellant are distinguishable because they do not hold that mere request for and acceptance of late value reports constitute a waiver of the value reporting provisions of a provisional insurance policy. There were additional and distinguishing facts present. In Columbia Fire Ins. Co. v. Boykin Tayloe, Inc., 185 F.2d 771 (4th Cir. 1950), the insurance company's agent would obtain the necessary information and prepare the monthly value reports which he then often submitted late to the insurance company. The insured had been led to believe that the agent could handle all of the reporting details for it and that it was fully complying with the reporting provisions of the policy by merely supplying inventory information to the agent when requested. The insured was never notified that its value reports were being filed late. Furthermore, the insurance company knew and approved of the agent's reporting arrangement. These circumstances are clearly different from an insurance company merely requesting and accepting late monthly value reports from the insured.

In Schenley Distillers Corp. v. United States Fire Ins. Co., 90 F.2d 633 (2d Cir. 1937), the time limit within which monthly valuation reports had to be filed with the insurance company was omitted from the insurance contract. The court intimated that in this situation the insurance company would have to notify the insured to file its report before the report would be deemed late. This is clearly a different situation from that in the present case.

Like the Schenley case, General Ins. Co. v. Killen, 270 Ala. 604, 120 So.2d 887 (1960) and Aetna Ins. Co. v. Rhodes, 170 F.2d 111 (10th Cir. 1948), both cited by appellant, involve situations different from the present case. In Aetna the court, faced with a situation different from that here, assumed that the evidence at trial established a waiver and did not consider what conduct by an insurance company could amount to a waiver of the reporting requirements. 170 F.2d at 115. In Killen the issue involved was the effect of the filing and acceptance of value reports with no specific address given for the insured goods.

We hold that an insurance company's request for and acceptance of late monthly value reports, by itself, cannot constitute a waiver of the reporting and related provisions of the insurance policies in this case.

Clerical Error

The September and October 1962 value report, which we have decided governed the insurance coverage in this case, stated the value of the merchandise as being $2,500. That this value figure was the result of a clerical error has never been disputed during this appeal, in the briefs or at oral argument. Appellant asks us to correct this clerical error and place the parties in the position in which they would have been had no mistake been made.

As we have indicated, the terms of the policies permit an insured to file value reports late. The risk to the insured in this procedure is the resulting fixed amount insurance coverage determined by the last late value report filed. The premiums are determined with reference to that amount. The policy clearly does not contemplate that an insured generally be allowed to upgrade the value stated in a late report after a loss has occurred and after this value has proven to provide insufficient insurance coverage. This would open the door to fraudulent schemes involving under-reporting of values.

However, in the present case we need not worry about fraudulent schemes, since there is no dispute over the fact that the $2,500 value reported was the result of an honest clerical mistake. Although the relief asked by appellant may be in the nature of reformation, and although reformation generally will not be granted for a solely unilateral mistake, we can perceive no policy reasons in this case for not correcting the mistaken value reported.

The insurance policy was intended to provide full insurance coverage to the extent desired by appellant. There is no dispute over the fact that appellant intended to and thought that it had reported a value of $25,000, and that the $2,500 value reported was the result of an honest clerical error. Appellant thus desired to have insurance coverage to the extent provided by a $25,000 value report, and the purpose of the policy weighs in favor of granting such coverage.

There is no evidence that the insured was involved in any fraudulent scheme involving the filing of late under-valued reports. The parties can easily be restored to the positions in which they would have found themselves had there been no error by an increase in coverage corresponding to the $25,000 amount which was intended to have been reported in the September and October 1962 value report, and a corresponding adjustment of the premium owed by the insured in accordance with the policies of insurance.

The judgment is reversed with directions to the superior court to base appellees' liability on the September and October 1962 value report corrected to report a $25,000 value. The increased premium incurred by appellant as a result of this correction is to be deducted from the appellees' liability.

ERWIN, J., not participating.